Cryptopolitan

China has rejected a reported US effort to make other governments choose a side in the ongoing US-China AI race.
The long ongoing competition between the US and China over the lead in the artificial intelligence industry is entering a new chapter where other nations can no longer sit and watch from the sidelines.
Will countries be forced to pick between US and Chinese AI technology?
Following a reported attempt by the US to pressure countries that had backed an AI cooperation pact, foreign ministry spokesperson Lin Jian said at a Beijing briefing that China opposes forcing countries into rival groups over AI, arguing that governments should be free to pick partners on their own terms.
Reportedly, Washington wrote to about 35 signatories of a joint statement on AI partnership, telling them they could not stay in the US-led AI initiative while also joining Chinese-backed efforts.
Reuters, which reviewed an internal US draft and spoke to a US official, reported that the government is preparing to tell dozens of countries they must line up behind a US-led AI coalition or be shut out of it if they sign onto Beijing’s rival framework.
Lin called for respect for each country’s digital sovereignty and urged both sides to drop what he described as zero-sum thinking in favor of a fairer global system for governing AI.
Lin’s stance on digital sovereignty is one that has been previously endorsed by Beijing in a published paper on global cyber governance built around national control over data and infrastructure.
By framing the actions of the US government as an attack on other countries’ right to choose, China’s position as the side that doesn’t demand exclusivity will allow it to win over the same governments Washington is lobbying.
How the US-China AI dispute has escalated
In late July, China’s Ministry of Commerce urged Washington to stop threatening Chinese AI firms with sanctions, calling the pressure a form of “AI hegemony” that it said lacked factual or legal grounds.
Those threats were due to claims from US companies that Chinese labs “distill” American AI models and use the outputs to train cheaper ones.
Cryptopolitan has reported that a White House memo from science adviser Michael Kratsios accused entities based in China of running industrial-scale distillation campaigns, and that Chinese regulators are moving to bar domestic firms from taking US investment without approval.
Beijing has noted that nearly 200 US startups asked their own government not to cut off access to Chinese open-source models, warning it would hurt American competitiveness.
Cryptopolitan reported in late July that US officials added foreign-built advanced robots to a restricted list, blocking new Chinese-made robot models from being sold in the US market.
The restrictions affected robotics maker Unitree (SSE: 688836), but the company still had one of the biggest stock market launches of the year with its shares jumping 629% on its first day of trading in Shanghai. At one point, the company was valued at about $66 billion.
Meanwhile, US officials are still checking whether China has managed to obtain restricted Nvidia (NASDAQ: NVDA) chips.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Washington has a pretty obvious problem with its Nvidia (NASDAQ: NVDA) restrictions. It can stop the company’s most powerful chips from being shipped directly into China, but that does not necessarily stop Chinese AI companies from using them.
If those processors are sitting inside a data center somewhere else, a Chinese developer can potentially pay to use the computing power remotely. Nvidia makes many of the fastest processors used to train AI systems, so the U.S. has made chip export controls a major part of its effort to stay ahead of China.
Nvidia’s GB300 is among the products that cannot currently be sold into China, but some less powerful chips can still be exported there.
But Chinese companies have reportedly been getting access to more advanced computing through cloud services running from other countries, including locations in Southeast Asia.
Chinese AI companies rent Nvidia computing power from data centers outside China
The problem is current U.S. restrictions are mainly focused on who buys the chip, where the chip is shipped and who physically has it. They do not automatically stop a company in China from paying to use a machine that is sitting in another country.
Cassia King, a senior researcher on the Compute Policy team at the Institute for AI Policy and Strategy, told CNBC that Moonshot’s reported use of the facility in Thailand could still be legal “so long as Moonshot isn’t actually buying and owning the physical hardware directly.”
Cassia also explained that the U.S. system “controls physical AI chips. It does not cover remote access to those chips.”
According to CNBC, the White House defended the administration’s current approach. “The Trump administration has implemented the most rigorous export control regime in modern history, and is committed to safeguarding America’s national and economic security,” the official said.
The company operating the service also said customers do not own the chips behind the infrastructure or gain physical control over them.
“The companies we service do not have ownership, potential future claim or physical access to the chips that power our solutions,” the spokesperson said. “Any permitted access to our services, infrastructure or technology is fully compliant with all applicable regulations.”
Nvidia builds deeper ties across AI infrastructure as retail investors buy more protection
The fight over chip access is happening while Nvidia shares remain up for 2026, although the stock has not climbed as quickly as it did last year. NVDA is trading around $225, after gaining about 39% last year, based on historical pricing data gotten from Yahoo Finance.
Bank of America (NYSE: BAC) analyst Vivek Arya believes the market is worrying too much about the risks that come with Nvidia taking on more financial exposure outside its main balance sheet. Vivek kept his Buy rating and left his $350 price target unchanged, which is a 55.5% surge from current price at press time.
Bank of America is also looking at Nvidia as more than a company that sells GPUs, believing it is getting involved in more parts of the huge AI buildout and securing things that are already difficult to get.
The company also brings in money by renting out GPU capacity and generates a large amount of cash. But if Nvidia keeps using some of its own money to support other companies, projects and infrastructure around AI, investors cannot treat every dollar of free cash flow as money that will eventually come back to them.
BofA deals with that by separating Nvidia’s cash generation into two parts. About 50% is counted as cash that can go back to shareholders through share repurchases and dividends, while the other half is money that could be used for investments, financing and other spending tied to Nvidia’s wider AI business.
The smartest crypto minds already read our newsletter. Want in? Join them.

SanDisk (SNDK) perp contract on Binance has surpassed two of the largest crypto assets, Bitcoin and Ether, in trading volume.
The TradFi asset SNDK attracted more than $7.38 billion in volume over the past 24 hours. During that period, BTC and ETH saw only $6.11 billion and $4.57 billion, respectively, according to Binance market data.
This wasn’t a sudden move, actually. As of August 17, SNDK was the third-biggest perp on Binance, with a 24-hour trading volume of around $3.71 billion.
As of then, the stock contract carried $1.73 billion in open interest, 1.86 times SPCX’s $928 million and 3.51 times SKHX’s $493 million.
SNDK pushed to a high of $1,693 earlier today before settling at $1,641 at the time of writing. The price has risen over 35% from the $1,213 low in less than two weeks.
Interest in TradiFi stocks is rising on Binance
It appears that interest in stock perpetuals is beginning to overshadow that in major altcoins, at least on Binance, based on trading volume.
Of all the top 10 traded perpetual contracts on Binance in the last 24 hours, half were TradFi contracts.
Following SNDK are SKHYNIX with $2.48 billion in volume, KORU at $1.84 billion, SOXL at $1.82 billion, and SPCX at $1.35 billion, according to Binance market data at the time of writing.
Binance began introducing tokenized stocks in June under the bStock program, which is now the second-largest issuer of tokenized stocks by market cap, per earlier reporting by Cryptopolitan.
On August 13, bStock recorded a market cap of up to $610.6 million, or 22.1% of the entire sector, surpassing xStocks at $601.2 million. Ondo Finance led with $951.8 million and a 34.4% share.
The smartest crypto minds already read our newsletter. Want in? Join them.

The Singapore International Commercial Court (SICC) has authorized to freeze about S$75 million ($58 million) in Bitcoin and USD Coin (USDC) in a crypto legal battle that has rolled on for years. The episode began after a customer received coins that were never meant for them because the unnamed exchange was looking at an outdated ledger when it initiated the transfer.
Observers are now drawing parallels between this episode and the Bithumb customers who quickly withdrew tokens that were wrongfully sent to them after an employee error.
The Singapore court granted an interim proprietary injunction on March 26 in the case listed on its eLitigation service as *DVA and another v DVC* [2026] SGHC(I) 4. International Judge David Goddard, who sat with High Court Justice Aidan Xu and International Judge Anthony Meagher, delivered the rulings.
The claimants appear as DVA and DVB, the customer as DVC.
How a Singapore exchange mistakenly sent tokens to a customer
The SICC documents did not name the crypto platform litigating the wrongful transfer case, only going as far as describing the claimant as one of the world’s largest digital-asset trading operations. But the judgment revealed that the customer has used the platform since around 2013, founded his own blockchain in 2016, and set up a cryptocurrency exchange of his own.
Apparently, the unnamed exchange discontinued support for its specialized self-custody wallet product in April 2018. However, users were not immediately cut off, as there was a period after the exchange wound the service could that they could still use a third-party open-source tool to access wallets.
As for how the wrongful transfer happened, the exchange’s systems continued to show that the customer named in the lawsuit still had 2,500 Bitcoin and 2,500 Bitcoin Cash in their specialized wallets when in fact, they had already emptied those accounts in March 2020.
The first transfer, per court records, saw 2,500 BTC leave the wallet on March 2, 2020, landing in an account registered on a cryptocurrency exchange that the customer founded.
The transfer of 2,500 BCH was initiated six days later. 250 of those coins went to Binance, ruling out the exchange as a potential claimant or defendant in this case.
The problem with these withdrawals was that the platform’s internal ledger just never logged them. Hence, the exchange continued to send reminders to the customer to move their tokens for up to four years after the actual withdrawals.
A help offer from a relationship manager in June 2024 turned up an automated “remediation tool” in July that ended up with the exchange sending 2,500 BTC and 2,500 BCH of its own holdings to the customer in what appears as a classic double-spend incident.
The platform clawed back the 1,700 BTC and 2,500 BCH in the customer’s wallet when it caught the mistake on January 29 2025.
The customer has resisted the refund request on the missing balance, insisting on their claim to the disputed tokens as part of a defense strategy that disagrees with the platform’s version of events.
What the injunction covers, and what the judges withheld
The interim order bars the customer from selling, moving, or reducing the value of about 780 BTC and 816,773 USDC, along with any profits, interest, or assets derived from them. The court also ordered him to disclose where the disputed coins and their proceeds now sit, which matters because later transactions have made some of them hard to trace.
The judges did not give the platform everything. They refused, for now, to let it use that disclosure to chase similar freezes in other countries, leaving it to apply again later if needed.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Iran is weighing strikes on US military targets in Europe if Trump escalates the war.
Bulgaria and Cyprus are among the locations being considered for possible retaliation.
Tehran has also looked at hitting subsea fiber-optic cables in the Strait of Hormuz as tensions rise.
Bitcoin whales are buying again, adding about 43,000 BTC worth $2.75 billion over the past 60 days.

Maya Protocol fell victim to a hacker this week who exploited a flaw that points to a bigger issue, namely, that tools for defense have issues keeping up with sophisticated hacks. Blockchain security firm CertiK has estimated that the total direct loss was about $1.7 million as the hacker deceived Maya into awarding a nonexistent subsidy and then repeatedly added and removed liquidity to extract assets from shared pools.
The aftermath of the attack hit Maya’s liquidity providers and CACAO holders straightaway. What is more important is that the vulnerabilities were not just the direct result of a broken line of code. It took advantage of regular protocol logic behaving differently than what was expected of it, making this type of attack undetectable until it was already too late to stop it.
How the attacker gamed Maya’s accounting
According to CertiK, which identified the event on August 19, the hacker caused a misrepresentation of Maya’s internal accounting through a fake subsidy, only to finally alter liquidity positions to withdraw an estimated 48.87 million CACAO and 98.82 LINK. DefiLlama Hacks and Exploits Database classified the August 18 occurrence as “Protocol Logic,” attributing a loss of $1.7 million.
The greater harm was considerably more significant. Developer Vini Barbosa called it a “sophisticated 6-bug exploit” that took over $1.36 million in hard assets out of the protocol but had an overall impact of close to $11 million given the collapse of the price of CACAO. According to him, the token fell from $0.115 to $0.013 in less than 240 blocks and registered a drop of nearly 89%.
Aaluxx, one of Maya’s founders, recognized the damage on the same day. He added that the team will “work to fix and recover in full.”
A pattern THORChain already lived through
Maya is a friendly offshoot of THORChain, which experienced its own $10.7 million hack in May. In its post-mortem analysis, THORChain said a new node operator took advantage of vulnerabilities in the network’s GG20 threshold-signature system, and the system’s solvency checks only noticed there was an issue after the hack was over.
Aaluxx subsequently clarified in THORChain’s community podcast that the hacking was based on three older bugs that were not dangerous on their own, but when combined could prove problematic. He went on to say that Maya also had this same latent flaw.
In order to determine the cause of the hack, an extremely in-depth forensic investigation was required. The team had to examine cryptographic configuration parameters and search for insignificant prime numbers that should not be present in order to pinpoint the infected vaults. This essentially opens up a general flaw for the defenders: a simple balance monitor can potentially detect losses once the funds have already changed hands, by which point it may already be too late to prevent the attack from continuing.
What upgrading the defense actually looks like
Aaluxx cautioned that AI technology is enabling small teams to investigate codebases from a larger number of perspectives simultaneously. This is useful for defenders, but it also results in more methods for attackers to uncover unique vulnerabilities that existing auditing processes may fail to detect.
His chosen solution was redundancy rather than depending on one system. Maya and THORChain did not go with an integration approach but chose to remain independent instead. In this way, Maya could continue carrying out swaps using a verifiable healthy vault while THORChain was inactive for weeks.
The comparison is particularly interesting because THORChain’s own post-mortem says its root cause was not one dramatic flaw but three older bugs that became exploitable when chained together. THORChain also explicitly notes that the same latent bug existed on Maya but had not previously been exploited.
Metric Maya Protocol / MAYAChain THORChain Incident date Aug. 18, 2026, around 17:30 UTC May 15, 2026 Loss About $1.7 million in total attacker value; roughly $1.36 million extracted to external chains About $10.7 million drained from one vault Assets affected 20.83 BTC + 48.87 million CACAO, plus other assets BTC, ETH, BNB and Base-chain assets Root cause Six chained vulnerabilities involving the Trade Account and outbound-flow logic; the attack exploited interactions between otherwise non-catastrophic bugs Three older bugs chained together, including a vulnerability involving the GG20 threshold-signature scheme Detection / response Exploit activity triggered an emergency halt; the attacker executed the exploit before the network was stopped On-chain investigators, including ZachXBT and PeckShield, flagged suspicious activity; THORChain’s automatic controls then suspended signing/trading Protocol relationship MAYAChain is a friendly fork of THORChain, sharing much of its architecture/code Original cross-chain liquidity protocol Security lesson Multiple individually manageable bugs can become catastrophic when combined with liquidity/accounting and outbound-flow weaknesses Redundancy, automated detection, and signing controls can limit the blast radius of a vault compromise
Maya vs. THORChain security incidents
How can DeFi protocols prevent chained exploits?
Protocols need to test interactions between security controls, not just individual vulnerabilities. Maya’s incident shows how false theft detection, faulty outbound-transaction handling, and liquidity-accounting errors could become dangerous when combined. Stronger defenses include invariant testing, adversarial simulation of multi-step attack paths, independent review of accounting logic, real-time anomaly detection, and automatic circuit breakers for abnormal withdrawals or pool balances. Research published this month similarly argues for multi-layer detection and defense rather than relying on a single security mechanism.
Time / Date Event How to cite it Aug. 18, 2026 Exploit occurs. On-chain activity shows the attacker exploiting MAYAChain’s accounting/outbound-flow vulnerabilities and extracting CACAO and BTC. Maya, founder of AaluxxMyth, publicly described the hack on Aug. 18. Incident date: Aug. 18, 2026 Aug. 18, 2026 MAYAChain halts trading/transactions to contain the exploit and begins remediation. Same incident date Aug. 18–19, 2026 Exploit analysis emerges. Researchers reconstruct the six-bug chain, including the false theft alert, incorrect compensation and accounting failure. Treat as post-incident analysis, not a second incident Aug. 19, 2026 Broader reporting and technical write-ups appear. CoinDesk’s report published Aug. 19 describes the Aug. 18 attack and reconstructs the losses. Reporting date: Aug. 19 Aug. 19, 2026 Some data providers/security feeds label the event Aug. 19, likely reflecting their UTC/time-zone convention or the date their incident record was created/updated. Do not use as the primary incident date without qualification
Incident Timeline: There is a concrete timezone explanation for at least some of the discrepancy. KuCoin’s incident alert, for example, says MAYAChain was hacked on August 19 (UTC+8). That corresponds to the evening/night of Aug. 18 in UTC, depending on the exact transaction timestamp.
These block-level figures are reported in the incident reconstruction and provide a particularly useful way to connect exploit mechanics → asset extraction → token-price damage.
MAYAChain block Event Quantitative impact 17,977,941 23-message exploit transaction executed Six-bug exploit chain initiated 17,977,971 Attacker adds/withdraws liquidity from manipulated pool ~48.87M CACAO extracted 17,977,998–17,978,008 CACAO rapidly swapped into BTC 20.83 BTC moved externally 17,978,094 CACAO reaches post-exploit low About 88.7% below pre-exploit level 17,978,500+ Partial recovery begins CACAO moves back toward approximately $0.03
CACAO price + block-height attack timeline
Damage containment
Data from the industry points to the importance of reinforcing defenses. According to TRM Labs, there were 207 hacks in the world of cryptocurrency in the first six months of 2026, the largest number ever recorded in a half-year period, and it states that smart contracts are increasingly being attacked in several different ways rather than just one flaw. The Maya exploit fits that trend perfectly.
The conclusion reached is that it can no longer be possible to secure DeFi without identifying and fixing particular bugs. Protocols must involve multilayer monitoring, use of various review mechanisms independently, installation of the quick halt mechanism in case of an emergency, and the ability to implement the fixes before the questionable transaction becomes irreversible.
Maya not only wants to resolve the issue at hand. Aaluxx indicated that the company intends to accelerate the development of Aztec Chain, an omnichain DeFi project based on what has been learned from their prior projects, Maya, THORChain, and Rujira. If those lessons could help them build better solutions rather than simply restart the cycle of fixing defects, the Maya incident can turn out to be more valuable than the loss of $1.7 million suggests.

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Meme token traders are still active, but only a small handful are making profits. In the past 90 days, meme tokens have mostly extracted value from the crypto space.
A running metric for meme token traders shows that the past 90 days were profitable for only 6.25% of wallets. A new dashboard for Solana meme token traders shows the trenches are mostly losing money.
In the past 90 days, a total of 304,161 traders were active on Solana DEX. Of those traders, only 19,003 were in the green, but the median trader lost $120.
The biggest problem with meme token traders is the overall extraction of value. According to on-chain research, as a group, meme token traders lost $1.26B based on paper valuations. Of all traders in the green, 88% made under $100.
The metric tracks realized profit and loss, rather than paper valuations from holding tokens. In most cases, meme tokens have low liquidity, even if their market capitalization is high.
Only 25 Solana meme token traders made more than $10K
Only 25 wallets made more than $10,000 in earnings, showing meme space has lost its ability to produce high valuations and is instead a P2P extraction market.
Meme token traders on the FOMO app showed a similar pattern, with only 25 wallets achieving realized gains of over $10,000. | Source: Dune Analytics
The current profitability of meme tokens is even lower compared to wagering or prediction platforms, or even general speculative crypto trading. Despite this, the meme token model is still appealing, with incentives like copy trading, influencer tokens, and point farming.
The meme model has also changed, with very few long-lived memes. The newly minted assets have an extremely short lifecycle and are not aiming to create a community.
The ongoing meme activity also pushed Pump.fun to a top-3 position as a fee producer. The platform now achieves over $4.4M in daily fees as of August 19, the highest level since January.
Meme tokens still survive despite the risk of losses
Despite the overall slowdown of the crypto market, meme tokens still survive as one of the viable models. The reason for this is that meme launches happen entirely on-chain, with no limitations and waiting times for listing on centralized exchanges.
On-chain listings also partially solve the issue of curated tokens that are inflated by market makers. Meme tokens with new apps allow for instant onboarding. New users also bypass BTC and blue-chip tokens, which are now closer to mainstream finance.
Robinhood boosted its usage by offering fast meme token launches. Up to 51% of Robinhood activity was based on meme tokens. As of August 19, the trend is starting to reverse, and Robinhood memes erased 10% of their value overnight, with over 30% in losses over the past week.
Another driver of meme activity was the launch of the FOMO app, which achieved another wave of mainstream adoption. The app allowed an inflow of newcomers who still discovered meme trading, leading to almost immediate losses. The social trading feature of the FOMO app allowed users to copy-trade influencers, which led to inflated valuations for several memes.
Meme tokens are also inviting new users through the latest platform launched by crypto influencer Ansem. In this case, using the platform is incentivized by airdrops, though trading itself may lead to similar losses. As with other launchpads, users are waiting for runners, while absorbing the losses for over 99% of meme launches.

If you're reading this, you’re already ahead. Stay there with our newsletter.

The AI inference sector could be getting its own version of the long-running rivalry between OpenAI’s Sam Altman and SpaceX’s Elon Musk after AI hardware and software startup tiny corp publicly questioned the credibility of the benchmarks behind the chip that inference rival Etched shipped to Jane Street on August 18.
While Musk and Altman have publicly swung mud at each other, the accusations appear to have moved in only one direction in this inference drama between the operator of the tinygrad framework and the firm that raised $700 million at a roughly $21 billion valuation, as reported by Cryptopolitan.
tiny corp: “Etched shipped a slogan”
tiny corp framed its latest criticism of Etched as a warning to buyers and investors, writing on X that it could not rule out that the firm led by CEO Gavin Uberti did not have a working chip or was simply propping up a poor one with “smoke and mirrors.”
The tinygrad X account also conceded that “It’s possible they have a great chip and just very distasteful marketing.”
tiny corp is asking Etched to put up the numbers to back its claim. Peak FLOPS, power draw, or third-party benchmarks that don’t reveal any proprietary tech would do, according to a follow-up tinygrad post.
tinygrad fanned the conspiracy flame by saying the company could be trying to avoid making outright fraudulent claims by not touting any numbers above its unflattering figures.
The tiny corp criticism goes as far back as June 2025, when the same account urged people to alert Etched investors of what it claims was a nonsense pitch claiming “20x gains from etching transformers into silicon.”
A different critic, Wesley Yue, a robotics engineer who claimed to have held roles at Kitty Hawk and Waymo, wrote on X: “This smells very bad. none of the claims make sense,” in a July 2026 scrutiny of Etched’s technical claims.
Yue questioned how much of an achievement it was to hit a high model FLOPs utilization number when peak throughput is low. He added that spending scarce bleeding-edge wafers on lower-performance chips does not add up.
“Etched shipped a slogan,” was the tiny corp’s conclusion.
Etched has quadrupled its valuation in eight months
As Cryptopolitan reported, citing the Wall Street Journal, Etched got $700 million in fresh backing via a Jane Street-led round. The funding came at a valuation that roughly doubled the firm’s $10.3 billion rating in a Series C round that closed just one month earlier in July. The firm was valued at $5 billion in December 2025.
Jane Street, unlike the tiny corp, was impressed by the hardware that Etched delivered, according to the funding announcement. “We tested the chip and are pleased with the early results,” the firm said.
Sequoia, Andreessen Horowitz, Peter Thiel, Bain Capital Ventures, SK Hynix, and Blackstone have now participated in Etched funding rounds.
Per Cryptopolitan, Etched reported first-pass silicon success on TSMC’s N4P process, employs more than 400 people, and has booked over $1 billion in orders. It runs a 2-megawatt site in San Jose and has opened a 10-megawatt facility in Milpitas. The company has also worked to shed its early reputation for etching a single model into each chip; its systems now run any frontier model, including DeepSeek, Qwen, Mamba, and Llama.
The smartest crypto minds already read our newsletter. Want in? Join them.

Kalshi is trying to take one of crypto’s most distinctive trading products into traditional markets. On August 18, the federally regulated exchange submitted two proposals to the CFTC for approval of perpetual futures contracts based on a major US stock index and copper. If approved, the contracts would bring a structure that originated in the cryptocurrency area to stocks and commodities.
This is important as perpetual futures, or perps, have become a fundamental part of crypto trading. Researchers from Cornell University estimate that perpetual futures comprise 93% of the entirety of crypto derivatives trading. The concept behind their popularity is simple. Traders can maintain their leveraged trades without the need to roll the contracts at expiration, and at the same time receive regular funding payments which keep the prices aligned with the underlying market.
The idea behind perpetual futures is not new. Economist Robert Shiller proposed a structure for perpetual futures trading back in 1993.
“A perpetual futures contract is proposed that would cash settle every day…” — Robert J. Shiller, 1993.
It was in crypto, however, where the product gained broader recognition.
A stock index and a metal, priced off a Pyth feed
The US500 Contract by Kalshi will monitor an index called the MerQube US Large Cap Index, which consists of the largest 500 companies that are listed in the US and uses float-adjusted market cap for their weighting.
The second proposed contract is called COPPERPERP, which is used for measuring the present price of copper in dollars per pound using the XCU/USD data feed from the Pyth Network.
At present, neither of the products has received approval. Their filings fall under Regulation 40.3, which means that Kalshi must wait for the CFTC’s sanction before launching the products.
Perpetuals differ in structure from conventional futures contracts in that they do not have an expiry date. Traders can hold long or short positions for an unlimited time period, where funds are paid out to keep the perpetual aligned with the underlying price.
From a Bitcoin contract in May to equities in August
Kalshi’s transition into non-crypto perps comes after the regulatory opening that took place earlier this year.
On May 29, the CFTC greenlit Kalshi’s BTCPERP contract and released a statement indicating that perpetual contracts other than this one would be reviewed under Regulation 40.3.
Bitcoin perps launched in early June and were later joined by Ether, XRP, and several other crypto-assets.
Cryptopolitan has reported that Kalshi now offers perpetuals across 13 cryptocurrencies.
The stock index and copper filings take the same idea much further. For an exchange best known for event contracts, they are another step toward competing as a broader derivatives venue.
Why crypto traders should care about a copper contract
However, the wider issue is not copper itself, but rather if an approach to trades that is rooted in crypto can be useful for traditional assets.
In its 2026 Crypto Market Outlook report, Coinbase Institutional put forward this idea.
“Equity perps could become the preferred choice for a new generation of retail traders.” — Coinbase Institutional
Coinbase highlights constant accessibility and efficient use of funds as the main selling points. Furthermore, it is convinced that perps are no longer merely leveraged products but are gradually becoming elements of lending, collateral, and hedging systems.
In case regulated exchanges in the US manage to include perps in stocks and commodities, crypto will have exported one of its most effective market structures to conventional financial markets. Additionally, it can create a more competitive landscape for trading volumes as traditional and crypto-native venues continue to intertwine.
A small book, ramping fast
At this point in time, Kalshi’s perp business still has a limited capacity.
As reported by Cryptopolitan, Kalshi’s daily open interest registered an all-time high of $17.98 million as of August 12. By contrast, Hyperliquid had around $11.7 billion in open interest over 377 trading pairs, meaning Kalshi is, for now, only about 0.15% of that amount.
Its expansion has, however, been rapid. Kalshi’s perps hit $1 billion in notional volume in a week of launch, while it took the company’s event-contract business around 40 months to hit this landmark.
However, there is some danger involved with this expansion. CME Group filed a lawsuit against the CFTC and its Chairman, Michael Selig, in June over the approval of Kalshi as well as general policy regarding perpetual contracts. CME says that this type of offering should fall under swaps and not futures.
That case, along with the CFTC’s review of Kalshi’s August filings, could ultimately decide whether US500 and COPPERPERP make it to market.

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Bitcoin is going to move back to $100,000, but it might take up to 18 months for the catalyst to align, according to Anthony Scaramucci.
The SkyBridge Capital founder told CNBC’s Squawk Box on Tuesday that “this is clearly a bitcoin bear market,” in response to the point that the price has not really moved much over the past months.
Scaramucci says $100,000 BTC might take a while
Scaramucci said bitcoin has been stuck in its narrowest volatility band in five years for about nine weeks now, with prices roughly flat since the war in Iran started in February.
He offered three explanations for the consolidation, which include miners redirecting computing power to AI, capital flight into AI stocks, and bitcoin’s four-year market cycle.
Despite all of these, Scaramucci pointed out that bitcoin has shown more resilience than in the past bear markets. He put the current market drawdown at roughly 55%, well short of the typical 75% to 80% collapses that defined earlier cycles.
“That’s actually a good sign that there’s a lot of net buyers going into the next bull phase of bitcoin,” Scaramucci told CNBC.
The catalyst for the bull run, he believes, will be bitcoin’s supply cut in the next halving, which is still around 18 to 19 months away.
“I think you’ll see the thing move back up over $100,000,” Scaramucci said. “But it’s going to grind for a while,” he added.
Standard Chartered calls $100,000 BTC by year-end
Standard Chartered is also of the view that bitcoin will return above the $100,000 mark, but as soon as the year’s end.
The price target represents a 55% increase from the current price of $64,350, at the time of writing.
“When we look back at the end of 2026 with bitcoin at $100k we will say this was the buying zone we all wanted,” Geoffrey Kendrick, the ​global head of digital assets research at Standard Chartered, said in a June report.
BTC 24-hour price chart. Source: Coingecko
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

The U.S. will wait three days before introducing new tariffs on Canadian goods as negotiations between the two countries continue. President Donald Trump commented, “I have paused the 50% Tariffs against Canada that were scheduled to kick in tomorrow morning for a three-day period.”
He further contended that the countries have already struck a deal and are now working through the remaining paperwork.
Canadian Prime Minister Mark Carney ‌also noted that the talks have moved forward considerably, though important work remains. Tuesday marked Trump and Carney’s second call this week, anchoring previous weeks of negotiations since July. Initially, the American president had set the new Canadian levy deadline for 19th August.
Trump had hinted that a deal could renew the Keystone XL pipeline project
The two nations have mainly been divided on U.S. auto tariffs and Canada’s provincial bans on American alcohol. In a post on the White House site, Trump said Canada pledged to attend to U.S. complaints about dairy, booze, and car duties.
He also hinted that a definitive trade agreement could facilitate the revival of the Keystone XL project, an oil pipeline intended to connect Alberta to the United States that was rejected during the Obama and Biden tenures.
So far, however, the pipeline has faced long-standing opposition from environmentalists and Indigenous communities, but Trump has repeatedly called for bringing the project back. It would carry about 830,000 barrels of oil daily.
The U.S. Trade Representative Jamieson Greer’s office also gave its statement. It noted that the U.S.-Canada deal should expand market access for U.S. goods while aligning the two countries on digital trade and economic security. It further noted that the agreement will have “many important provisions that will continue to protect our market and American workers, along with our Canadian partners.”
Canada’s Carney, more recently, nonetheless, has insisted that they will prioritize creating an economy that is stronger, more self-reliant, and better able to compete globally. Though the Canadian negotiators and cross-border businesses have welcomed the three-day extension after repeatedly warning that the new tariffs would harm both economies.
Sources say the two nations were considering 15% levies
Trump’s latest proposed tariffs targeted Canadian wine, dairy, cement, apparel, and hockey gear, adding to existing U.S. levies on Canadian steel, aluminum, autos, and lumber. Canada’s focus remains on securing a trade pact that rolls back or slashes U.S. tariffs in these core areas.
Meanwhile, the U.S. wants Canada to lift its retaliatory auto tariffs and open up its dairy quotas to allow more American cheese imports. The U.S. is also asking Canada to lift the ban on American booze that most provinces rolled out last year to push back against Trump’s tariffs.
Before the Wednesday deadline, U.S. and Canadian negotiators were working on a deal that could bring tariffs on Canadian autos down from 25% to 15%, according to sources. However, insiders also claimed that the two were split over which vehicles receive the lower rates, with the U.S. holding out for cars that use mostly American components.
Furthermore, Canada must secure the cooperation of provincial premiers to resume the sale of American alcohol, as liquor regulation falls strictly under provincial, not federal, jurisdiction. Ontario Premier Doug Ford, whose jurisdiction is most heavily impacted by United States automotive tariffs, has this far, indicated a willingness to rescind the restrictions only upon achieving an equitable agreement.
The three-day pause gives negotiators a narrow window to resolve the remaining disagreements before the higher tariffs potentially take effect. A failure to reach an agreement could revive concerns among manufacturers and businesses that rely heavily on cross-border supply chains.
The United States and Canada have deeply integrated economies, with billions of dollars in goods moving between the two countries each month. Higher tariffs could therefore increase costs for businesses and consumers while disrupting industries that depend on parts and raw materials from across the border.
For Canada, the negotiations also come as Ottawa seeks to reduce its dependence on the U.S. market by expanding trade relationships elsewhere.
Carney has increasingly emphasized economic resilience and diversification, suggesting that Canada wants any agreement with Washington to strengthen its position rather than leave the country vulnerable to future tariff threats.
The next three days could therefore prove critical in determining whether the two sides can turn the latest progress into a broader and lasting trade agreement.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Small batches of Nvidia’s H200 AI accelerators have reportedly started arriving in China following the relaxation of Washington’s licensing rules. This development comes at an uncomfortable time for Nvidia since, although Chinese demand for its products is high, the company’s share in the world’s second-largest chip market is dropping below ten percent.
The situation has effects that extend beyond Nvidia. The company’s GPUs enable much of the training of some of the world’s most sophisticated AI models, which means that decisions regarding who gets access to them are going to shape where the next generation of AI technology is developed.
Case-by-case licenses replaced the outright ban
Earlier, the Bureau of Industry and Security of the Department of Commerce announced that export applications for Nvidia’s H200, AMD’s MI325X, and similar processors would be considered individually rather than being banned altogether. This shift happened after President Trump announced on December 8, 2025, that approved Chinese clients could purchase H200 chips if they abide by certain rules.
Exporters will continue to face stringent regulations. It is essential that they convince the authorities that their sales will not lead to a reduction in supply for American clients. In addition, it is required that the Chinese buyers follow export regulations and conduct customer verification procedures, while the chips undergo independent testing in the US.
“Export controls should evolve with changes in technology, while protecting national security,” Under Secretary of Commerce for Industry and Security Jeffrey Kessler said.
Beijing has set its own limitations. Cryptopolitan reported in July that companies such as Alibaba, ByteDance, and DeepSeek were poised to receive the chips, but authorities would only allow orders of about 200,000 H200 chips to be fulfilled, which is significantly less than requested, and they are also limited to AI training on public data only and not for inference or any use in sensitive workloads.
Nvidia budgeted for zero China revenue
Nvidia’s projections are not predicated on a substantial recovery in China.
On May 20, the company announced its results for the first quarter of FY 2027. The company’s revenue reached an unprecedented $81.6 billion, representing an increase of 85% compared to the previous year, while the Data Center revenue reached $75.2 billion. Nevertheless, the management’s guidance for the following quarter, which is around $91 billion, did not incorporate any Data Center compute sales to China.
CEO Jensen Huang has opened up about the reason.
“Huawei is very, very strong,” he said in May, and he claimed that Nvidia had “largely conceded that market to them” after enduring years of increasing restrictions in the US.
For a company at the hub of global AI infrastructure spending, not including China in their forecast shows how much the competitive landscape has changed.
Domestic chips are taking the budget
In a survey cited by TrendForce, executives said Chinese-made chips are expected to account for 46% of their AI accelerator spending over the next year, up from about 30% today.
Bernstein made an even bolder estimate that Nvidia’s share of the AI chip market in China could drop to about 8% by 2026 from nearly 40% just a year before, while Huawei rises above 50%.
The change is being facilitated by government-funded investment. TrendForce noted that China intends to invest around 2 trillion yuan (about $294 billion) in data centers over the next five years, with at least 80% of the core technology, including chips, expected to come from local manufacturers.
This allows Huawei and other Chinese chip manufacturers to have a larger domestic market to develop their products, while also making big buyers, like Tencent and Alibaba, to build more of their artificial intelligence systems using domestic equipment.
Controls that helped build a rival
A March 2026 CSIS commentary found that US and allied export controls, first imposed in 2022, did slow China’s progress at the technological frontier. But the restrictions also gave Beijing’s semiconductor self-reliance push new urgency and created a captive customer base for domestic chipmakers.
At the same time, the gap in performance that those controls intended to safeguard is much smaller now. The Stanford AI Index 2026 found that the gap between the top AI models from America and China is almost gone; as of March 2026, the American model is ahead only by 2.7%.
That means that the shipments of H200s are much more important for the information that they convey than for their total numbers. While Nvidia might be able to have some sales in China again because of new permits, years of bans had already directed clients to local products. America might manage to open the door a bit, but fixing the ecosystem created during its closure will be much more difficult.

If you're reading this, you’re already ahead. Stay there with our newsletter.

In terms of total value, tokenized stocks is among the fastest growing categories within the RWA space. Year to date, the total value of tokenized stocks has gone from $683.6 million to $2.399 billion or a roughly 250% growth, according to RWA.xyz. This makes tokenized stocks the second-fastest-growing category among RWAs, just behind Venture capital, which saw growth of around 270% in the same period.
The adoption becomes even more stark when we look at onchain trading volume. Tokenized equities have cleared $9 billion in trading volume, according to Blockworks data. This number back in January stood at $1 billion and $300 million a year ago. The majority of that move happened within the span of two months.
June Is Where the Line Bends
For most of the year, the volume trend looked steady and gradual. April saw volumes reach $2.7 billion and the following month reached $3.6 billion, the kind of natural rise that reads as normal category growth. June is where the numbers grew exponentially when it doubled to $7.2 billion before growing another 25% in July. Monthly gains of this size cannot be attributed to existing users trading more, but rather on new platforms coming to the fore.
The spike in volume in June and July lines up with the launch of Robinhood Chain and Binance’s bStocks. Both the products have brought stock tokens to millions of existing users.
Distribution Changed, the Product Didn’t
For most of last year, tokenized equities lived on crypto-native DEXs and stayed under $1 billion a month. Buying Tesla exposure onchain meant finding the right pool, trusting an issuer nobody had heard of, and accepting spreads that made the trade barely worth doing. That friction is what capped the category.
What changed is where the products sit. Tokenized stock moved into brokerage and exchange front ends with millions of existing accounts, which collapsed the onboarding step entirely.
The demand underneath is an access arbitrage. Nobody is buying a tokenized Apple share because they cannot buy Apple. They are buying the wrapper: 24/7 trading, fractional size, stablecoin settlement, and availability to users outside the US who cannot easily open a domestic brokerage account. That last group is the real volume driver, and it does not show up in US equity market data at all.
Regulation stopped being the blocker somewhere in that window. Post-GENIUS stablecoin rails gave the settlement leg actual legal footing, and a friendlier posture toward tokenized securities meant issuers stopped waiting for permission that was never going to arrive as a formal blessing.
Nasdaq Is Aiming at the Same Pitch
Which brings up the awkward part of the growth story. Nasdaq plans to extend trading to 23 hours a day, five days a week, putting the largest incumbent venue directly on top of the always-on argument tokenized equities have been built around.
If a regulated exchange offers near round-the-clock access with full settlement finality and no counterparty questions, the crypto-native advantage gets narrower. What survives is weekends, global access without a US brokerage relationship, fractionalization at very small sizes, and composability with DeFi protocols. That is a real list. It is a shorter list than it was.
Nasdaq compresses the moat rather than closing it. The question the next few prints answer is whether $9 billion was the start of a curve or the top of a venue-launch bump, and August data is the first month where new-listing effects should have washed out.

According to reports, the software firm known as Temporal, which developed the “durable execution” technology that allows AI agents and other long-running applications to recover from malfunctions without having to start again, is in the process of seeking approximately $500 million at a valuation of not less than $12 billion.
If the funding round comes to an end at that level, Temporal’s valuation would have increased twofold within a period of six months. It would also indicate the gradual transfer of investor focus from only “frontier-model” developers to infrastructure providers ensuring the stable running of AI systems in operations.
This change is significant for the larger market of AI. Although model developers continue to lure in the biggest rounds of financing, companies that tackle the less visible issues that come with the mass implementation of the technology are now demanding higher market valuations. According to Temporal, OpenAI, ADP, Block, and Yum! Brands are among their clients who have deployed the technology.
A valuation that has quadrupled in a year
The worth of Temporal has surged rapidly. The firm obtained funding of $146 million in the first quarter of 2025, which valued it at $1.72 billion and then got a secondary deal of $105 million in the month of October that increased its valuation to $2.5 billion.
Andreessen Horowitz led a Series D of $300 million in February that valued Temporal at $5 billion post-money based on the participation of Lightspeed Venture Partners, Sapphire Ventures, Sequoia Capital, etc.
A valuation of $12 billion would put Temporal at over four times its valuation from about 18 months ago. The company hasn’t confirmed the news about the latest fundraising talks.
Why agent builders are paying up for durability
The reason behind Temporal’s appeal lies in the technical challenge that has bigger consequences continually.
The company describes its platform in terms of its Durable Execution service, which allows applications to “ride out” the crashes, retry certain operations on their own automatically, and continue from the point of failure as opposed to having to restart the process from scratch.
This is particularly essential with AI agents. While a simple chatbot query can only be answered quickly, an agent can operate for hours or even days; it can go through several steps, rely on external services, and even change its path depending on the outputs of its model. It is obvious that the longer the process goes on, the more chances that something will go wrong.
Temporal says demand is already showing up in its numbers. When it announced its February funding round, the company reported revenue growth of more than 380% year over year, weekly active usage up 350%, and installs up 500% to more than 20 million a month. Temporal Cloud had processed 9.1 trillion lifetime action executions, including 1.86 trillion for AI-native companies.
Venkat Venkataramani, OpenAI’s VP of App Infrastructure, said in Temporal’s February 17 announcement that “as AI systems become more complex and long-running, durability is as important as performance.”
The cost math driving the infrastructure bet
The investment case becomes clearer when the cost of running AI is considered.
On August 17, Gartner has made a prediction that the expenditure for agentic AI workflows will increase by five times by 2028. They have called this phenomenon the “Inference Paradox”, which means that while the cost of each token goes down, the total expenditure for AI increases significantly due to the higher number of tokens involved.
As per Will Sommer, a senior director analyst at Gartner, product leaders “cannot rely on more efficient token economics to rationalize AI costs.”
This means reliability is not only a technical matter, but a financial one, too. Every time a workflow fails and needs to restart, it translates to more model calls, more tokens, and higher computing costs. Temporal believes organizations using AI on a large scale will be willing to spend to prevent this type of waste.
There is evidence of a very similar sentiment among investors. Crunchbase reported that global venture funding in the first half of 2026 reached a record amount of $510 billion. More than 70% of all funding in the second quarter went to AI startups, with OpenAI and Anthropic alone accounting for $217 billion, which represents 43% of startup funding in the first half of the year.
A Temporal round at a $12 billion valuation would be more than just another substantial AI transaction. It would signal a wider trend occurring in the marketplace: investors are betting billions of dollars not only on AI models themselves, but also on the infrastructure that is necessary to ensure the functioning, scalability, and profitability of these models.


If you're reading this, you’re already ahead. Stay there with our newsletter.

On August 18, Onchain Lens, which tracks blockchain transactions, reported that Multicoin Capital had moved about 172,710 HYPE (about $10.15 million) to Coinbase Prime, a destination that has often been associated with selling by traders.
Given that HYPE is currently among the top 10 crypto tokens, the action by one of its major institutional backers is likely to attract attention far beyond Hyperliquid users.
This is important as HYPE’s market cap is almost $13 billion and is currently priced at about $58.59, as per DefiLlama. Given that market cap, the token movements are far from taking place independently. Major transfers to exchanges can quickly shape traders’ perceptions of imminent risks across the market.
A transfer that still leaves Multicoin heavily exposed
When viewed relative to Multicoin’s current holdings, the $10.15 million transfer is actually rather insignificant. Onchain Lens estimates that the fund has approximately 2.16 million HYPE left, currently valued at around $126.63 million. To put it differently, the Coinbase Prime transfer actually seems to be a trim rather than an exit.
Coinbase Prime is the institutional platform of Coinbase, which provides services such as custody, execution, and funding for the institution’s clients. It also serves as a facility through which institutions sell large amounts of their positions using the regulated market. Onchain Lens classified this transfer as “likely to sell”, thus alerting the trading community.
Why the crypto market reads exchange inflows as a warning
This anxiety isn’t without some recent background. HYPE is trading well below its June high of $76.87 according to DefiLlama; meanwhile, open interest is at almost $11.8 billion. This creates a large number of leveraged positions, which can amplify any sudden downward movements.
Traders consider the transfer of tokens to exchanges to be a sign of supply entering the market. With HYPE’s liquidity still smaller than that of big cryptocurrencies, even an experienced trader may find the order book hard to manage. After a summer of observing the movement of tokens from Hyperliquid’s biggest holders, any other Multicoin transfer keeps the market on alert.
Multicoin has insisted it is not selling
Multicoin previously opposed similar rumors. For example, in the latter part of July, after Multicoin and Paradigm unstaked a total of $291 million worth of HYPE tokens, causing the price of HYPE tokens to fall below $60, the co-founder of Multicoin, Tushar Jain, issued a statement saying, “We did not unstake to sell,” and stated that the unstaking had occurred solely for privacy and wallet rotation purposes, as reported by Cryptopolitan.
At that time, on-chain specialists at Markets Alpha confirmed this reason, showing that the tokens were moved to custody instead of going to exchanges. The recent transfer is important due to the change in destination: Coinbase Prime has a more direct path to settlement. This does not prove that the tokens have been sold, but it explains why a $10 million movement is on the front pages.
What still underpins HYPE
The negative evaluation also needs to be considered alongside Hyperliquid’s fundamental figures. In its valuation report for June, Multicoin noted that the platform earned approximately $873 million in revenue on trading volume of about $2.9 trillion in 2025, which accounts for over 59% of open interest in the DeFi perpetual markets.
HYPE’s token economics are another part of the bull case. Roughly 99% of protocol revenue is directed toward HYPE buybacks, with the purchased tokens subsequently burned. Multicoin argued in the same report that HYPE could reach around $319 by 2028.
The transfer of $10.15 million does not conclusively resolve the debate between fundamentals and prevailing short-term selling pressure. However, given that Multicoin continues to hold more than $100 million in HYPE, major events involving Multicoin’s HYPE wallets are bound to have a significant impact on market movements.


If you're reading this, you’re already ahead. Stay there with our newsletter.

OpenAI’s revenues increased by 18% in the second quarter to reach $6.7 billion, while Anthropic’s revenue doubled to more than $11.6 billion, according to the Wall Street Journal’s report on August 18.
The annualized revenue run rates demonstrate an even bigger difference in revenues between OpenAI and Anthropic. OpenAI’s revenue growth increased from about $20 billion by the end of 2025 to more than $40 billion, while Anthropic has jumped from around $9 billion in revenue to over $65 billion in late July.
Run rate is not the same as audited annual revenue, and private companies don’t publish their financials in a manner similar to public companies. However, investors are still interpreting the disparity as an indication of the movement of enterprise AI demand.
The implications reach beyond the two companies. The anticipated initial public offering of Anthropic may occur even before that of OpenAI, with investment bankers saying the first major AI IPO will help set a precedent for the entire industry, according to Cryptopolitan. This means OpenAI’s slower rate of development gives relevance to every AI company trying to justify its very high valuation.
Where the business money is actually going
The shift in the nature of business demand can be clearly seen in the valuations of the two companies by investors, with Anthropic’s last funding round resulting in a value of $965 billion, as compared to OpenAI’s $852 billion last value.
The gap could increase even more, once these firms become public, as Anthropic investors are reportedly considering the firm’s future at $2 trillion, which is a much larger figure than OpenAI’s target figure of up to $1 trillion. This allows to conclude that the value at which the companies are being traded demonstrates not only the recent growth of Anthropic but also the growing importance of the enterprise-demand gap.
The valuation gap can be better understood through Ramp’s AI Index for August 12. Anthropic succeeded in getting 43.5% of American companies to subscribe to AI services, a rise of 1.1%. OpenAI only grew by 0.23 points to get 39.7%, while xAI went up by 0.94 points to 4%.
Yet Anthropic’s newest model complicates the picture. Ramp economist Ara Kharazian called Fable 5 the best model to reach the market, but it represented only 6% of Anthropic tokens purchased by businesses in its first month and 11.4% of spending. OpenAI’s GPT-5.6 Sol accounted for 25% of its tokens and 23% of spend.
Fable 5’s cost is around $10 for every million tokens used; therefore, Fable 5 is twice as expensive as the competition. This means that effective and better performance does not automatically guarantee that companies will opt for its heavy use given the cost.
It should be also mentioned that Ramp points out that the data from models used in this case is much more technology-heavy than the general AI Index, so the outcomes obtained by Ramp must be interpreted more as partially illustrating the desire of corporations rather than as giving a complete overview of the situation.
The spending ceiling everyone is now watching
Data provided by Ramp indicates that the amount of money spent on corporate AI is still not the same for everyone. According to them, in July, the top 1% of businesses spent, on average, $7400 per employee on AI, the top 10% spent $650, while the median company spent only $11.95 per employee.
According to Kharazian, the future growth will be contingent upon the companies which are already spending a good amount of money. The companies which belong to that group also try to explore the possibility of open-source or low-price alternatives and run the risk of forcing the frontier laboratories to prove that their high price actually gives enough value to customers.
On May 11, Goldman Sachs also expressed the same worries. James Covello, who leads Global Equity Research, stated that even if semiconductor firms had reported their highest sales and profits, the majority of AI companies have not yet achieved significant profit from the sector. He regarded the situation as “unprecedented and unsustainable.”
On August 17, Gartner raised another alert, projecting that AI inference costs associated with agentic workflow will increase by more than five times by 2028. Its “Inference Paradox” explains that although tokens are getting cheaper, total costs are increasing because of the consumption of tokens by more sophisticated AI agents.
OpenAI continues to enjoy its existing advantage of scale. In April, the company indicated that ChatGPT had 900 million users per week, whereas as reported by Reuters in June, about 2 million enterprise clients of OpenAI were contributing to 40% of its revenues.
However, size by itself might not trigger the subsequent stage of competition in AI. The main focus is which corporation will be able to convert model capability into quantifiable business success at prices customers will be willing to pay. In case price discipline plays an important role in the industry, inexpensive models and platforms used for allocation of workloads among providers would gain along with frontier laboratories.


The smartest crypto minds already read our newsletter. Want in? Join them.

Meta rejected accusations on Tuesday that it deliberately built Facebook and Instagram to addict children, as opening statements began in Oakland federal court in a case brought by 29 states. California, Colorado, Kentucky and New Jersey are leading at trial.
California deputy attorney general Megan O’Neill told the jury Meta’s model was to “hook the users, hold them for as long as they can,” and then to take what those users generated and keep the company’s own findings from the public. It had worked particularly well on children, she added.
She said the states were not trying to put the company out of business and acknowledged that social media benefits some people. Meta shares closed down 4.4% at $543.67.
Meta Platforms Inc shares | Source: Google Finance
States say Meta made teen engagement a product goal
O’Neill said Meta had studied the psychology of youth seeking rewards and validation from others and used that knowledge to develop its product. She provided an example of an internal email to Adam Mosseri, CEO of Instagram, about targeting teen usage time, and how employees had referred to Instagram as a drug and themselves as pushers.
Meta’s own findings favoured early adoption, she said, the earlier the better. As for users below the age limit, O’Neill made it clear to the jury that once the company identified a Facebook user as being under 13 years old, it would disable that profile while keeping the Instagram profile active.
Paul Schmidt, lawyer for Meta, said to the jurors that there is no disagreement about some people having difficulty; however, he said that there was no proof that the use of these social media platforms by teenagers harms their well-being, and that addiction is not an agreed-upon description. He admitted that sometimes employees used casual language, but jurors would hear what the employee who called Instagram a drug did next.
He says that during the past four years, the company has identified 1.4 million accounts that displayed evidence that they belonged to someone underage. Meta has separately said the states are seeking penalties as high as $1.4 trillion, roughly its entire market capitalisation.
Kentucky Attorney General Russell Coleman called the case the largest consumer protection lawsuit in American history.
Former safety executive says Reels launched without enough safeguards
The states called Arturo Béjar first. He was an engineering director at Facebook from 2009 to 2015, whose cyberbullying work brought him public prominence, and he returned as a contractor on safety from 2019 to 2021.
Béjar testified that what the company learned about safety never made it into the products, citing eating disorder content where engineers “had very good ideas on how to make it better.” He described a pattern of releases in which safety was not weighed at all, naming Reels.
Zuckerberg and Mosseri are expected to testify. The trial is scheduled for six weeks, with an advisory jury and US District Judge Yvonne Gonzalez Rogers deciding liability.
A New Mexico judge ordered $567 million, then refused the design changes
The trial follows two adverse outcomes. As Cryptopolitan reported, a Los Angeles jury in March ordered Meta and Google to pay $6 million combined to a woman who said the platforms harmed her as a child. That same month a New Mexico jury found 75,000 violations of state law and imposed $375 million.
On August 6, Judge Bryan Biedscheid added $567 million for a youth mental health abatement fund, bringing New Mexico’s total to $942 million, and ordered five years of safety changes. He declined to order the structural remedies the state wanted, including removing infinite scroll, auto-advancing video and visible like counts, ruling that mandating design changes would run into the First Amendment and Section 230. Meta is appealing.

The smartest crypto minds already read our newsletter. Want in? Join them.

Five economists at the European Central Bank published an analysis on Monday arguing that stock market valuations are likely to correct, and that the argument holds whether or not today’s prices are rational.
The post, by Malin Andersson, Johannes Breckenfelder, Stefano Corradin, Kalin Nikolov and Maria Antonietta Viola, puts US valuations close to their historical peak on the CAPE ratio, a measure comparing share prices against inflation-adjusted earnings averaged over the preceding decade.
Euro-area valuations have risen as well, by less. Research on past technological revolutions, the economists write, points to “a worrisome conclusion.”
The views are the authors’ own and do not necessarily represent the ECB.
Rational valuations and investor exuberance point to the same outcome
This post gives both the rational and behavioural explanation. According to the rational explanation, extreme uncertainty with regard to a new technology’s efficiency would be enough to justify the valuation of its stocks at very high levels since one loses nothing much by trying it out and there is no way to tell the limit on the upside. This asymmetry creates an option value which increases price-to-earnings ratios for early adopters.
Nvidia is used by the economists as an example where the rationale of investors has been that a firm will be the next Google. This theoretical framework is based on research conducted by Ľuboš Pástor and Pietro Veronesi in 2009.
Prices can still fall from there. While a technology sits in a few firms, failure is diversifiable. As adoption spreads, the same uncertainty becomes economy-wide and can no longer be diversified, so investors demand a higher risk premium. Profits need not fall for prices to. Adoption helps cash flows, but historically the rising premium prevails unless profit growth is strong enough to compensate.
The behavioral approach goes hand-in-hand: overconfident investors bid beyond fundamentals, and once that overconfidence diminishes, the market may crash even further. “The exact moment cannot be predicted in advance,” state the economists, and “these sequences can be identified only retrospectively.”
Euro-zone households have €440bn exposure to US tech sector
Most euro-area investments in the Magnificent Seven are via mutual funds and exchange traded funds, as opposed to being direct shareholdings. Households, increasingly channelling money into low-cost ETFs, carry around €440 billion of exposure to US technology equities without necessarily being aware of the concentration risk.
Insurance companies and pension funds hold significant positions too. The holdings data is measured as of the third quarter of 2025.
The fund structure is itself a transmission channel. A market correction can compel funds to liquidate in order to satisfy redemptions, starting with their most liquid holdings and finishing with their troubled assets, thereby driving prices down and causing further redemptions. This is why the economists consider the Mag7 market correction to be a matter of financial stability for the euro area.
“The real risk,” they say, “is not only the equity market correction but one that takes place in an environment where the authorities have much less room than normal to ease monetary and fiscal policies in order to alleviate the impact.”
Europe looks less stretched than 2000, but remains exposed to a US selloff
A home-grown crash looks less likely, the post argues. Euro-area price-to-earnings ratios remain considerably lower than US levels, productivity and markups in the information and communication technology sector are rising, and the business climate in euro-area digital services does not appear exuberant.
Firms’ AI adoption is rising notably a few years after ChatGPT launched in 2022, and digital investment across the region over the past decade grew more than three times the cumulative growth in GDP.
As Cryptopolitan reported in December, the ECB drew the dot-com comparison in its financial stability review then as well, and Morningstar chief equity strategist Michael Field noted that the Magnificent Seven accounted for 40% of the Morningstar US index.
The limiting factor for such protection, however, lies in correlation. US and euro-zone stock markets have traditionally moved in tandem, and the economists predict that an AI catastrophe in the US would not just remain a problem for the US.

If you're reading this, you’re already ahead. Stay there with our newsletter.

Three engineers who once helped build rocket engines for SpaceX raised $15 million to open an automated steel-fabrication plant in Cincinnati.
Startup 1872 held a ribbon-cutting at its Factory One site on July 22, 2026. Its purpose is to provide steel components for AI data centers and small nuclear reactors.
Raptor team turns to steel skids
The three founders, Dan Summers, Brian Mongilio, and Michael Grant, all worked at SpaceX. Summers, now CEO of 1872, led the team that integrated and fabricated the Raptor engines that powered the Super Heavy booster for the Starship launch system.
He said the Raptor team paired software engineers with hardware engineers to build not just the engine but the system that made it.
That approach, he said, got Raptor from a first full-scale concept to production in three years compared with a jet-engine development cycle that he said can run past two decades.
1872 produces steel skids, which are the rectangular frames that act as a moveable base for modular buildings.
Summers described them as lower-precision parts than an aerospace component, which leaves more margin for an automated system to get things wrong and still turn out something usable.
The American Welding Society projects that the U.S. will need 320,500 new welding professionals by 2029. This demand is driven by retirements and increasing needs from data centers, chip fabs, and shipyards. Tighter immigration policy under the Trump administration squeezes welding-heavy sectors further.
“The problem that we are trying to solve is, how do we build more things with a decreasing pool of skilled labor to do it with,” Summers said.
1872 is looking to sell its skids to companies building AI data centers or small modular nuclear reactors. Both consume large amounts of fabricated steel.
Robots weld at $0.12 per inch
The welding is the result of a partnership with Columbus-based Path Robotics, whose robotic arms are used for automated arc welding.
Path Robotics says its systems get first-pass yields of 95% to 100%, the percentage of parts that go through inspection without rework or scrap.
Its arms run with the arc on about 70% of the time, versus 10% to 12% for human welders, who spend more of the task positioning and repositioning metal.
Path Robotics puts robotic welding at ~$0.12 per weld inch versus ~$0.78 by hand, an 85% reduction.
Welding a skid runs two to four hours. But assembling the cut components beforehand can eat four to five days, Summers said. “The whole name of the game is how do we keep that machine fed,” he said.
1872 describes a software stack in two parts. An “Architect” system takes a customer’s digital design files and builds a full manufacturing plan, including pricing and material sourcing.
A “Conductor” runs the floor, shuttling material and coordinating robots, which could include self-driving vehicles delivering parts or rail-mounted arms along a production line, Summers said.
He added that 1872 may reach 80% autonomous operation and just stop there if the quest for 100% stops paying off. It hopes to have its prototype factory automating most of the fabrication process by 2027.
1872 said the $15 million round, led through private funds advised by The O.H.I.O. Fund, ranks among the largest seed investments in Ohio history.
The plant sits in Camp Washington’s Spring Grove Avenue industrial corridor, inside the former David Hummel Building Company site, a 1903 structure whose namesake worked on Cincinnati City Hall and Union Terminal.
“Camp Washington has been one of Cincinnati’s most important industrial districts for more than 150 years,” said Jill Meyer, a founding partner of The O.H.I.O. Fund, who added that 1872 shows “its next chapter is being written right here.”
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.