PIZZA看加密

This summer, PayPal’s board received a cash acquisition offer. After reviewing it, they felt the price was too low and sent it back. The buyer didn’t walk away right away—negotiations dragged on for more than a month. In the end, they decided not to raise the price and simply ended the talks. The news broke on Thursday night, and by the next session the stock opened and immediately plunged.

$PYPLB is trading on Binance’s spot market at $51.97 right now, down 15.63% over the past 24 hours. At one point it was smashed down to $50.07—an 8:08 reading.

The consortium of Advent and Stripe offered $60.50 per share, implying a valuation of more than $53 billion. The board thought this number undervalued the company. Today’s trading price landed even lower than that figure.

The first half of August shows the expectations even more clearly. On August 15, PYPLB touched $64.54, trading above the bid. People who entered back then were betting that the consortium would add more. That bet has been cleared today.

So the remaining question is: apart from the M&A premium, what is this company worth by itself? The board said it’s above $60.50; today’s price says $51.97. That gap of about 16% has to be found in the financial statements.

I went through the July 28 earnings report from top to bottom.

Revenue was up 5% year over year, while total payment volume was up 10%. The money flowing through PayPal’s pipe is still growing at double-digit rates, but the portion the company keeps for itself is only growing at single digits.

Go one layer deeper and it becomes clearer. Divide transaction revenue by total payment volume: in Q2, the transaction fee rate was 1.61%, versus 1.68% in the same quarter last year. Seven basis points doesn’t sound like much—until you spread it over a quarterly flow approaching $500 billion, and it’s not pocket change. In the same quarter, transaction profit in dollars was up only 1%. After stripping out interest on customer balances, it was up 3%. When volume grew 10%, the gross margin the company actually pocketed rose by only 1%.

PayPal handles more money year after year, yet each dollar leaves less behind as time goes on. Operating profit margin keeps sliding, and non-GAAP EPS fell back slightly year over year. On the same day, management raised its full-year guidance, pointing non-GAAP EPS to $5.38—up a little from $5.31 last year.

That “little bit” can be traced directly on the books. Over the past twelve months, the company repurchased about 111 million shares, reducing the float by nearly 6% over the first half of the year. The numerator barely moved, the denominator shrank quickly—so EPS barely held steady.

Buying back your own stock with free cash flow is a reasonable move at this price. In the first half, free cash flow was $2.678 billion—this business is still making money. Based on today’s price, the market capitalization corresponds to fewer than 10 times this year’s earnings guidance.

Buybacks can solve for per-share numbers, but they can’t fix every “payment.”

The company’s own answer is hidden in the restructuring on April 29. Lores took over as CEO from Alex Chriss on March 1, and the board’s reason for the management change was that execution was too slow. In his first month, he split the company into three parts. One of them is called Payment Services & Crypto, housing Braintree, merchant processing for SMBs, and PYUSD. Digital assets got an independent slot for the first time within #PayPal .

When the cut on the button becomes harder to defend, you go toward the pipe and the settlement layer—where you can earn from clearing, float, and cross-border money. Stablecoins serve as settlement tools on this route.

Where that path goes can be seen on-chain. Today I directly pulled up the contract: total PYUSD on Ethereum is about 1.8 billion coins; on Solana it’s about 680 million. On a whole-chain basis, it’s more than 2.7 billion dollars. The peak on March 5 was 4.2 billion—down about 35% from this year’s high.

If you stretch the timeline, the shape matters more than the drawdown. Last September, PYUSD across all chains was still only a little over 1.1 billion. By December it reached 3.8 billion. A stablecoin that surged more than threefold in three months doesn’t come from merchant settlement demand behind the scenes. That period corresponds to a round of high-interest reward programs. Once rewards were reduced, the money left.

So this year’s supply dropped by one-third. You can’t directly read it as PayPal’s payments business shrinking. A lot of what rose earlier was rented. What I care about more is whether it can keep climbing on its own when there are no rewards.

After the news broke, Thomas Hayes of Great Hill Capital publicly applauded the board, saying you can’t let the buyer take away upside from existing shareholders. His rationale was that the $60.50 offer is still less than nine times free cash flow. The company should continue executing independently, including buybacks.

By his math, the offer really isn’t expensive. My disagreement is in the premise. Nine times free cash flow might be cheap for a payments company with stable gross margins, but for a company whose fee rate declines by seven basis points every year, it may not be. Put a “cheap” multiple onto gradually thinning gross margin, and the outcome might only be that it falls more slowly. The consortium is essentially saying the same thing. After reading the financials, they decided not to raise the price—suggesting that in their model, the company can’t support a higher number.

I’m not on the “value trap” side either. The buyback intensity and the cash flow are real. Venmo was singled out as an independent business unit. Reading it, I felt like they were paving the way for a spin-off or sale. If that route works, it’s the second way to make up the discount.

What the board is betting on is that its transformation can outperform the decline in the fee rate—and the money for that bet comes from shareholders. Today’s share price is the market’s price for that bet.

I’ll watch three things. In the Q3 report in late October, whether the decline in that seven-basis-point transaction fee rate narrows. Whether the growth rate of transaction profit in dollars can rise above 3% once interest is stripped out—showing the new money starts moving faster than the payment flow. And finally, the PYUSD supply curve when there’s no reward campaign; anyone can check the on-chain data themselves.

Of the three, two have turned. The board scrapping that offer is the right call. All three are still mostly the same; the $60.50 that was pushed away might end up being the best price this company will see in the coming years.

If you have a position—or you just want to see whether this transformation can work—you can note down the single line about the transaction fee rate from the Q3 report. You can also save a copy of the PYUSD on-chain supply curve; it updates faster than research notes.

When it comes to making Bitcoin quantum-resistant, several different paths suddenly emerged this week, and they are completely different in how they’re carried out. The first one to go viral is also the easiest to understand: Bitcoin’s rules don’t change at all—repeatedly re-compute the signature for a payment until a quantum computer can’t do anything about it, and then hand the transaction straight to the miners for packaging.

The premise is a bit awkward: ordinary nodes simply don’t relay transactions in this format—you have to knock on a mining pool’s door yourself.

This transaction landed in block number 964,199. The scheme was designed by Avihu Levy of StarkWare, and his colleague Tomer Giladi routed it through MARA’s Slipstream channel. I pulled it apart on-chain and checked it: the input protected by the quantum-resistance mechanism is 10,000 sats, which at today’s price comes to less than eight dollars. Levy has explained the principle: the wallet doesn’t accept the first valid signature it computes. Instead, it keeps generating candidate signatures over and over until it finds the one with the right shape—this process consumes several hours of computing power.

Eight dollars, several hours. Pulling this comparison out isn’t meant to mock; it actually shows what this proof demonstrated. Under Bitcoin’s consensus rules today, a spend that does not leak a public key throughout can indeed be packed into the main net.

The scheme’s boundaries are something StarkWare itself describes more honestly than the people who circulated it. The company states clearly that QSB did not make Bitcoin quantum-resistant. It protects only certain transactions. Specifically, it only works for addresses whose public keys have already been exposed on-chain; it can’t save others. CEO Eli Ben-Sasson puts it even more plainly: he still wants Bitcoin to do a soft fork, and he believes it will happen in the end.

The trouble is right here. Bitcoin’s quantum risk isn’t distributed evenly across all coins—it is concentrated only on the subset of addresses whose public keys are already written into blocks. An address that has spent money belongs to this category. As of March 1 of this year, more than one-third of all Bitcoin on the network has already exposed public keys on-chain. These coins won’t become automatically safe just because there’s a new way to spend them. Either the owners actively move them, or they just sit there indefinitely waiting. QSB is a tool prepared for the portion that hasn’t exposed public keys; that’s exactly the group that most needs rescue, but it can’t reach them.

The second path goes after this gap. In the same week, Blockstream’s Jonas Nick formally published BIP for SHRINCS. This is the first quantum-resistant signature scheme specifically tailored by trimming according to Bitcoin’s ledger structure. The foundation is still SHA-256, and it doesn’t rely on any new mathematical assumptions. The cost is made explicit: today a Schnorr signature is 64 bytes; SHRINCS has a minimum of 548 bytes, and in the worst case it can grow to 4,619 bytes. It also requires state: with the same private key, each time you sign the key grows a bit longer. If the device is lost, you need a fallback transaction of more than 5,000 bytes to recover the funds. In the BIP document, there’s also a line that has not been deleted: the security proof hasn’t been completed.

The third path goes even further. In the same week, Blockstream also released an evaluation of lattice-based signatures. Falcon-1024 is the most space-efficient in that category: public key plus signature together totals 3,073 bytes. But the research team itself didn’t recommend deploying it right now, and NIST’s standard text isn’t finalized yet. Their suggested order is: use the hash-based approach first, and only consider a hybrid once the Falcon standard is settled.

Once the routes are laid out, the shared point can’t be hidden. Besides QSB, the other two paths require changing the consensus layer. QSB doesn’t because it bypasses the entire P2P network: nodes don’t recognize such transactions, so miners have to receive and package them separately. Engineering-wise, Bitcoin today isn’t short of answers—it’s short of someone who has the authority to decide for the one-third of coins.

This contradiction has already been brought to the surface this year. In February, BIP-360 was merged into the official repository, defining Bitcoin’s first quantum-resistant address type. In April, Jameson Lopp and five other developers published BIP-361, setting a five-year sunset period for old signature types. Coins that haven’t been moved by the deadline will no longer be recognized by the network as spendable, including the batch widely believed to be Satoshi’s. Adam Back is explicitly against forced freezing; he argues that quantum-resistant functionality should be made an optional feature now, so people can move their own coins. The most accurate summary came from Marin Ivezic, who works on post-quantum security; he said the true constraint for Bitcoin’s quantum migration isn’t cryptography—it’s governance.

I agree with that judgment, and this week’s news provides a perfect footnote. Cryptographers have finished the multiple-choice part: QSB is what can already be used; SHRINCS is what can be brought into the protocol; and Falcon is the option that saves space. The remaining controversy no longer belongs to technical selection—it’s whether to set a deadline for the coins owned by some people. Bitcoin’s governance structure is capable of adding features—Taproot is proof of that. But when it comes to taking away rights, it has never succeeded. The original design was meant to block exactly this kind of thing.

This judgment can be falsified. Over the next few months, if BIP-360 or SHRINCS enters substantial activation discussions and rejects the kind of signaling schedule Taproot used that year, the governance bottleneck might not be as stuck as I imagine. Another signal could be even more direct on-chain: if large addresses that haven’t moved in more than ten years—and whose public keys are already exposed—start relocating in bulk, then the debate about freezing versus not freezing will automatically be downgraded. Neither of these has happened yet.

And there’s no need to be scared by this week alone. The market’s starting point for quantum anxiety is late March and June. Google Quantum AI improved the resource estimation for Shor’s algorithm on elliptic curves by an order of magnitude, and Justin Drake’s long write-up spread it widely in the community. Resource estimation improvement is not the same as actually building machines—the former only shifts the timetable forward a bit. #Bitcoin is now 79,891, and overall this week it’s still moving upward, basically unrelated to the quantum timeline. The value of $BTC is still running along with macro factors and the ETF schedule.

If you want to do something for yourself these days, you can check whether the commonly used address you control has spent funds on-chain. If it has, that means the public key is already exposed, and later—no matter which path Bitcoin chooses—the addresses that will need to be moved proactively are exactly this kind. It’s still far from that day, but knowing which side you’re on is more useful than remembering which week someone proposed which scheme.

$SOL beat the broader market. Today, both BTC and ETH are just grinding sideways, while the one that really moved was the one that “walked out on its own.” In the Chinese community, people credit the US spot ETF with the win. That explanation isn’t wrong, but it leaves out another thing happening on-chain on the very same day. Solana validators are voting, and the item being cut is exactly the kind of thing ETF buyers have been coming for.

In the past 24 hours, SOLUSDT is up 7.22%, while BTC and ETH over the same period are only up by a little more than that. On-chain, Solana is, for the first time, moving a complete governance process on-chain: three proposals are being voted on at the same time, with all firepower focused on SGP-0002. It aims to raise Solana’s annual discount rate from 15% to 30%, doubling the speed at which inflation comes down. The endpoint stays the same, but the time to reach it is brought forward by three years. The accompanying SGP-0003 remakes transaction fees: it splits them into one payment as a bundling fee to block proposers, and another resource fee charged based on computational consumption and fully burned.

These two things collide in the same week, but they point in opposite directions. As of August 26, the US spot SOL ETF’s cumulative net inflows are $1.26 billion, a record high. Of the nine products, Bitwise’s BSOL alone captures 77% of the cumulative net inflows. Morgan Stanley’s MSOL has a lower fee rate than BSOL, yet the money still went first to BSOL. It won by being the first to pass through on-chain staking rewards to holders. The money follows staking rewards, not much to do with the issuer’s branding. The ETF buys yield; what SGP-0002 is meant to do is to push that yield lower.

The number most often cited in the promotional talking points is that 18.9 million fewer SOL will be issued over six years—sounds like a supply shock. Based on current protocol inflation, Solana issues about 64,000 new SOL per day. Over six years, the “missing” 18.9 million SOL is actually less than even the amount of additional issuance in just the past ten months. The proposal documents are more honest than the people reposting them: their stated scope is 2.6% below the current issuance schedule. Don’t overestimate the “burn” line either. Right now, the entire network burns only a little over 600 SOL per day. Based on the proposal’s own calculations, once the new rules run, that could reach 7,500 to 9,000 SOL. Sounds like a huge multiple, but set against more than 60,000 SOL of daily additional issuance, it’s still a small fraction.

Both sides argue without ambiguity. On August 14, Helius’s mert publicly said that some so-called stakeholders have motives to profit for themselves by diluting coin holders through increased issuance; he believes this whole argument doesn’t hold water. Standing on the other side is Solana Company, listed on Nasdaq. In the second quarter, 99.4% of this company’s revenue came from staking rewards, and on August 21 it announced opposition to SGP-0002. Of course that stance has self-interest, but the issue it points to is real. If this cut goes through, the first pain will be felt by validators and institutions that survive on staking income. The proposal document itself also admits that under the new table, there will be a group of validators dropping earlier into a non-profitable position. Another listed Solana treasury company, SOL Strategies, took the opposite route: its four validators all voted in favor, with three votes each. Same kind of company that makes its living off SOL, one opposes and the other supports— the difference is that the former sells staking rewards, while the latter sells validator services. Grayscale’s estimate lands somewhere in the middle. They think that if both Ethereum and Solana’s token-economics changes in this cycle are implemented, Solana’s annual inflation would be pushed to just over 1% around 2031, scarcity would increase, and staking returns would move down in tandem.

I support this proposal, but I don’t buy the promotional spin. Using increased issuance to pay staking participants’ modest nominal gains is essentially an internal transfer payment among token holders. Non-stakers get diluted; stakers get it back. The network doesn’t actually gain any new value as a result. Pushing this curve down is the right move. Changes on the supply schedule are limited; the pressure will transmit to the demand side. Once yields move lower, the 77% BSOL share will face a test: when that money originally came in, it bought SOL or it bought yield? I’m inclined to believe buying SOL has a larger share, because on August 26, the cheaper MSOL’s single-day net inflows already surpassed BSOL. Fees and channels themselves are also at work. After the accelerated discount rate takes effect: if the BSOL share keeps falling while total inflows don’t, then my view holds. If total inflows collapse along with the yield, then I was wrong.

There’s also another risk: the voting rules for this round are fighting each other. The Governance FAQ requires that one-third of the network’s staked participation is present, and that two-thirds of the votes cast in favor must be “yes” for it to pass. In the proposal repository, it says there is no participation threshold: as long as the proportion of yes votes out of yes plus no votes is two-thirds, it passes; abstentions don’t count. In the August 26 snapshot, the yes votes are close to seven times the no votes. Under the repository’s rule, it passes easily; under the FAQ’s rule, the staked amount participating would be less than 24% of total active stake, far from one-third. Same vote, two rulebooks. #Solana ’s first time using on-chain governance hits this mismatch—its level of trouble isn’t lower than the proposal itself.

The voting ends with the close of epoch 1023. The earliest estimate from the official side was Thursday. I ran the timing based on the current block production speed; the landing point should be sometime tomorrow night Beijing time. Epoch length already drifts with block production speed. Given the dispute over the accompanying interpretations of the situation, it most likely will come out within these next couple of days.

$SOL Next, you can see whether the spot ETF money keeps going into BSOL. After the yield is cut, whether that 77% share stays put or disperses will explain who’s behind this inflow better than any talk about how many fewer coins are issued on the inflation chart.

The hardest-to-read part of Nvidia’s quarterly report isn’t in the income statement. The fact that revenue has doubled is already baked in by the time the seller-models start their calls—there’s no real suspense. The disagreement is concentrated in the footnotes on the following pages: whom the company has guaranteed leases for, how many years’ worth of purchase orders it has signed, and which customers it has extended payment terms to. Put these three items together, and they explain how the money from this round of AI capex circulates better than any year-over-year figure ever could.

Let’s get what’s on the surface out of the way. In the second quarter, revenue was $96.2 billion, with data center accounting for more than 90%. Gross margin held steady at above 70%, and the third-quarter guidance came in at $108 billion, higher than the Street’s consensus. These numbers aren’t controversial—they just confirm something that was already known: Blackwell Ultra is still ramping up, and whatever capacity it can produce, it sells.

The new element was a table CFO Colette Kress chose to present proactively. Nvidia’s supply and production commitments jumped from $119 billion in the prior quarter to $279 billion, an increase of $160 billion in a single quarter. Her explanation was that the money is mainly going into memory procurement. The explanation is plausible, but the implications are heavier than what it sounds like. Memory contracts are long-term—when the deal is signed, both price and quantity are locked. In effect, Nvidia is betting on the shape of demand over the next three years, and it has already paid a deposit. On the call, Kress said that instead of letting this table become a hanging question, it should be made clear outright.

Accounts receivable at quarter-end was $63.1 billion, and DSO stretched from 45 days the previous quarter to 60 days. In the 10-Q, the company spells out the standard: for large purchases from investment-grade customers, the payment terms can be extended to 90 days, and up to a year, to align with customers’ large data center construction. The same filing also includes another line noting that the combined receivables balance for five direct customers accounts for 70%.

Read these two lines together, and the picture comes into focus. Chips ship, revenue is recognized in the period, but the cash arrives only three months to a year later—and the unpaid amounts are concentrated among a few customers. Nvidia is using its own balance sheet to help customers finance working capital turnover. This isn’t automatically “bad debt,” and investment-grade customers will likely pay—but it shifts part of those customers’ credit risk onto Nvidia’s books.

In August, Nvidia also signed a guarantee with a cap of $105 billion. The guarantee covers an SB Energy campus in Pike County, Ohio, supporting the credit for land, power, and the factory, with the tenant being an OpenAI affiliate entity. The scale is roughly 4.25 gigawatts. This guarantee isn’t an investment, and it isn’t a cash outlay. It will only take effect gradually as the data center is built in phases and the lease becomes effective. The first tranche is expected in fiscal 2029; each time OpenAI pays rent for a tranche, the exposure decreases a bit. In exchange, the campus will deploy only Nvidia equipment.

Huang Renxun has long denied that this is cyclical financing. His rationale is that rent is paid by OpenAI, and Nvidia locks in resources only where it can see demand. On the call, Kress added one more point: the demand such cooperation can generate is roughly a quarter of next year’s business, and Nvidia’s platform is general-purpose and durable, making the risk therefore controllable.

The counterargument also comes with names and titles. Bill Birmingham of Rex Financial said the guarantee amount shrank from the more than $200 billion level previously rumored in the market down to $105 billion. The market read it as demand shrinking, not as risk falling. Nvidia, he said, lost $250 billion in market value because of this. Melissa Otto of S&P Global Visible Alpha took the other side; she said the whole market was shocked by the 70% figure.

The 70% refers to the annual guidance Huang Renxun gave as an exception during the call: revenue growth of 70% in fiscal 2028, while the Street consensus was only 44%. He also added that this was calculated based on supply capability, and that actual demand would be higher than that. Those words were the turning point in the trading that night. In the first hour after the earnings release, the spot order on Binance priced at $NVDAB was briefly smashed to just above 204. After the call started, it kept being pulled back, and within 24 hours it was up 3.2%. In the same #Nvidia earnings report, the income statement makes people tense, the forward-looking guidance makes people feel reassured, and what lies in between is how to read these footnotes.

My own judgment leans toward acknowledging Huang’s supply-logic explanation. Memory is the toughest bottleneck right now. Locking supply three years in advance makes sense commercially, and the $279 billion figure looks more like抢产能—grabbing capacity—than hard-building demand. But I don’t accept the claim that risk hasn’t changed. In the same quarter, Nvidia’s cash flow from operating activities was $24.1 billion, and GAAP net profit was $59.7 billion—more than double the gap. The shortfall mainly came from accounts receivable and inventory. It also issued $25 billion of senior unsecured bonds. For a company with ample cash, extending payment terms to customers while still needing to issue debt indicates that the funding pressure from this expansion is already starting to transfer onto the company itself.

There’s also an easy-to-miss accounting perspective. This quarter, GAAP earnings per share were $2.46, while non-GAAP was only $2.22. GAAP was actually higher. The difference came from $7.8 billion in equity investment gains, which were unrealized gains from Nvidia’s holdings of equity in AI companies. Non-GAAP excludes those. When the valuations of the companies Nvidia invests in rise, it directly lifts Nvidia’s reported book profit. This link is a plus item in an up-cycle—and when the direction reverses, it’s just as responsive.

Under what circumstances would I admit I’m wrong? If over the next one or two quarters DSO shrinks back toward the 45-days range, and operating cash flow catches up to net profit again, then this payment-term loosening would just be a timing difference for a few large orders. In that case, my concern would be overinterpretation. Conversely, if DSO keeps moving upward, and the concentration of receivables continues to exceed 70%, then the “investment-grade customers” wording will carry too much weight, and the market will eventually demand that Nvidia clearly disclose the names of these few customers.

The next quarter’s focus won’t be whether revenue can reach $108 billion—it will most likely. Watch how these tables in the 10-Q change: whether supply commitments add more, which direction DSO moves, and whether another name shows up in the guarantee schedule. The main text of the earnings report is written for everyone; the footnotes are written for people willing to spend an extra twenty minutes reading.

Tesla's Cybercab will make its official debut in Austin next week—no steering wheel, and no pedals. Nevada’s regulators had just, before that, loosened its grip on it in Las Vegas by allowing it to operate paid, driverless rides. This was supposed to be a fairly clean bullish story. But when the market opened on Monday, the entire rally from the previous week was fully given back.

$TSLAB hit a mid-session high of 366.42 last Friday, closed at 349.53 on Monday, then traded sideways for the following few days, and is now hovering around 349. The pricing window for this permission-related news is only a day and a bit. With the same news, bulls see it as a commercialization turning point; the seller’s money is treated as a one-time positive catalyst, and the gap in between is what this article is set to unpack.

That Nevada Department of Transportation Services vote raised Tesla’s cap on driverless taxi vehicles in Clark County from 10 to 5,000, effective for the next year. At the same meeting, Waymo and Uber also received licenses, but at a smaller scale. This is the ceiling for permissions—it’s a different thing from the number of vehicles already deployed. How big is the difference? Tesla’s own people put it more directly than anyone else. After the meeting, Cybercab chief engineer Eric Early said that 5,000 has always been the upper limit they were given. By this time next year, Tesla won’t be able to deploy 5,000 vehicles either; he said the bottleneck isn’t technology. He added that being able to do a little over 2,500 vehicles would already make them very satisfied. For a company’s chief engineer to proactively push down expectations on the very day it obtained the license—nobody says something like that casually.

The event scheduled for September 3 is set in Austin. What’s confirmed so far is that it will be invitation-only with the entire session livestreamed. Seats were given to the highest-scoring group of Robotaxi passengers from an in-app raffle. Tesla’s Model Y driverless cars have already been running for a while in several cities in Texas and Florida, and this event is more like inserting the Cybercab into a fleet that’s already in motion. The “getting it started” piece has happened long ago. A launch event that puts a new vehicle on stage, versus an operational change that can be recorded in the income statement—those should be two different prices in the stock market.

The reason behind Monday’s long black candle wasn’t actually about autonomous driving. News about a new round of auto tariffs weighed down the entire U.S. auto-plant sector. At the same time, China announced a recall covering nearly 3 million vehicles. The reason: after severe collisions and electrical circuit failures, the mechanical emergency door handles are not easy to find, which could block escape and rescue. The numbers look alarming, but the remediation is limited—OTA software pushes plus warning labels. The actual money being paid out is limited. This round of checks on hidden door handles targeted multiple automakers at the same time; Tesla is simply the largest by market size among them. In Monday’s drop, the sentiment drag was more than the bookkeeping loss on paper.

To judge how far robotaxis have progressed, mileage is the toughest set of data. Tesla’s Q2 earnings call revealed that cumulative supervised/unsupervised driving exceeded 380,000 miles. The company said that so far there hasn’t been any noteworthy accident. In the same metric for passenger-carrying, Waymo’s driverless ride miles by mid-year have already approached 200 million miles. The company said those miles can still keep rolling up by a large margin each week. Yet the gap between 380,000 and 200 million is a difference in scale, not the kind of small remaining segment left on a progress bar.

Robotaxis account for less than 0.5% of Tesla’s revenue last year, but in Morningstar’s valuation model they represent more than 30%. That firm’s current fair value estimate is $450—placing today’s price in the undervalued range. But this is a model from one institution; the market hasn’t formed that consensus. On the bullish side, the more aggressive view comes from Wedbush’s Dan Ives, with a $600 target price. His logic is that Tesla sells cars at near-cost prices to lock in the installed base, then recovers gross margin through software and mobility subscriptions.

The most specific argument from the bears comes from Gordon Johnson of GLJ. He counted an active Robotaxi fleet of only 31 vehicles; the number truly operating in an unsupervised manner is even smaller. They are all constrained within geofenced areas, and remote human staff are always ready to take over. He cites crowdsourced data to claim that on FSD v14, models like the AI4 need takeover about once every 40 miles. He then pulled up collision records over the past year involving safety drivers in the vehicle. His conclusion is that the market’s valuation for robotaxis and the level of revenue this business can generate right now are not on the same scale. His reliance on crowdsourced data is a soft spot, and Tesla also hasn’t provided better public numbers to rebut either the fleet size or the takeover interval.

I agree with Morningstar’s framework. Robotaxis are indeed the main driver of this stock’s valuation; the auto-selling portion can’t currently support that price. In Q2 revenue, the company hit a record high, but operating margin fell to 1.4%. Earnings per share were far below market expectations, and capital expenditures are still moving higher. The core business is having cash eaten away along the autonomous-driving line; the pace of monetization must outpace the pace of consumption. But on the timeline, I weigh Early’s statements more than Ives’s target price. For the $600 case, the subscription revenue and fleet revenue must bring gross margin back in next year. Early’s original wording indicates that the capacity and operations side aren’t ready yet—and that side can’t be accelerated by just writing code.

So on September 3, there is only one direction of signals that can change the judgment: whether there is a safety driver in the car, and whether the ride is actually paid. If either of those two points truly lands, Johnson’s argument will immediately weaken significantly, and Nevada’s 5,000-vehicle cap will shift from paper to a production-scheduling issue. If what’s unveiled on stage is a vehicle without a steering wheel, paired with a livestream segment, then September 3 is a launch event—not a commercialization timing milestone.

The downside risk cuts just as sharply in the opposite direction. Musk himself said on the call that safety is the biggest constraint right now; one serious accident with casualties can become a global headline. Bad news along this line doesn’t need to be proportional—one incident is enough. The string of updates tied to #Robotaxi over the next two weeks means that rather than fixating on daily ups and downs, it’s better to record the safety-driver issue in your notebook. Whether the valuation model can hold hinges on that single variable.

This rebound has been noisy for almost a week, and the least clear part is where the money actually came from. One camp insists that the shorts were squeezed through, while the other says that the real cash from ETFs has returned. These two claims point to completely opposite outcomes: a short squeeze is one-off—once it’s squeezed, the momentum dies; if subscriptions can keep going, that’s a different mechanism. I broke Binance’s futures and spot data down day by day and matched them against each other. Each camp is right for part of the story, but neither has told it completely.

$BTC is currently at 79,039, down 2% over the past 24 hours. You have to look back a week to see the full range of this move. At the August 18 close it was still 64,725. On August 25, the intraday high touched 81,272. In just a little more than seven days, it ran up about 22%. Since May, it’s the first time a number starting with “80,000” has shown up on the board, and most headlines only go as far as that.

The short-squeeze theory holds—but only for that single day, August 19. On that day, the price rose 7%, yet open interest actually fell 2%. The ratio of aggressive buy volume to aggressive sell volume was 1.22, the highest reading for the entire stretch. Price moved up while positions moved down—this is the shape you see when old positions are being closed out. New money flowing in would produce the opposite curve. That day, they really did clear out a batch of shorts.

The problem is that August 19 contributed only those seven points. The bulk came out after August 20. During that period, the position size barely moved; today’s open interest is basically the same number as just before the rally began. In dollar terms, the notional value of positions has increased a lot—that’s because the price lifted it—not because the number of contract lots expanded. Throughout the whole rally, leverage positions didn’t really grow, which runs counter to most people’s intuition.

The large buy order wave after that could only come from spot, and the ETF data lines up with it too. In the week of August 17 to 21, US spot Bitcoin ETFs saw their strongest net inflow in ten months, according to Bloomberg’s Aug 24 read. BlackRock’s IBIT got 503 million dollars in a single day on August 20. Four trading days later, it was cut in half to just over 200 million. Money is still coming in, but the pace has slowed.

From August 22 onward, the aggressive buy/sell ratio dropped for four straight days to below 1. On the order book, the volume that was aggressively smashed down is larger than the volume that aggressively swept up. Yet the price didn’t fall—it just traded sideways between 77,000 and 81,000. The portion of aggressive selling was likely absorbed by passive buy orders resting underneath, which matches the ETF’s daily subscription rhythm. What’s propping this price level right now is that mechanically-entered passive buy flow; the trading desk itself has already been stepping out. The risk is now out in the open: this price level depends more than a week ago on that daily ETF inflow, and that inflow is getting smaller.

There’s also a disagreement hidden in positioning. Binance’s large account long/short positioning ratio has been climbing all the way: it was 1.47 on August 17, and today it reached 2.26, the highest in this period. Retail accounts show the opposite trend: their long/short ratio fell from more than 2x back to around 1x. Big money is adding longs; small accounts are shrinking their exposure. Both sides have effectively taken the opposite positions. This situation doesn’t forecast direction—it only suggests no one is fully confident at this price.

Galaxy Research put out a set of statistics on Twitter on August 21. They define a bear market round as a period when the daily closing price retraces more than 50% and lasts at least 90 days. Since 2011, there have been 6 such rounds. During those periods, Bitcoin regained the 50-week moving average 13 times; after 11 of those instances, it never printed a new low again. Their current moving-average level is 82,470. I recalculated using Binance weekly data and got 81,822. The difference is a few hundred dollars because the data sources use different conventions, but the conclusion is the same. The high point on August 25 is still short of that line by the last small stretch.

Galaxy also wrote down the cost of paying for this signal. In three bear markets that lasted relatively long, the first time the price recovered the 50-week moving average happened 130 to 284 days after the lowest point. By then, Bitcoin was already up 60% to 80% from the bottom. When the signal finally confirms, most of the “meat” has already been eaten.

Measure today against that line and it becomes clear. This round’s low was July 1 at 57,800. From there to now is under two months, and the rise is about 37%. Both on the time and magnitude dimensions, it’s earlier than those historical confirmation points. There are two ways to read it: either this bear market is simply shorter and shallower, so the signal comes early; or there’s still distance to true confirmation, and the fact that August 25’s high got pinned back by the moving average is evidence for that.

I lean toward the latter. My basis is the aggressive buy/sell ratio above. On the demand side, if there really were a full pivot, the trading desk wouldn’t keep net selling for four straight days after price had stood above 80,000. The current structure looks more like passive buy orders are propping the price one-sidedly, while the trading desk is using this level to rotate positions and exit. How long this structure can hold depends on how long the ETF subscription pace can remain sustainable—and it already has begun to slow.

What would overturn this view? I said it earlier. If this coming Sunday’s weekly close is above 82,470, and at the same time ETF net inflow for a single day returns to a level above 400 million, then it would mean the passive buying hasn’t weakened and is actually accelerating. In that case, Galaxy’s dataset should be taken seriously. The real judge is Sunday’s weekly closing price; whether it touched 80,000 at any point during the day means nothing.

Record-high ETF listings or inflows are often treated as top signals, and people in both the Chinese and English communities say that. Bloomberg Intelligence’s James Seyffart commented on Twitter on Aug 21 once. His counterexample was January 2024: when spot Bitcoin ETFs were launched, the price was still in the 40,000s, and the top didn’t appear until nearly two years later. That rebuttal holds up. ETF inflow is a flow indicator—it only says who is buying during this period, not that it sends a top signal.

In the next few days, you can watch whether ETF daily net inflow can keep holding above the 300 million line, and also when the aggressive buy/sell ratio can flip back above 1—that would be the sign that the trading desk is re-entering. Both numbers are public, so there’s no need to wait for anyone to interpret them. As for whether #比特币 has actually exited the bear market—where that Sunday weekly candle closes will be cleaner than any analysis published today.

This year’s harshest research report on Circle on Wall Street—when it was published, it happened to coincide with the stock’s intra-year low. Morgan Stanley cut its rating to “Sell” (reduce holdings), slashed the price target by more than half; on the same day TD Cowen flipped the script and initiated a “Buy.” Three weeks later, the price of $CRCLB had left both firms’ targets in the dust. Needless to say, the bearish note was wrong; but even the bullish one didn’t keep up.

On August 3, Morgan Stanley’s James Faucette downgraded CRCL from Hold/Watch to Sell (reduce holdings), setting a target price of $38. On the same day, TD Cowen initiated coverage with a Buy rating and a target price of $82. That day CRCL closed at $60.35—the lowest day in three months. On Monday, on the NYSE, it closed at $87.72. On Binance’s spot market, $CRCLB is now quoted at $85.9, and in pre-market trading it has pulled back slightly, tracking the crypto-concept stocks.

This rally has little to do with reserve income. In Q2, Circle’s reserve income year over year increased by only a low double-digit percentage—or rather, just a single-digit gain. Its reserve yield was also lower than last year, and total revenue failed to beat market expectations. What the market bought were two other things. At the end of July, Circle fully acquired IBM’s blockchain patent portfolio, instantly becoming the U.S. company with the most blockchain patents in hand. Then on August 19 it also announced that Arc mainnet is scheduled for September 16. The day that news broke, CRCL’s trading volume more than doubled versus the prior day.

The money-making machine at Circle hasn’t been fixed yet. Today, USDC circulating supply stands at $73.8 billion—almost unchanged since the end of June. This year’s high was $79.6 billion on March 18. Coincidentally, CRCL’s own closing high this year also occurred on March 18, at $132.84. Both curves peaked on the same day. Before that, the market’s valuation of Circle was basically calculated based on circulating supply. Faucette’s downgrade landed right on that point. He cut his assumptions for USDC in 2027 and 2028 by 30% to 40%, and added a more unkind line: stablecoin usage is always concentrated in crypto trading and has not expanded into payments. Tokenized money market funds and tokenized deposits—both of these “things” steal both balances and fee rates.

Stopping circulating supply is only the surface. Circle’s trouble is buried in the cost line. In the same quarter, reserve income was $668 million, while distribution, trading, and other costs were $412 million. More than 60% of the money changes hands and gets paid out. The main recipient is Coinbase: all reserve income attributable to the portion of USDC held on the Coinbase platform goes entirely to Coinbase, while the two other counterparties split it fifty-fifty for USDC held elsewhere. At the end of June, about 30% of USDC was sitting on Coinbase’s books. This profit-sharing agreement was renewed in August under the original terms through 2029—effectively “welding shut” the cost structure for the next three years in advance.

Coinbase signed that agreement and is also one of the initiators of Open USD. This alliance includes more than 100 institutions—Visa, Mastercard, BlackRock, Stripe, Google are all in it. The rule of the game is: reserve earnings minus management fees are returned to the participating merchants. The slice of income that Circle and Tether have survived on by earning the spread has been carved out in this design. BlackRock is also a collaboration partner for Arc. These institutions sign on both sides; what they’re betting on is the #稳定币 track itself. As for who the issuer ends up being, they don’t care as much.

This year, the interest-rate line has actually been a tailwind. Warsh is scheduled to speak at Jackson Hole this Friday—his first public appearance since becoming chair of the Fed. Inflation hasn’t yet returned to target. In the July minutes, three votes favored tightening. Market pricing for a September rate cut has been drifting downward for weeks. Higher rates are good for Circle’s reserve income. If the shorts bet that rate cuts would thin the spread, then this year’s logic has not played out. The pressure comes from stalled circulating supply and that profit-sharing table, not from the Fed.

The market has shifted the way it values Circle—from an interest-rate ledger to a network-usage ledger—and in terms of direction, I agree. Faucette’s $38 target treats Circle as pure interest-rate beta and cut it too aggressively. Things like licenses, patents, and a clearing chain that is about to go live—none of that belongs in a spread model. But at $85, the market is already buying September 16 as the realization date. Arc going live is only the start of work; even if the testnet is bustling, it still hasn’t reached the point of charging.

The doubled full-year other-income guidance in Q2 included roughly $180 million from staged confirmations of Arc token pre-sales—one-time revenue. If you strip that out, look at whether other income in Q3 and Q4 can stand on its own. If it can, then this round of gains has a basis for accounting. If you strip it out and it’s still only in the low tens of millions range, then what you’re buying with $85 is simply the story.

On September 16, the Arc mainnet launch coincided with this round of the FOMC. If you want to follow the path of #Circle , you can compare the USDC circulating supply before and after with the actual settlement volume on Arc itself—it’s more useful than staring at the order book.

In the White House meeting room last Wednesday, a row of exchange and brokerage executives sat there while the president urged the Senate to pass the CLARITY Act as quickly as possible. Pushing the timeline forward by a day, the SEC had already posted a set of rules for issuing crypto assets. The committee approved them unanimously, with not a single dissenting vote. The market put these two developments into the same pocket, and the label on the outside read: regulatory tailwind.

Once that label was attached, that week’s weekly chart printed its biggest bullish candle since March 2023. On a weekly basis, Bitcoin closed up 23.58% this week; I personally cross-checked using Binance’s weekly K-line. Today the price is still hovering around the 770,000 level, with little pullback.

With a candle this large, it could be that someone is buying with real money, or it could be that the shorts are being carried out on the shoulders. On the K-line chart, these two things look identical. Only when you dig into the position data can you tell them apart.

First, look at the futures. On Binance, $BTC perpetual contracts had an open interest of 110,900 coins on August 17, and after a week’s rally it was 105,500 coins on August 24. In dollar terms, open interest did rise—it rose because the price rose. Over the whole week it moved up by more than 20%, but the coin-margined contract positions increased by not one coin; instead, they fell by more than 5,000 coins.

The funding rate also didn’t move. From August 22 to now, every eight-hour settlement has stayed at 0.0100%, exactly the benchmark level. In the three most intense days of the move, longs didn’t pay even a single cent of additional premium.

What jumped was the big-account long/short ratio. When price action kicked off on August 19, it was 1.43; today it’s 2.09. Big players only chased after the price had already stood up, and they didn’t really add positions during the rally. In the same period, CoinGlass data shows that the scale of short liquidations exceeded $4 billion.

Put the three sets of data together, and they point to the same action: shorts are closing, while longs are chasing higher.

But it’s also not right to write the entire week as a full-on short squeeze. The spot ETF side is bringing in real money. On August 20 alone, net inflows were $606 million—the biggest day since May 1. After that, net inflows continued for five straight days. That money has nothing to do with contract liquidations; it’s someone buying shares at the current price. A weekly gain of more than 20% can’t be sustained by just these few days of net inflows—it’s more like confirmation of the move than the engine.

Mark Cuban’s explanation is the most direct: he said the government manufactured a short squeeze that pushed the price up, and nothing else really changed. Mike McGlone of Bloomberg’s industry research also doesn’t accept this as a trend reversal; his wording was a bounce within a bear-market purge. On the opposing side, 21Shares points to an on-chain selling-pressure indicator, which has already fallen to the lowest range since 2010—there aren’t many people willing to sell at this price.

On the mechanism, I agree with Cuban: this week’s thrust really came from liquidation flows. On the conclusion, I don’t agree. The proposal filed on August 18 matters more than this week’s rally.

The SEC proposed two tiers of exemptions: one that allows projects to raise $5 million over four years, and another that relaxes it to $75 million within 12 months. The cost is additional disclosure and audited financial reports. The conditional safe harbor goes further. If the issuer can prove that the management work it promised has been completed and that it no longer takes on any new material management responsibilities, the token can exit the definition of an investment contract. The anti-fraud provisions keep applying. #crypto_regulation

The difference between this proposal and the CLARITY Act is that they run on two different clocks.

The SEC runs on an administrative clock. After the proposal is published in the Federal Register, there’s a 60-day comment period. After the comment period ends, the final vote is taken, and the whole process doesn’t require Senate approval. Also, this time it was approved through written voting by three Republican commissioners. After the last Democratic commissioner left office in January last year, there was nobody left in the commission to vote against.

CLARITY runs on a political clock. The House already voted on it in 2025, but the Senate dragged it until today. On August 8, the Majority Leader submitted a procedural motion; the vote is scheduled for September 15, with a 60-vote threshold. Even if all Republicans vote yes, it’s still not enough—you’d need to pull someone from the other side. The point where it gets stuck is that Democrats require the provisions on conflicts of interest and anti-illicit finance to be written in. Galaxy Research put the probability of this bill becoming law within the year at 50% at the end of July, then cut it to 30% in August.

Last week’s market action essentially bought both lines together. The vote in mid-September will split them apart.

A constraint breaking-news item that’s not commonly written about: Hester Peirce, who leads the SEC’s crypto task force, plans to leave in November this year to teach at university. Her term expired in June last year. Under the rules, she could have stayed until December this year, but she chose to leave early. The concept of the safe harbor was originally proposed by her. If the rules move from proposal to final form, it’s best to push them through while she’s still on the commission.

My assessment also has weaknesses. What’s on the table for now is still only a proposal. The comment period is where everyone comes in to amend the text. Both the industry and the consumer protection side will push their own opinions into it. How far the safe harbor needs the issuer to prove things—and what the final draft ends up saying—may differ a lot from the current wording. If it ends up being narrowed, this line has to be recalculated.

The Treasury’s long-term bond buyback round hasn’t finished either. The execution window is September 9 to November 4. The money hasn’t really been spent yet; the liquidity impact is still coming later. This is one factor that can support staying bullish. Also, the judgment I made using the funding rate has limited evidentiary strength. 0.0100% is Binance’s benchmark funding rate. Under a neutral market, it naturally would sit there; it can show that longs aren’t crowded, but it can’t tell you more.

In that week, $XRP rose by more than 50%, and $ETH also outperformed $BTC . The assets that are most sensitive to the regulatory lens ran out front. That suggests the market really is buying the line about rules—not just liquidity.

Next, you can look at how the market reacts on September 15. If the procedural vote doesn’t pass but the price holds, it means the market has already learned to treat that SEC line separately. If they send it back together, then what last week’s buyers bought would only be the Senate. You can also keep an eye on whether coin-margined positions and funding rates on the contract side have caught up. In a rally taken over by real money, these figures will move.

Google signed a custom chip agreement with Marvell at the end of last month, but it only came to light in regulatory filings in mid-month. After the news broke, the market slapped a fresh label on Marvell, and the stock price kept climbing. The reason: it finally landed custom AI chip positions across Amazon, Microsoft, and Google.

But if you look at the company’s own guidance, you’ll see a different picture. The fastest-growing business this fiscal year is the one that sells fully stocked connectivity chips—both optical interconnects and electrical interconnects—into AI racks. Custom chips are next, but they’re still far behind.

The story on custom chips is that growth is lagging; on interconnects, it’s where the momentum is. The quarterly report that Marvell has to file after the close next Thursday for $MRVLB will test whether this mismatch holds.

Management’s guidance for this fiscal year is that interconnect business will grow year over year by more than 70%, while custom AI chips will grow by more than 20%. Full-year revenue guidance is about $11.5 billion, and the target for next fiscal year has been raised to $16.5 billion. These numbers were given during the May earnings call. Three months have passed—now the market needs to know whether they still hold.

First, let’s talk about Google’s deal.

The headline says “up to $120 billion,” but if you scroll down one level, the structure is completely different. What Google got is a warrant to purchase shares, corresponding to roughly 7% of Marvell’s equity. Most of it hinges on Google placing purchase orders to unlock it. There are 240 “tranches,” and for each $5 billion of custom product purchases, one tranche unlocks. The window runs from this fiscal year’s third quarter through fiscal 2033.

The “$120 billion” figure means that all 240 tranches are fully filled. Compared with the $11.5 billion full-year revenue guidance, this headline number behind the scenes assumes seven years’ worth of orders that don’t even exist yet. How much Google buys, how much of the warrant vests. If it doesn’t buy a tranche, it doesn’t vest.

The market is also doing the math. Last Friday, the stock’s rally driven by the agreement was partly reversed—on that day the share price fell by about 6%. The portion that comes from warrant dilution got put on the table. By the weekend close, $MRVLB spot was stuck at $234.82, and the stock board barely moved.

There’s also an older issue on custom chips. In December last year, Benchmark’s Cody Acree downgraded Marvell from Buy to Hold. His reason was that the two generations of Amazon Trainium3 and Trainium4 designs landed with Taiwan’s Wiwynn.

At the time, Matt Murphy responded that when switching the main XPU customer from the current one to the next, everything was already accounted for in the numbers he provided—orders and visibility included. Acree didn’t accept that explanation. He believes Marvell’s revenue forecast relies on existing shipments of Trainium2. Whether Trainium3 can smoothly take over is still an open question.

That’s been over half a year. This earnings report may or may not provide a definitive answer. But it helps explain why the growth guidance for custom chips is only a bit over 20%. If a socket that’s already in mass production is lost, growth will be cut in a very real way. The new signed products from Google won’t start counting in revenue until after this fiscal year’s third quarter. Winning designs at hyperscalers isn’t a renewal contract—each generation has to compete again. Even winning this generation doesn’t guarantee being in the next one.

On interconnects, the guidance has been raised once this year, and the reason is that 1.6T optical interconnect demand is taking off. The number of compute chips in the racks is rising, and for each chip, the number of accompanying optical modules, DSPs, and electrical interconnect chips rises as well. At this stage, Marvell looks more stable in interconnects than in custom chips. It doesn’t have to join design bids again every generation, and it isn’t betting on whether any one customer and any one project will tape out on time.

With these accounts in mind, I’m more inclined to view Marvell as a high-speed networking chip company, with custom chip design in second place. A forward P/E of about 52x—if it’s being propped up purely by the custom chip story—requires that Google’s attach products and the next generation of XPU both show up in revenue on schedule. If the network business is what’s carrying it, then you’re betting on how long this 1.6T replacement cycle can stretch. In the latter case, you can see orders now; in the former, you’re still waiting for orders.

What’s most worth watching in this report is whether management has moved those growth assumptions—whether the interconnect growth gets pushed higher, and custom chips stays flat or goes down. If that happens, it suggests the story the market is buying and the money the company is making are continuing to diverge. Then the $16.5 billion target for next fiscal year will have to carry even more weight. If management does the opposite—raising the custom chip growth rate, or providing a production ramp timeline for the next-generation XPU—the mismatch would close itself, and you wouldn’t need to bring up those concerns anymore.

The bull case isn’t weak. After the agreement was announced, Roth Capital’s Suji Desilva maintained a Buy rating and raised its target price. He’s focused on the stickiness of attach products. Once things like storage controllers, network cards, and memory interfaces are integrated into TPU racks, the cost of switching is higher than switching out a computing chip.

On the bear side, in mid-July Erste Group cut its rating from Buy to Hold, citing customer concentration, valuation, and slower profit growth. The first risk points to the same hidden issue as the Trainium “old case.”

There’s also a timing coincidence. This earnings report is scheduled one day after Nvidia’s, and the market will likely read the two together. What Nvidia’s report tests is whether general compute is hot or not. In Nvidia’s upcoming capex cycle—#AI芯片 —how much goes to custom chips and the networking side depends on Marvell’s report to provide the numbers.

If you want to judge this earnings report, you can first shift some attention away from whether revenue beats expectations. The company did beat last quarter. After beating, which direction the stock moves depends more on how segment growth is defined. The growth gap between the interconnect line and the custom chips line tells you more about what valuation multiple the stock should be priced at going forward than the few tens of millions of dollars by which total revenue exceeds expectations.

On Twitter, now you can see the same question after only a few posts. Who can explain in the simplest terms why Zcash is actually going up? The answers underneath are almost identical: the ETF has been approved, and trading starts next week. The screenshots with the highest repost count have “confirmed” written directly in the title.

Grayscale’s 8-K doesn’t say it that way. The original text says the shares are expected to begin trading on the NYSE Arca around August 25 under the ticker ZCSH, and that the trust will also be renamed The Zcash ETF. Immediately after that comes this line: all of the above must obtain the relevant regulatory approvals; there is no guarantee that the shares will be listed according to the issuer’s expected timetable, nor any guarantee that they will ultimately be listed.

A timetable written with “no assurance” held onto by oneself was treated as a fait accompli on both the Chinese and English timelines. The price ran up right out of this mismatch.

In a week, $ZEC rose by more than sixty percent, reaching a level it hasn’t been at in eight years; its market-cap ranking is back to roughly the top ten. What drove it was two consecutive filings: the fourth amendment on August 18, the fifth amendment on August 21—and on the same day they also filed that 8-K. Off-exchange, the Winklevoss brothers also kept showing up with support for several days. Tyler said Zcash is the most exciting and least discovered opportunity in the crypto world—crypto’s version of Bitcoin. Their controlled company Cypherpunk announced on August 18 that it had started mining ZEC on its own; prior to that, it had already been the publicly listed company with the most ZEC holdings in the world.

The market’s misread doesn’t stop at this one approval.

Another widely circulated claim in the Chinese community is that DCG plans to buy more than $100 million worth of ZEC. The prospectus says something completely different. A wholly owned indirect subsidiary under DCG is conducting non-binding discussions that could involve authorized participants subscribing for trust shares; the consideration would be approximately 200,000 ZEC. That’s using coins they already hold to exchange for shares—so the trust’s size grows, but there won’t be more buying demand for a single coin in the market. Non-binding discussions are still a long way from signing.

This time, Grayscale is simply taking a trust that has been running since 2017 and switching it onto a listed exchange venue. No new fund is bringing fresh money to build positions. As of June 30, this trust held 2.3% of circulating ZEC—these coins have long been sitting on-chain. The real change brought by the conversion is that the subscription and redemption channels were opened. For holders of the old shares, their cost basis is far below today’s price; the redemption channel is first and foremost an exit for them.

Even the annual fee number can be read as a message. This product is priced at 2.5%. Before this, the most expensive crypto ETF across the whole market was Grayscale’s own GBTC at 1.5%, while BlackRock’s IBIT charges 0.25%. A 2.5% fee is a quote this category has never seen. Grayscale is willing to open at this price because it judges that, in the foreseeable future, nobody will be able to offer a second channel like the same one—and it also judges that these buyers won’t shop around just for fee rates.

So in this rally, how much of the story versus the compliance channel is actually driving it?

The evidence on the “story” side is harder than I expected. Zcash’s shielded pool currently holds about 26.8% of the supply, and this ratio is still trending upward this month. People willing to lock coins into the shielded pool are using the chain’s functionality—they don’t look at the order book, and they don’t care when the ETF starts trading.

But in today’s price, the channel component is heavier. On the derivatives side, 24-hour trading volume is seven times that of spot. A market that’s truly hoarding privacy assets wouldn’t look like this—hoarding money goes into spot. At the same time, the funding rate is still sitting at the 0.01% benchmark every eight hours; the longs haven’t been squeezed into a frenzy. Leverage is piled up quite a bit, but it hasn’t reached the point where someone is gambling their life. The money entering is chasing an event, not building a long-term position.

Yesterday intraday it was hitting new highs; today $ZEC is back around 788, and over the last 24 hours it’s down. The ETF hasn’t even started trading, yet the price turned first and fell.

The “narrative” layer is real—it will remain. The “channel” layer is one-off, and before August 25 arrives it has basically already been priced in. There isn’t much sell pressure on the day of listing; what I’d really like to know is whether, after listing, there is a second wave of incremental capital coming in.

It’s only been a bit more than two months since that June incident. During an audit, security researcher Taylor Hornby found a vulnerability sitting in the Orchard shielded pool for four years that could mint unlimited ZEC. It was discovered on May 29, fixed on June 2, disclosed on June 5—on that day ZEC plunged 38%. The development team said there was no trace on-chain of exploitation. The trouble is that Orchard’s entire design goal is to make traces invisible: this can neither be proven nor falsified.

Putting this next to the ETF creates the contradiction. The ETF works via authorized participants subscribing and redeeming at net asset value; the whole mechanism is built on the premise that NAV can be verified and calculated. But the supply of ZEC can never be fully determined. You’re trying to put a deliberately uncheckable asset into a shell that must calculate NAV day by day. Custodian Coinbase Custody can solve custody, but it can’t solve this layer.

The risk in Europe points in a different direction. New EU anti–money laundering rules require regulated platforms to stop providing services for anonymity-enhanced tokens after July 2027, and Zcash is on the list. While the U.S. is opening a compliance gateway for privacy assets, Europe is shutting down existing gateways. The same asset getting opposite answers in two jurisdictions can’t persist long-term.

The strongest argument on the other side is one I also agree with. If ZCSH really gets listed on time and starts seeing sustained inflows—and for the first time, a compliance channel becomes available for privacy assets—this precedent can support a premium for a not-short while. What I said above about the one-off effect would then be directly falsified.

The only readings that can truly distinguish narrative from channel are where the shielded pool ratio goes after this round of行情. If it keeps moving up, it means someone is using the #Zcash chain; if it turns and moves down, it means only traders are touching that ticker. This data is public, so you don’t have to wait for anyone to interpret it—you can take a look for yourself after August 25.

The last time Nvidia reported earnings, revenue set a record—data center revenue was nearly doubling, and the company’s guidance for the next quarter was also higher than the market consensus. After the close that day, the stock price dropped.

This isn’t the first time. There’s nothing obviously wrong with the earnings report itself. The market stopped pricing it based on how much they expect to make in that particular quarter. Next Wednesday after the U.S. stock market closes, Nvidia will report the quarter ended July 26. The company’s guidance is around $91 billion, plus or minus 2%, with institutional consensus slightly higher than that number. Binance spot $NVDAB has been tracking the U.S. stock market closing prices these past few days. The weekend is just waiting for this earnings report. And what I care about has little to do with whether revenue beat expectations.

Let’s go back to that previous quarter’s report—the line that’s easiest to skip. Revenue was $81.6 billion, up 85% year over year. Everyone saw that. Look at the next line: net profit was $58.3 billion, even higher than operating profit for the quarter. The extra portion goes into other income: $15.9 billion, and most of it is Nvidia’s unrealized gains over those three months on publicly traded stocks it holds. In a manufacturing company’s GAAP earnings, the GAAP EPS ends up higher than non-GAAP.

In Nvidia’s income statement, there is now a piece dedicated to the market value of the AI companies it has invested in.

And it keeps adding to it. In the same quarter, the company spent $18.6 billion buying non-public securities. In the same period last year, it was only a small fraction of that. The scale of non-public equity on the balance sheet nearly doubled within a quarter.

This line has moved to a new level recently. On August 17, Nvidia filed an 8-K. The company signed a set of residual value guarantee agreements with SB Energy, a unit of SoftBank, tied to a large AI data center campus in Pike County, Ohio, where the tenant is an entity affiliated with OpenAI. Nvidia’s payment obligations under these agreements have a cumulative maximum cap of $105 billion.

The terms themselves are more worth reading than the headline number. The guarantee takes effect only after the landlord completes the delivery conditions, which the company expects to begin in 2028. If OpenAI defaults on bankruptcy or fails to pay rent, only then does Nvidia have to make up the difference between the guaranteed minimum lease value and the actual proceeds from disposal and recovery. One of the termination scenarios is that OpenAI receives a qualifying rating. In addition, it commits that any advance payments Nvidia makes will be fully repaid by OpenAI. The first version of this arrangement was much larger: the number reported by the media at the end of July was $250 billion. Over the following weeks, investor complaints about revolving financing were made public, and only later did it settle at $105 billion.

With this filing, what used to be argued about only through rumors can now be discussed based on the actual terms. Bears have been repeating this for a year: before those several-hundred-billion-dollar AI data center lease obligations are formally put into use, they sit off-balance-sheet. Once they get concentrated onto the balance sheet, the situation could resemble Cisco in the year 2000. Michael Burry has been the loudest voice for this view. On the accounting front, he isn’t wrong—the 8-K was filed for the portion of obligations arranged off the balance sheet.

But I don’t agree with equating it directly to Cisco. $105 billion is a capped maximum amount. It doesn’t equal expected spending. It only gets triggered if OpenAI defaults, and the advances must be repaid by OpenAI once the rating requirement is met—the guarantee is automatically released when the rating criteria are satisfied. This looks more like Nvidia lending its own credit to a customer that hasn’t yet obtained an investment-grade rating, in exchange for the campus’s land, power, and compute bays. Calling it a ticking time bomb, or calling it a purely commercial arrangement, both amount to laziness.

What makes me think the demand side still holds up is another set of numbers. As of April 26, Nvidia’s manufacturing, supply, and capacity commitment totaled $119 billion, of which $95 billion needs to be paid off within the remaining time of this fiscal year. This is the company putting real money behind orders upstream—harder than any adjective you’d hear on an earnings call. If they’re willing to sign this many orders for capacity years ahead of a few quarters, it shows they believe those goods will be sold.

On August 26, I’ll start by flipping through the guidance for the third quarter. The numbers for the second quarter have already been boxed in by that $91 billion figure. What the market lacks is the metric for the next quarter. Gross margin is another place to look. The company’s guidance for Q2 is around 75%. During the ramp-up period for the new architecture, gross margin is typically pressured. Whether it can hold that line matters more than selling an extra few billion—because it indicates how much pricing power it still has with customers. Going further into the commitments section, it’s also critical which direction the $119 billion moves. That determines whether the company continues to ramp up or whether it signals something it sees and chooses to reduce. And there is the finalized text of the residual value guarantee: in the 8-K, the agreement form is listed as an exhibit, submitted together with the 10-Q for this quarter—meaning next Wednesday’s filing. Across the entire earnings cycle, this is the only truly new piece of material.

There is one more metric that’s easy to overlook. When the company gives $91 billion guidance, it assumes that China data center compute business revenue is zero. If that assumption stays unchanged, the China market looks more like an option layered on top of the current guidance—any loosening would only add upside. But once the assumptions change, the market’s full-year model needs to be revised accordingly.

On the risk side, I don’t want to be vague. This company’s customer concentration keeps rising. In the prior quarter, three direct customers accounted for 21%, 17%, and 16% of total revenue, respectively—compared with only two customers exceeding that threshold in the same period last year. The top three also represent an even higher proportion within accounts receivable. And these same customers are also spending heavily on developing their own chips. In late May, Joseph Moore of Morgan Stanley maintained a Buy rating, and the reason is exactly here. He argues that for those hyperscale vendors developing their own ASICs, the next two years will actually be when Nvidia’s growth is fastest—because a shortage of compute makes everyone scramble for chips first before negotiating roadmaps. The logic will start to loosen first once compute capacity is no longer tight.

There are two things that could overturn my view: either third-quarter guidance is clearly below the market’s current expectations, or supply commitments pivot downward in the new quarter’s financial statement. If either happens, the thing to worry about is that demand itself is already loosening first. How the accounting is recorded would then be secondary.

Before next Wednesday, the market will likely keep circling the revenue number. If you want to spend a little time, you can read the 10-Q filed alongside the earnings report—especially the commitments section and the residual value guarantee provisions in their original wording. #英伟达财报 is often the clearest indicator of where the risk truly sits—precisely in those parts that nobody reads.

On Wednesday afternoon, the Ministry of Finance did something rather unusual: it went into the market to buy its own long-dated Treasuries. By Friday’s close, the bond market returned the favor—promptly reversing it. Yields on the long end were back to where they stood before the Ministry’s move. Meanwhile, in the same set of trading days, crypto here didn’t give anything back at all.

First, let’s look at what the Ministry of Finance announced. In an announcement dated August 19, it singled out two maturity buckets—10 to 20 years and 20 to 30 years—and raised the single-transaction cap for buyback liquidity support from $2 billion to at least $4 billion. The policy starts on September 9 and runs until November 4, covering the refinancing season. The official rationale was to provide more liquidity support on the long end.

$BTC that day climbed from the 64,000s straight up to above 70,000, and the market instantly gave the move a name: the Ministry of Finance was stepping in to prop up the market.

“Propping up the market” is an early read. The Ministry of Finance can’t create money. The cash it has only comes from two sources: taxes and borrowing. To buy back long-dated Treasuries, it has to make up the funds with newly issued short-term debt. Once the full circle is completed, the market ends up with no more money than before—it has merely swapped long-duration paper for short-duration paper.

This round of action changes long-end supply and demand, but overall liquidity doesn’t actually change. Even at the maximum scale—say, $30 billion per quarter in buybacks—spread over a stock of $4 trillion in outstanding debt, it’s barely enough even to count as a rounding error.

The bond market’s reaction was more honest than the reactions on Twitter. The 30-year yield closed the day before the announcement at 5.28%, was pushed down to 5.19% on the announcement day, and then finished Friday at 5.27%. After three days of a round trip, it was basically as if nothing happened.

In a Thursday report, the FT put it plainly: this intervention by Bessent didn’t manage to soothe investors’ nerves. The Wall Street Journal, in the same day, quoted its Federal Reserve reporter Nick Timiraos saying that long-bond yields had given back all or part of the downward move brought by the buybacks. The amplified buybacks don’t even begin until September 9. For those three days, the market’s trading is purely a reaction to the paper announcement.

Crypto didn’t “give it back.” $BTC is at $78,568, up 5.52% in 24 hours. I personally ran the numbers off a Binance weekly chart: the increase for the whole week was 24.6%. Lining up all 400 weekly candles from the end of 2018 to now, only four are more explosive. The last time something at this level appeared was March 2023. $ETH and $SOL over the same period were up roughly 30%—more than even BTC.

Even gold rose the same week by 5%.

At this point, that popular explanation can’t hold. If the reason for this rally is that the Ministry of Finance pushed yields down, then since yields weren’t actually pushed down, the rally should have already been reversed. It wasn’t. That means the market wasn’t buying the “yield decline” story.

I lean toward another interpretation: the market is buying the fact that the Ministry of Finance had to take this action in the first place. On Thursday, Jeffrey Currie, head of global commodities research at Goldman Sachs, spelled this out: a sovereign that has to go into the market to buy its own bonds to set prices has already implicitly admitted that the market won’t give it that price.

With the 30-year yield still sitting at 5.27% and debt only just crossing $4 trillion, the Ministry of Finance blinked first. Following this line of thought, the question becomes where the money is likely to hide—and the answer points toward the “dilution” side. Gold and crypto rising in the same direction in the same week is itself evidence that this rally isn’t crypto’s own separate story.

The counter-argument also needs to be laid out. On August 19—the same day—the White House held a crypto summit. Trump directly urged the Senate to pass the CLARITY Act, and the heads of Coinbase, Robinhood, and Kraken were all in attendance. This is real, tangible policy-positive news—and it did happen on that same day.

But it can’t explain why gold rose at the same time, nor why after August 20 and 21 when the bond market had fully digested the buyback-positive news, crypto could still keep climbing for two more days. The bill itself was still stuck in the Senate; only after the Senate reconvenes on September 14 is there an actual window. Right now, it’s just a promise.

Beyond the magnitude of the rally, there’s another layer. This move puts the Fed in an uncomfortable position. In a July 29 press conference, Waller explained why the Fed wouldn’t hike rates. The reason wasn’t that inflation had fallen. His exact words were: “In these 42 days, we haven’t done much, and the market has done quite a bit.” He also said that during this period between meetings, financial conditions tightened—giving the Fed some confidence that it can achieve price stability. The bond market did the work of adding the hike for him, so he didn’t have to move.

What the Ministry of Finance is doing now is exactly aimed at dismantling that tightening. If it truly dismantles it, then Waller would lose the very reason he quoted himself—rate-hike pressure would return to the Fed’s hands. As it stands, it looks like the tightening wasn’t dismantled; with the long end still hanging around 5.27%, he can continue using the market as a shield.

No matter which route you take, it doesn’t lead to a “looser policy” script.

The push toward rate hikes is also building. At the July meeting, the three regional Fed presidents voted for a hike; it was the most consistent disagreement since September 2016. The minutes published the same day said that most participants believed that if inflation doesn’t fall, a rate hike may be necessary. Market pricing for September hikes is roughly one-third. Meanwhile, tensions in Iran haven’t eased; oil has been climbing all week, and this line also adds to the inflation case.

Where might my judgment be wrong? Let’s set a check point. If after the September 9 buybacks truly kick off, the 30-year yield stabilizes below 5% while gold turns downward, then this rally really was a liquidity story, and the “dilution pricing” interpretation should be discarded. Conversely, if the long end continues to hover in the 5-handle and both gold and crypto keep strengthening together, then the nature of this rally would be set.

Next Friday, at 10 a.m. Eastern time, Waller will deliver his first remarks since taking office at Jackson Hole. This year’s theme is “the impact of financial innovation on payments and policy.” Only 19 days remain until the September 16 rate meeting. You can watch whether he continues to credit the market with tightening financial conditions. If he changes his tune and says financial conditions have already turned looser, then everything that has risen this week must be recalculated. #JacksonHole

Robinhood finished filing its financial reports by the end of July, and the market had already settled the story within a few hours: for the first time, forecasted-market revenue outpaced crypto, and this brokerage finally doesn’t have to rely on coin prices for a living. Over the next month, sell-side reports and the bullish crowd on Twitter basically repeated that same line.

Revenue really did set a record. But when you line up the events contracts from the two quarters, the way they rose doesn’t match the way the number of contract units rose.

In Q2, event contract revenue was $156 million—more than a tenfold year-over-year increase. In the same quarter, crypto trading revenue was $100 million, down by nearly 40% year-over-year. Other than crypto, nearly every line at Robinhood hit record highs: stocks, options, net interest, and Gold card subscriptions all climbed steadily upward—CFO commentary in the report was essentially firing on all cylinders.

The problem is the denominator. In Q1, customers traded 8.8 billion event contract units, generating $147 million. In Q2, units surged to 13.6 billion, but revenue increased by less than 10%. Money left for the company per contract fell from 1.67 cents to 1.15 cents.

We need to give a discount here. The $147 million in Q1 was reported under the accounting line “other transaction revenue,” mostly made up of event contracts; actual event contract revenue would have been lower than that. So the decline in revenue per unit was slightly less than 30%, but the direction was unchanged: units were climbing, while the unit price was falling.

Why did the unit price fall? Two things in June explain most of it. Robinhood launched Rothera in June—a regulated exchange with a licensing-approved venue and its clearinghouse—jointly operated with Susquehanna and run independently. And in the same month, the World Cup kicked off. According to Bernstein’s statistics, the World Cup accounted for 93% of Robinhood’s forecast-market trading volume from June through early July. Sports contracts had very high ticket counts but extremely thin revenue per trade; customers cycled back and forth dozens of times a day, and the company took only a little each time. Over on Polymarket, they also pushed down taker fees and offered rebates to limit-order makers—so overall take rates across the whole segment kept falling.

Put together, most of the earnings from the event contracts leg now comes from the tollbooths on sports events.

Legal risk therefore became more concrete. On June 17, Kentucky Attorney General Russell Coleman filed a lawsuit against Kalshi and Polymarket. His argument was that the sports contracts offered in his state require a state license—and both companies lacked one. Robinhood, Coinbase, and Webull were named together as distributors, and similar lawsuits have already been laid out across more than a dozen states. On April 6, the U.S. Court of Appeals for the Third Circuit ruled that the federal Commodity Exchange Act takes precedence over relevant state law—essentially issuing a “get-out-of-jail” card for the industry. But there are moves going the other way too: at the end of March, lawmakers introduced a bill in Congress to remove sports and entertainment event contracts from the hands of regulators. The ceiling for this leg is now drawn by courts and Congress, not by the product team.

Looking back at the crypto line. Crypto was $100 million, down 40% year-over-year, and the crypto notional amount shown in the app fell by 35% as well, landing at $18.0 billion. But that was a quarter when Bitcoin slid steadily from its mid-year highs, and volume was already contracting. The environment has since changed. In the past three days, $BTC went from 64,725 to 77,325; as turnover expanded in sync, the volume trend has clearly turned.

So I don’t buy this “passing the baton” narrative. The market’s pricing of Robinhood is that it’s a growth brokerage shedding the crypto cycle. Bernstein estimates its forecast-market revenue could reach $586 million this year. I’d rather read this earnings report the other way around: forecast markets are shifting from a scarce business into a more competitive, share-taking one. Volume is still rising, but unit price has been pressured down by competition across the venue and by the structure of sports contracts. Meanwhile, the crypto leg was marked down at the bottom of the cycle—its elasticity hasn’t disappeared, it just didn’t have its turn that quarter.

One more accounting point I’d like to flag. In Q2’s diluted EPS of $0.62, $0.14 mainly came from a one-time gain related to the Robinhood Ventures fund no longer being consolidated. Excluding that, EPS would be $0.48. Comparing $0.62 with market expectations would make this quarter look better than it really was.

The third-quarter report in late October will test this view. Watch two lines: whether revenue per contract unit can stabilize near 1.1 cents, and whether contract units in August and September collapse after the World Cup ends in mid-July. If revenue per unit stops falling and starts climbing again, and units also continue trending upward, then I’d be wrong—this leg would be holding the pricing power, not merely catching a single event.

On the market side: on Binance spot, $HOODB is quoted at $100.19, roughly matching the pre-market quote in U.S. stocks, and it barely moved over the day. Within the same 24-hour window, $COINB rose 7.9%. The elasticity in this Bitcoin rebound was first “assigned” to the purely-crypto brokerage market. If crypto trading volume really can come back like this, then that leg—written off as a drag—might actually be the first place for incremental upside to show up at Robinhood in Q3. If you want to track this line, you can compare the two firms’ moves side by side on the same chart and look at them together.

In Saylor’s update last Monday, there was no mention of buying. What he talked about was how many more days the duration of the U.S. dollar reserves had been extended, how many basis points STRC’s credit spread had tightened, and how many preferred shares had been repurchased in the week. In the past few years, every Monday this company’s fixed routine was to announce how many more bitcoins it had accumulated. But in the same spot this time, what’s laid out reads like a cash management report submitted by the Treasury.

Go back week by week through Strategy’s quarterly 8-K filings and look at the final bitcoin purchase. It stops in the week in mid-June. After that, through August 16, the buy column in eight consecutive weekly reports has been blank. In the meantime, there were also transactions in the opposite direction: two bitcoin sales from late June to early July, and two more sales from late July to early August—four trades totaling 6,916 bitcoins. Now the position stands at 840,447 bitcoins.

The footnotes to the 8-K also spell out where the money went. Last week the company issued common stock via an ATM, netting $333.7 million—none of it went into bitcoin. That money was split into three parts: $52.4 million to pay STRC dividends, $132.2 million to repurchase STRC in the open market, and the remaining $149.1 million went directly into U.S. dollar reserves. In the prior week, after selling that batch of coins, the proceeds were also all used to repurchase STRC.

U.S. dollar reserves are now $4.8 billion, down from about half that amount at the beginning of July. The company has already defined this pot of money in writing—meant to support preferred-share dividends and debt interest. Stopping bitcoin buys, continuing to issue shares, and stacking up cash—these three actions point to the same move: Strategy is thickening the safety cushion on its liabilities side.

Why is this happening? In his investor Q&A on August 17, Saylor laid out the rules himself. His trigger line is 1.0 times mNAV. If the share price trades above that line, issuing common stock to pay STRC dividends is worthwhile: the assets you get back for selling each share are more than what gets diluted. If it falls below the line, the same action turns into pure dilution—at that point, it should buy back common shares. The moderator of that Q&A, Natalie Brunell, later condensed the whole thing into a single sentence: fix the credit, and you fix the equity.

Following that line, the machine’s input variables have already changed. It used to consume the upside in bitcoin; now it consumes the premium of MSTR relative to net assets. STRC’s dividend was increased starting in July to an annualized 12%, paid every half-month. This bill isn’t affected by how bitcoin is trading—on the due dates, it has to be paid. As long as the premium remains, the issued funds can both feed the dividends and buy bitcoin; once the premium disappears, the issuance capacity is only enough to keep dividend payments and the preferred-share price propped up first.

This logic isn’t unique to just one company. For all companies that stockpile bitcoin by issuing shares, the wheel’s driving force comes from the same place: the premium of the stock price relative to the assets in hand. When the premium exists, issuing shares is like letting existing shareholders pick up bitcoins for free—bitcoin rises, the premium expands, then more issuance, and the loop climbs endlessly. When the premium disappears, the same loop turns the other way: issuance becomes dilution, and management is left with selling assets and shrinking the balance sheet. Strategy reached this stage earlier than its peers because its liabilities side is more complicated—four series of preferred shares plus convertible notes, and each layer requires cash to be set aside. If you understand what it has been doing over these eight weeks, you’ll have a pretty good sense of what other bitcoin-hoarding companies will face next.

As for the claim that Saylor is dumping, that’s wrong. Of the 840,447 bitcoins fewer than at the June peak, the missing portion is less than 1%. If you spread the total sell volume across bitcoin’s average daily trading volume, it’s almost invisible and has virtually no direct impact on price. The company is also supplying counterevidence to the market itself: among the top fifteen institutional shareholders, twelve added to their positions in the second quarter. That is company-provided data; at minimum, it suggests that institutions willing to buy MSTR haven’t collectively retreated.

The other half—I disagree with that. This week, bitcoin rallied from above $63,000 to above $74,000. The spot price is 74,472 dollars ($BTC ). Up 7.87% over the past 24 hours. In this rally, the buyer with the largest market-wide footprint, the one least sensitive to price and least in need of timing, didn’t show up. Marginal buying has changed hands—replaced by whom is something no one yet has been able to answer clearly with data.

There’s yet a more unflattering layer. The company’s disclosed average cost basis for all its holdings is $75,385, and today bitcoin’s intraday high has already touched above that line. The nearly seven thousand coins it sold from late June to early August all transacted at prices just above the $60,000s. That doesn’t have much to do with judgment. Preferred-share dividends are a bill with fixed dates: no matter when bitcoin comes back, the bill won’t wait.

The next developments that could change the machine’s operating conditions come from the index side. On August 14, MSCI opened a consultation. The proposal would exclude companies whose core operating asset ratio is less than half from the global investable market index, and using May data for backtesting, Strategy would be removed from ACWI IMI. This round of screening deliberately avoids using the words “digital assets,” treating all assets neutrally. Feedback ends at the end of September, results are due in mid-October, and if approved the change takes effect in the November index review. Strategy has already publicly responded: the index provider’s job is to measure the market, not to decide what assets a company is allowed to hold. #Bitcoin

When most companies are cut out by the index, the trouble stays in liquidity. This company will transmit the impact all the way to the financing end, forcing passive capital to sell and push down mNAV—and mNAV is the switch that controls this financing machine.

My view is that this eight-week pause is structural. As long as the premium of common stock relative to net assets can’t return above the line, there’s no reason for the buyback-by-issuance to restart; the money raised will continue to flow first to dividends and the preferred-share price. Under what circumstances would I change my view? If mNAV stays at or above 1.0x for the long term, if STRC’s market price holds above par, and the buy column in the weekly reports shows up with numbers again—then these eight weeks would just be a temporary brake period during a cash-flow squeeze.

Tokenized $MSTRB is quoting $116.55 today, up 8.93% over the past 24 hours, running ahead of bitcoin. People willing to pay a premium for leveraged exposure are still there. In the coming weeks, keep an eye on the footnotes in each Monday’s 8-K about ATM usage—the lines of small print will tell you what this company is thinking earlier than the position numbers. #MSTR

A company announced it would sell a tranche of debt, and the stock price was hammered down by a large margin that same day—this happens in the U.S. stock market every week. Nebius is worth spending extra time on, because after you read through the pages of the terms in the announcement, you’ll find that the place where the market hits the stock doesn’t match where you should really be looking. The information is hidden in the interest-rate spread between two bond tranches, and that spread is already talking about more than just this one company’s situation.

First, let’s lay out the facts. Nebius rents out AI compute capacity. It builds its own data centers, buys GPUs, and sells its compute power under long-term contracts to customers that want to do training and inference. The founder is Arkady Volozh. On August 19, before the market opened, the company announced it would issue two series of convertible senior notes: one maturing in 2030 and the other maturing in 2034. The stock slid steadily throughout the day, closing down 10%. During the session it dipped even further. By that evening’s pricing, the offering size was increased to $5 billion. In the same announcement, it also included an exchange arrangement to swap some of the older notes maturing in 2029 and 2031 for Class A common stock. This was the third convertible-debt deal in less than a year.

A big chunk of that drop was mechanical. The company even spelled it out in the announcement: holders participating in the exchange, after receiving shares, may sell them directly in the public market, or they may unwind the hedging positions they previously set up for these convertibles. For convertible-arbitrage institutions, the strategy is always to buy the bonds while shorting the stock at the same time. When the exchange and the new bond pricing are squeezed into the same day, stock supply and short hedges will surge together. You can’t tell, just by looking at the size of the drop, how much came from valuation judgments versus how much came from this flow. If you only look at how much the stock fell, you’ll end up drawing skewed conclusions.

The terms themselves are more worth focusing on than the magnitude of the selloff. The 2030 tranche pays 0.50% coupon, and the 2034 tranche pays 4.50%, and at maturity the latter must be repaid at 125% of the original principal. Same company, same day, same term sheet: borrowing money for four years or less is almost like getting it for free; borrowing for eight years costs the price of a high-yield bond. Back in March, for that round of convertibles, both tranches’ coupons were still below 3%. Five months later, Nebius’ revenue is rising and contracts are increasing, yet the cost of long-dated funding has been pushed up dramatically.

Put this into the macro backdrop and it all fits. This week, the yield on the 30-year U.S. Treasury rose to 5.31%, the highest level since 2007. Barclays’ Anshul Pradhan decomposed the rise in long-end yields into higher fiscal deficits, higher term premium, and AI-related issuance competing for the same pool of buyers as Treasuries. The same institution estimates that this year’s net corporate bond supply will increase noticeably, and a substantial portion comes from technology companies financing data centers. There just aren’t many lenders willing to lend for more than ten years; AI infrastructure is absorbing that funding, so whatever remains naturally gets more expensive. Nebius’ 4.50% 2034 bond is a specific readout of how this environment lands on an individual issuer.

So why can it still get a 0.50% coupon on the short tranche? Convertibles sell equity options. The conversion price was set more than 40% above the stock price on the announcement day. Investors accept an almost zero coupon in exchange for the right to convert after the stock price rises. CoreWeave in the same sector took a different route: unsecured high-yield debt is priced purely on credit, with no option to sell—so the coupon has to be much higher. Nebius currently stands on the side where it can use equity to get low interest. That is the dividing line between it and peers that have already been priced as highly leveraged assets.

What supports that position is the order book, not the narrative. In its Q2 earnings, Nebius had a contract backlog of $40 billion. The biggest source of funding this year is customer prepayments. New signed contracts cause customers to pay up front for a large part of the construction costs. It also did an asset-backed financing in July, with collateral being the cash flows from contracts of a certain investment-grade customer. Customers pay first, the company builds the facilities next—this kind of money is cheaper than any kind of debt. That is the real reason convertibles can be pushed down to a 0.50% coupon. Along this #AI算力 pathway, companies that can secure large prepayments versus those that can’t will see their financing curves diverge ever more.

The risks are on the other end, too—also very specific. In Q2, quarterly capital expenditures were $5.66 billion. This $5 billion funding deal this time roughly covers just a bit more than one quarter of construction. The burn rate determines that the company has to keep coming back to the market. Every time it returns, it must re-paper the terms at the long-end pricing of that moment—and long-end pricing is moving in a direction unfavorable to it.

My view is this. The market read this announcement as yet another bout of dilution. That reading isn’t entirely wrong, but it’s too shallow. I agree with the seller-side premise: the contract backlog and customer prepayments are real, and this company doesn’t finance itself by storytelling. I disagree with the conclusion that follows from that—because the financing channel is still open, the problem isn’t severe. What got “more expensive” in the terms is the 2034 tranche, which indicates that the funding providers are willing to bet on the stock price within the next five years, but are less willing to bet on its ability to repay eight years later. These are two completely different kinds of trust, and the market separated them by the price it offered this time.

In what situation would I be willing to admit I was wrong? I can say it plainly. If the next round of financing still manages to secure coupons close to zero within five years, and customer prepayments continue to cover a larger portion of capital expenditures, then it would mean that what’s expensive this time is only the duration—not much to do with the company itself—and my line would have to be adjusted. Conversely, if it’s forced to switch to secured high-interest debt, or if the share of prepayments declines, the “route” CoreWeave took—where borrowing gets more and more expensive—would be right in front of it.

Back to the tape. Over the past day, $CRWVB moved mostly sideways, and peers in the same sector weren’t hammered along with it. $NBISB current price is $221.24, down 5.59% over the last 24 hours, almost back to the level of the close on August 19. This is the pre-market period in U.S. trading; liquidity is thin, so don’t treat this readout as if the market has already digested everything. $ORCLB in the same timeframe edged up modestly; the name with traditional business as a base has actually been steadier when long-end rates are rising.

Next to watch is the terms of the next financing, not the intraday volatility of the stock. Look at what maturity the company chooses, and where the coupon lands in those tiers. These two numbers explain the true cost of funding for the AI compute business better than any single intraday selloff ever could.

In the on-site video of that White House event, there’s a part that’s easy to gloss over. Someone asked the President whether the U.S. government would actually buy a substantial amount of Bitcoin and other crypto assets. He didn’t give a plan or a timeline. The gist was that this has been something people keep talking about; he’d probably rely on Paul and that whole group to make the decision first and then tell him. Paul was Paul Atkins, the chair of the U.S. SEC, and he was seated right next to him.

A few hours later, the headlines across various fast-news outlets all standardized to: the United States is considering making a major buy of Bitcoin—and the market basically moved in line with that headline.

I cross-checked that answer with the executive order dated March 6, 2025 that established a strategic Bitcoin reserve. That document spells out that the reserve would be filled with Bitcoin seized by the Treasury, and it authorizes the Secretary of the Treasury and the Secretary of Commerce to create a budget-neutral accumulation plan, on the condition that it doesn’t add extra costs to taxpayers. The government can buy—but it has to do it in a way that doesn’t cost money. That authorization has been sitting there for seventeen months, and the President’s “consideration” said last night has not moved forward even one step.

The other half—the part about other crypto assets—can’t survive scrutiny either. The same executive order writes different rules for reserves other than Bitcoin: aside from seized proceeds, the government will not go out and acquire any new assets for that portion. The half-sentence that most excited people in the headlines is precisely the one with the least basis.

If the surge last night was driven by expectations of buying into the reserve, Bitcoin should have been the one to rise the most.

What actually happened is the opposite. As of 11:00 this morning, $ETH is quoted at $2,233, up 16.84% over the past 24 hours; meanwhile $BTC is quoted at $69,072, up 7.34% over the same period. Ethereum’s relative price versus Bitcoin surged 9.69% on a single day yesterday, but before that, across the entire prior week, this ratio was basically a flat line. What the market did was to lift the price of altcoins relative to Bitcoin across the board by one rung.

To explain that one-rung move, you need to go back one day.

On August 18, the U.S. SEC released a proposal called Regulation Crypto Assets. It opened two exemption routes for token funding—under the larger tier, projects could raise no more than $75 million within twelve months. The second half of the proposal was even tougher: it provided a safe-harbor path for tokens. The founding team would permanently stop the key managerial contributions promised in investment contracts. If the network can run on its own, then the token can shed its securities status. The conditions for meeting this are written in a way that can be verified point by point. What used to require a lawsuit to reach that conclusion now becomes a checklist.

Bitcoin has never needed this path; its commodity-like attributes haven’t been seriously challenged. What needs this path is everything else.

On the day the proposal landed, the market barely reacted—Ethereum was up 0.22% at the close. The real launch happened after the White House event began last night, when the first high-volume bullish “big green candle” appeared around before 11:00 p.m. Beijing time. In the same event, the President specifically mentioned Hyperliquid, saying the Commodity Futures Trading Commission is pushing to make it enter the U.S. in a compliant, legal way. On OKX, HYPE moved from about $59.40 at the August 18 close to a little over $69 now; in two days it climbed about 16%.

Up to here is the list of things that got repriced last night: what exactly these assets count as under U.S. law. #CLARITY法案 is doing exactly that—it determines which tokens fall under securities regulation and which fall under commodities regulation. The SEC’s proposal is the technical add-on to the same issue. Bitcoin got lifted along the way; Ethereum is actually the target.

In the event, Coinbase’s Armstrong steered the conversation to September 15, when the Senate would vote on a motion to advance the bill. It sounds like a countdown.

But four days earlier, Galaxy’s research head Alex Thorn had just cut the probability of the bill becoming law within the year to 10%, and the reason had nothing to do with the White House commotion. The Senate only reconvenes on September 14, leaving a window of just two or three weeks. The ethical provisions involving officials in crypto hadn’t been ironed out, and the banking industry had been pressuring nonstop on stablecoin yield. Nothing in last night’s event resolved any of it.

Even the basic arithmetic of procedure doesn’t favor optimism. To advance the motion, 60 votes are needed; the Republicans have to get them, meaning at least seven Democrats must cross over. And at least two known Republican senators oppose the motion on substantive terms. In the conditions proposed by the Democrats, the ethical provisions directly target the President’s family’s crypto business. Once the President personally steps to the front of the stage, that knot becomes even harder to untangle.

So that’s how I interpret the jump last night. The part about the government buying coins contains almost no new information, so you can cross it out. The part about securities status is real, and the direction is correct. But the market has raised the probability that the September 15 vote will pass to a position far beyond what the procedural arithmetic supports. The SEC’s proposal still has to go through sixty days of public comment. In the middle, it can be narrowed; the next commission can also change it. The law itself can’t. This round of market action will ultimately be settled in the Senate—not in the White House.

What could overturn this view, and what would make it clear? If last night was only short-covering, then the perpetual funding rates should already be extremely tight now—but Bitcoin and Ethereum are still sitting near their benchmarks, more like spot assets swapping hands. If the relative price ratio before September 15 were to give back most of yesterday’s 9.69% and the funding rates never really rose, then it would mean that the structural change I’m seeing doesn’t exist; only positions moved.

If you follow this line, the relative price ratio might be more informative than looking only at the dollar price of any single coin. Over the next few weeks, it will answer for you: is the market really paying for a rewrite of regulatory identity, or is it only paying for a press-conference performance.

Storage chips are arguably the most profitable theme in this year’s US equities, and up until mid-August there hasn’t been much controversy. On Tuesday night, the bond market was the one that moved first. There was no bad news from the demand side, and no cloud provider came out to cut orders.

The day before’s script was completely the opposite. Musk publicly mentioned that memory is becoming a bottleneck, the storage supply chain rallied across the board, and Micron surged to near its intrayear high. One trading day later, the Philadelphia Semiconductor Index fell 4.98%, Micron dropped 7.02%, and the ADRs of SanDisk and SK Hynix fell even harder, nearing 10%. On the same day, the yield on the 30-year US Treasury briefly rose above 5.33%, the highest in 19 years. On the Binance spot side, $MUB quoted at 949.19, down 1.76% in 24 hours; $NVDAB and $SMHB also fell more than that—dumping was concentrated in storage and didn’t spread across the entire compute supply chain.

The explanation given by the US stock market is valuation compression. Higher interest rates raise the discount rate; the farther a company’s profits are in the future, the harder it gets hit. Storage has risen the most this year, and therefore it also falls the hardest. That view is correct only halfway. Micron is up 255% year-to-date—sounds like a bubble—but its forward P/E ratio is only 6x, and its gross margin last quarter was 84.9%. You don’t need a story to justify a price like this.

Where I disagree is in the second half. Treating interest rates only as the discount rate would miss a new “connection” that has just grown out during this storage cycle—interest rates are now also a demand variable.

In the past couple of years, the money cloud providers spent on compute and memory mostly came from cash they generated themselves. This year has changed. By FactSet’s definition, the share of interest-bearing debt in capital expenditures for these five cloud providers was 9% in fiscal 2024; by mid-year, based on trailing twelve months, it has already become 32%. Total capital expenditures for the same group of companies in the current fiscal year are expected to exceed $690 billion; aside from Google and Microsoft, free cash flow for the other firms is projected to drop to around zero, or even turn negative.

For the hand buying Micron stock, one out of every three dollars is borrowed. The cost of borrowing has just hit a 19-year high—this isn’t just something that exists in valuation models.

The credit market started pricing this earlier than the stock market. Oracle’s 5-year credit default swaps jumped to 203 basis points, the most expensive in 18 years, up nearly half over the course of the year. Meta’s protection cost also roughly doubled. The outstanding notional size of CDS for large tech companies set a record—Oracle alone accounts for more than half. People buying this protection likely don’t believe Oracle will default; what they want is a hedge tool that can be entered and exited at any time, to bet on whether the #AI资本开支 chain will snap.

The evidence for the bulls is also solid. Bank of America on August 17 reiterated its buy rating on Micron. The reason: information disclosed by the SanDisk analyst day suggests this storage upcycle has a structural component, unlike the ordinary upcycles of the past two years. Micron CEO Sanjay Mehrotra is even more explicit: he says the timing when supply will catch up with demand is still not visible. In this quarter, DRAM contract prices are close to doubling; Micron’s 2026 HBM capacity has already been sold out under fixed-price contracts—these are all on the books.

On the other side, Steve Eisman—who became famous for shorting 2008 mortgage loans—told CNBC that as soon as cloud providers start cutting capital expenditures, the market will move straight down. When you apply that line to storage, it lands even harder. Cloud providers’ capital expenditures are storage’s orders.

For Tuesday’s selloff, I think it was rational—only the market attached the wrong label to it.

The orders for 2026 are already locked in. Fixed-price contracts won’t disappear just because bond yields move up by dozens of basis points. What changes is the 2027–2028 segment—those orders depend on customers continuing to issue debt to be bought. How long can the storage upcycle last? Whether the investment-grade credit market is willing to keep “tying the door shut” to it now—this structure has not appeared in prior storage cycles.

It isn’t hard to overturn this view. If, after financing costs rise to this level, cloud providers continue to raise capital expenditure guidance, it suggests they can still comfortably absorb the money; then interest rates revert to playing their role as only a discount-rate variable. If Oracle’s CDS curve starts moving back, it would imply that the credit market has loosened its stance, and the reasoning above would be invalid.

There’s another angle I might be narrowing too much. Storage is a cyclical stock; a gross margin of 84.9% hasn’t been sustained long-term at this level in this industry. Once supply catches up, the “knife-edge” position will shift from valuation to earnings itself. Interest rates are just the noisiest variable right now—other variables are queued up behind it.

The next point in time that can give the answer is Nvidia’s earnings report on August 26. The revenue numbers probably won’t look bad. What to watch is how management describes customers’ purchasing cadence and payment terms in the guidance—whether this is a demand-side problem or a financing-side problem will show up in the wording. If you want to track this theme, you can also add Oracle’s CDS quotes and the 30-year Treasury yield to your watchlist—they will likely move before the chip company’s quarterly report does.

BlackRock released a research note yesterday, putting a label on this round of Bitcoin’s decline. The world’s largest asset manager said that the drawdown from historical highs this time is a position adjustment. As an emerging substitute for global currencies and a diversification tool, Bitcoin’s core investment logic has not changed.

The line circulated on Twitter all day, and people who got trapped as well as those watching from the sidelines each heard what they wanted to hear. But defining the cause is one thing; “position adjustment” and the behavior pushed out by weakening buy pressure are completely different. If the issue is simply in the chip/position structure, then it’s essentially the same asset at a 50% discount—what has been dropped will eventually have to be repaid. If, on the other hand, the side with buy demand really shrank, then after the “half-price” adjustment it becomes a new reasonable position; it won’t wait for a fix. So this statement should be broken apart and tested.

BlackRock listed crypto-native leverage as a main driver, and this point holds up in the data. In October 2025, open interest in the overall Bitcoin futures market briefly exceeded $90 billion. On October 10, when Washington announced a new round of tariffs on China, forced liquidations wiped out about $20 billion in open interest within a single day. That selloff was a positioning accident and had little to do with Bitcoin’s story itself.

This level today already looks nothing like what happened back then. Binance’s BTCUSDT perpetual funding rate in the latest period is 0.0024%, and over the past month it has stayed around this range. Longs basically don’t have to pay any premium to hold positions. Open interest for the same contract is equivalent to $6.8 billion; over the past month it has been mostly flat, and leverage has not rebuilt. The current price at $BTC is around $64,400, up 0.46% over the last 24 hours. The historical high was $126,199 in October 2025, implying a drawdown of 49%. The low of this leg was touched on July 1 at $57,800, after which the market has continued to trade in a tight range.

So I agree with BlackRock’s diagnosis of the underlying cause for the first half of this decline. In the first part, leverage and positioning were indeed clearing, and there’s no sign that anyone collectively changed their mind about the long-term Bitcoin story.

The problem lies in the second half. The term “position adjustment” hides an unfulfilled implication: after clearing is complete, the price should move back, because buyers are still waiting where they were. This part can’t be backed by data right now.

The US spot Bitcoin ETF fund flows have already completed a round trip within two weeks. The week from August 3 to August 7 saw net inflows of $854 million—the best week since April 17. The bulk of the money that went in was BlackRock’s own IBIT, and the market at the time interpreted it as institutions returning. Immediately after that, the week from August 10 to August 14 saw net outflows of $390 million, the largest single-week outflow since the end of June. Put together, the two weeks show that buy pressure has not yet formed continuity.

The signals from Glassnode are even colder. Bitcoin spot trading volume has fallen to the lowest level since 2019, and exchange deposit/withdrawal volumes have retreated to the quietest position in recent years. They describe this state as participants’ indifference, which is common in the middle of a bear market. I cross-checked with Binance’s own data and the direction is consistent: the monthly spot trading value of BTCUSDT is down more than 60% compared with last October, and after most of August has passed, it is still following the same rhythm.

Citigroup’s report dated July 1 was speaking from the buy-demand side. They cut their 12-month target price for Bitcoin from $112,000 to $82,000, and at the same time set their assumption for net inflows into future one-year spot crypto ETFs to zero. The reason: weakening investor interest, ETF flows turning negative, and stalled US digital asset legislation. This is already Citigroup’s second cut this year. If you isolate the “set to zero” assumption, it actually doesn’t clash with BlackRock’s story. BlackRock explains where the selling pressure is coming from; Citi calculates how much buy demand is left. The two sides are talking about the same thing—just from different angles.

There’s another layer readers need to keep in mind. BlackRock isn’t a commentator on the sidelines. IBIT is the largest spot Bitcoin ETF in the world, and in that best week for the market, most of the money went into—its own account. This doesn’t mean the report’s content is invalid; the portion about leverage has already been tested with data just now. It’s just that when an institution gives a long-term characterization of an asset class it manages, readers have reason to demand something harder than characterization.

Putting both sides together: BlackRock’s explanation of what has already happened is accurate, but its inference about what will happen next is missing one link. Once leverage has cleared, it only ensures that further declines are no longer driven by forced liquidations—it does not guarantee that prices will move back on their own. The missing link is continuous spot buying pressure. Until now, neither trading volume nor ETF flows have provided it. Until it appears, #Bitcoin is more likely to keep grinding within this range.

What could overturn this view? If over the next three to four weeks ETFs show consecutive net inflows, and if spot trading volume rises from that 2019-like level, it would suggest that buy demand really is returning, and the paragraph above should be invalidated. Conversely, if open interest starts moving clearly higher while spot trading volume stays pinned to the floor, then any rise would be driven by leverage—something that has already been demonstrated last October in terms of how such a rally ends.

There’s one more variable to watch tonight. The Federal Reserve’s July meeting minutes will be released tomorrow early morning Beijing time. In that meeting, the rate was held at 3.50% to 3.75%, and three regional Fed presidents voted against a rate hike. If the minutes’ hawkish wording exceeds market expectations, then among the volatility factors BlackRock listed itself, the Fed would become the dominant variable again. This can’t be solved by internal clearing within the crypto market.

In the next few weeks, you can watch whether the two lines—ETF weekly flows and spot trading volume—start to turn in sync, which may be more useful than staring at the price.

In the quarterly holdings report that Druckenmiller’s family office filed, a Nasdaq-listed company appeared that is linked to Hyperliquid—something it hadn’t previously held. The filing was submitted on Friday as scheduled. On Monday at the open, that stock jumped up by a noticeable amount. Overnight, English-language headlines in one voice turned it into a legendary macro trader betting on Hyperliquid. That statement isn’t entirely wrong, but it drops several key assumptions—assumptions that ultimately determine how much weight this position should carry.

First, let’s look at what he bought. The company is called Hyperliquid Strategies, Nasdaq ticker PURR. Its core business is holding and collateralizing $HYPE . The family office’s reported ending market value is $23.15 million, accounting for 0.44% of its entire portfolio. There’s a trap to get around first: on the Hyperliquid blockchain, there is also a PURR token with the same name, which has nothing to do with this stock. People in the English-speaking market have already mixed the two together. Second, these kinds of filings only disclose long positions in U.S. stocks. A macro trader’s on-chain holdings or whether they short to hedge simply aren’t covered in the document. Inferring that big money started allocating to HYPE by using this to connect two separate things is a logical jump.

The time aspect also needs to be made clear. The report reflects positions as of June 30, but it wasn’t made public until mid-August. In the two-month gap, he may have added to the position—or he may have exited completely. What the outside world sees is always an expired snapshot. In the same quarter, he also opened a batch of new positions, and the overall portfolio expanded significantly. PURR was just one of the names. At 0.44%, this is a probing allocation for someone of his size—orders of magnitude smaller than the weight he has historically put behind positions he truly leaned into.

Next, look at the company’s books. Its most recently disclosed fiscal-quarter reported net income is $152.5 million. The number looks impressive at first glance. But when broken down, only $2.6 million comes from collateralized income; the rest is almost entirely unrealized gains from HYPE price appreciation—leaving little, if anything, on a per-share basis. Push three quarters earlier and the company as a whole was still losing money. It almost doesn’t generate operating cash flow on its own, and the market has understood that. Its current market cap is under $900 million, yet the holdings it disclosed at the end of April were 20 million HYPE tokens—which, at current token prices, are already worth more than $1.1 billion. What the market is willing to pay for this company is less than the value of the pile of coins it holds. This year, several crypto treasury companies have traded below their net asset value, and PURR is only one of them.

So that purchase looks more like buying a discounted exposure to a bucket of HYPE rather than a sign of strong conviction in any particular narrative.

The second line concerns Hyperliquid’s expansion into the prediction market—and this matters far more for HYPE’s long-term valuation than any single holdings report, besides being the easiest area to get distorted by mismatched wording. HIP-4 is an event contract framework that launched on May #Hyperliquid . It lets people bet on outcomes in the real world—such as BTC price and macro data—on-chain in a fully collateralized manner. Galaxy’s research said that in its first month, it captured roughly 20% of the daily trading volume in BTC prediction markets combined on Hyperliquid and Polymarket. Another metric is that on launch day, its share of the entire prediction-market order book was under 1%. Both numbers are true—the difference is in the denominator. The first counts only BTC binary contracts and compares them only to Polymarket; the second includes Kalshi and all non-crypto event categories as well. Which one someone cites basically reveals what they’re trying to prove.

The expansion pace also has hard constraints. Builders deploying event contracts on HIP-4 must first collateralize 500,000 HYPE and then lock it in long-term. That threshold limits how quickly the category slate can grow. In the first phase, what can be traded is still mainly crypto price binary markets. Categories that could truly expand the market—sports, macro data, and the like—have to wait for the next phase of permissions.

The most weighty bullish statement on the long side came from Matt Hougan, investment director at Bitwise. Over the past couple of days he posted on X that people think Hyperliquid is a crypto app; back then, people also thought Amazon was a bookstore. The remark clearly carries an agenda, and Bitwise itself has issued a spot HYPE ETP.

On the direction, I’m on his side. Hyperliquid consolidates perpetuals, spot, and event contracts into the same matching engine and the same liquidity setup. There’s currently no second company with this structure. Using an exchange valuation framework to value it may indeed end up undervaluing it. But I don’t agree that Druckenmiller’s position should be used as supporting evidence. This is a probing allocation. The report is also of a two-month-old snapshot, and it only shows the U.S.-stock side—so it can’t support conclusions that large.

The real test is the next phase of HIP-4. If non-crypto categories start producing truly substantive trading volume and open interest, the Amazon analogy would hold up. If, three months later, the tradable markets are still only those few BTC binary contracts, then the claim that prediction markets are becoming the “second pillar” of the ecosystem would need to be reeled back.

HYPE perpetuals on Binance are a little above $59, and they’ve barely moved over 24 hours—so this news hasn’t left any footprint on the token price. Instead, it was the PURR stock that jumped on Monday, with all the attention concentrated on the stock side. This money bought a U.S.-listed container, while the on-chain setup is still one layer removed from it.

Rather than obsessing over who entered, it’s better to check after some time what categories can actually be traded on HIP-4 and how many open positions there are in each category. The category list is more honest than the holdings report—it at least reflects what’s true right now.

The US stock market hasn’t opened yet, but storage chip stocks have already been bought up. This weekend, the U.S.-stock capital on Binance almost only took one direction: SanDisk, Micron, Western Digital, and Hynix all surged across the board, while the broader market didn’t move along. Before Monday’s open, someone specifically added to positions in storage.

Counting from the Friday U.S. stock close, $SNDKB is up 8.6%, leading the sector; $MUB has broken through the $1,000 level, and the tokenized versions of Western Digital and the storage-theme ETFs have also each risen by nearly 5 points. SanDisk traded over $43 million in volume over the weekend, surpassing SpaceX to rank first among all bStocks; meanwhile, $SPYB just went sideways. These orders are essentially betting on one thing: storage stocks will gap up on Monday.

To understand this bet, you need to look back at how wild this sector has been over the past month or so.

This year, storage has been the most frenzied line in the AI trade. Data centers’ appetite for memory is growing faster than capacity can. Prices have been soaring nonstop, and JPMorgan even coined a term for this phenomenon: chipflation. The DRAM market rose by 30% quarter-over-quarter for two straight quarters. Samsung itself has said that the supply shortage could last until 2028. Micron has gained more than two times within the year, and at one point it was the most crowded long trade in all U.S. stocks.

SanDisk’s position in this cycle is somewhat special. It’s a pure-play flash memory maker. In the first half of the AI rally, HBM and DRAM were the focus, so money flooded into Micron and Hynix, leaving flash as the “lagging second.” Then the shift happened this year: data centers started stacking enterprise SSDs for AI training and inference, and a flash shortage also emerged. SanDisk, once considered a latecomer, squeezed its way from catch-up mode to become a leader. And again this weekend, it is the one charging at the front. The supply side also helped: during the prior down-cycle, flash makers cut capital expenditures to the bone. New capacity ramps up over years, so when demand suddenly turned strong, the shortfall proved harder to fill than people expected.

At the end of July, this line hit the brakes hard. The market feared the cycle was topping. Micron, Hynix, and SanDisk all fell more than 20% from their highs, slipping into a technical bear market, and calls for the overall AI trade to pull back briefly took the lead. But after the dump, only a few days passed—SanDisk rebounded 26% in a single day, leaving the shorts who rushed to chase downside hanging in midair.

On August 13, an investor day reignited the rally. SanDisk management laid out growth models for fiscal years 2028 to 2030: revenue would maintain an annual growth rate in the high single digits to low double digits; gross margin targets were drawn straight around the 80% range; and any excess cash would be returned to shareholders. The stock jumped another 14% that day. The old playbook for memory stocks is that manufacturers go on a wild expansion spree at the peak of the cycle, then crush prices, and investors have taken enough losses from that pattern over decades. This guidance is management signaling in plain sight: no expansion this time—divide out the money. The aggressive buying on Binance this weekend is buying into the continuation of that message.

The disagreement between bulls and bears is sizable right now. The bulls have already laid out their case: the shortage is real, capacity can’t be expanded in the near term, and the manufacturers have set rules for themselves this time. The bears also carry weight. BTIG’s Klynski warned that this AI pullback might not be over yet. Another, more specific sell-side view: memory prices will peak within the next two quarters. If it really gets there, then Micron’s low-looking P/E—because it appears cheap—would actually be a trap; in a cycle stock, when profits are at the top, valuations often look the lowest. Over the past several decades, this industry’s script has indeed always been the same: shortage, price hikes, expansion, oversupply, collapse—no round without exception.

I’m on the bull side, but only about half a step. The mid-term logic is supported by real data—supply gaps and manufacturer discipline. But after Micron has doubled and doubled again within a year, volatility at this spot can only get worse. The kind of “clean-out” at the end of July—where it dropped 20% in half a month—would likely come again. The weekend’s upside also deserves an extra discount: it ran up from a smaller-cap venue on Binance, where depth is limited, and sentiment readings are more reliable than pricing reads. Put together, this feels more like a holder’s market: if you already have inventory, keep holding; chasing from flat cash now means you’re taking the risk of that top-of-cycle tug-of-war. When it’s my turn to act, I won’t chase Monday’s gap up. At this level, patience is worth more than speed. I’ve also written down the conditions for admitting I’m wrong in advance: DRAM or NAND spot prices turn from month-over-month increase to month-over-month decline, or if any major maker’s capex comes in above expectations—then the whole logic is void.

No need to wait too long to verify. Monday night’s U.S. market open will determine whether SanDisk’s gap-up magnitude matches the 8.6% move that the Binance weekend board delivered—these bids will be proven right away as either a head start or an overshoot. And there are two more tests ahead: Nvidia’s earnings report next week, and the Jackson Hole meeting at the end of the month. Once Nvidia clarifies its capex guidance, the order expectations for HBM and enterprise SSDs will find footing. At Jackson Hole, what the market wants to confirm is whether the rate-cut path will still keep growth stocks in the game. If you’re holding storage positions, you can mark these two dates on your calendar.