戈多Godot

I’ve seen some KOLs say that the turning point for semiconductor stocks will come in 2030.

The reasoning is that new capacity will gradually come online around 2030, easing the supply crunch, and then stock prices should fall.

If you think about it carefully, there’s an implicit assumption here: that the industry’s turning point and the stock market’s turning point will coincide. When the industry peaks, stock prices peak too.

That sounds reasonable, but in reality, umm, not necessarily.
——————
Insufficient capacity seems more like a basic condition supporting stock prices right now—it can only ensure that prices don’t plunge.

Yesterday, Samsung reported operating profit up nearly eightfold year over year, exceeding 100 trillion won. In any normal year, those results would be extraordinary.

And what happened? The stock price fell. Of course, trading factors such as overly high expectations and profit-taking played a role.

But the market is starting to focus on questions further upstream than current semiconductor profits.
For example, whether AI can actually generate profits.
——————
Yesterday, the Financial Times reported that OpenAI’s annualized revenue in September was close to $50 billion, below the roughly $70 billion figure previously circulating in the market.

There are differences in how the figures are calculated, so you can’t simply interpret this as OpenAI’s business suddenly shrinking by $20 billion.

But the market doesn’t care about the different accounting methods—it just sends the stock tumbling.

Sure, capacity is tight, but the market no longer cares. All it wants to know is whether you can ultimately turn those costs into revenue.
——————
On top of that, consider the current macroeconomic environment:

Rising yields on long-term government bonds mean a higher discount rate for future cash flows.

Debt and fiscal problems in Europe are increasing the global risk premium.

The U.S. midterm elections are adding uncertainty around fiscal policy, regulation, and industrial policy.

While none of these factors affect how many chips are sold, they can all weigh on market valuations.

That’s why semiconductor stocks are diverging so sharply right now. They’re difficult to trade, and the market is likely to go through a fairly prolonged correction.

Just discovered that, in fact, the KOSPI has already seen net fund outflows for 9 consecutive trading days.

The core asset being sold off is SK hynix.

$SKHY

$MU Risk alert for Micron: As many as 99% of actual voters in the Taoyuan, Taiwan union support a strike.

The union has even said it “won’t rule out a surprise strike”! No date has been set yet.

Taiwan is one of Micron’s key DRAM/HBM production bases.
#美联储纪要聚焦10月暂停加息 #美光

$BWXT looks like there may be a weekly-level opportunity here. Buy on-chain in Alpha. #美联储纪要聚焦10月暂停加息

The big events really start tonight.

First, the FOMC minutes offer a look at the rate-cut path. Then TSMC’s September revenue will provide a direct test of AI momentum, while Taiwan’s exports will offer another read on semiconductor demand.

French government bonds and Europe are also worth watching.

Finally, there’s the 30-year Treasury auction and the University of Michigan’s inflation expectations.

Rates, AI, and long-term bonds are all in play—and together, they could set the market’s next direction.

NVIDIA is discussing with insurance companies to provide insurance for GPU-backed mortgages, helping small AI cloud providers finance the purchase of chips 🧐

These firms need to buy GPUs first, build clusters, and then gradually recoup their investment by renting out computing power.

When funding is insufficient, they use GPUs as collateral to borrow money.

What lenders are worried about is that if the borrower defaults, the resale value of the old GPUs may not be enough to repay the loan.

The solution NVIDIA is discussing is to have insurance companies cover part of the losses as agreed, thereby increasing lenders’ willingness to issue loans.

At present, the talks are still in the early stages and nothing has been finalized.

——————

Previously, Coatue and MatX discussed financing to support storage procurement and capacity. This shows that the financialization of AI hardware is moving forward.

The equipment value of GPUs and rental income are being used to support loans, and the storage supply chain related to HBM has also started exploring capacity financing.

Bringing in financial capital could turn potential demand into actual orders, and it may also make industry sentiment increasingly dependent on financing conditions.

When credit is easy, companies can purchase in advance.

Once computing rental rates and equipment residual values decline, both customers’ repayment ability and the collateral value may deteriorate at the same time, tightening loans would in turn affect new orders.

If NVIDIA needs to participate in providing guarantees, it would also face potential payout liabilities.

But I still want to bet on NVIDIA.

$NVDA

The House of Representatives vs. the Senate: Which one has more impact on U.S. stocks? | U.S. midterm elections and U.S. stocks (Episode ②)

First, let’s clarify the transmission mechanism: elections change expectations for fiscal and trade policies, which affects inflation, government borrowing, and the yield on the 10-year U.S. Treasury—ultimately influencing the valuation of technology stocks.

Will a divided government limit fiscal expansion, or will it lead to a situation where spending is hard to curb, deficits are hard to reduce, and policy conflicts become more frequent?

Taxation and appropriations both require approval from both chambers. Either chamber can block major legislation. The Senate is also responsible for confirming key personnel nominations.

Path 1: New fiscal stimulus is constrained, and yields may fall.

The Republican Party will roll out new tax cuts or spending plans. If the Democrats win the House of Representatives, the likelihood of these plans being passed could decrease.

That would mean easing pressure on future demand, growth, and inflation, and government borrowing could be lower than previously expected. As a result, the market may downgrade the future interest-rate path, supporting 10-year Treasury prices and lowering yields.

Path 2: Both sides compromise by increasing spending, so the deficit may not fall.

As of Sept. 3, UBS CIO gave scenario probabilities: Republicans keep both chambers 10%; Democrats win the House while Republicans keep the Senate 50%; Democrats win both chambers 40%.

Under a divided Congress, both sides may reach an agreement to boost defense spending and delay cuts to social programs.

In the scenario where Democrats control both chambers, higher taxes may still face resistance from the president. The eventual compromise could bring a wider deficit and less fiscal tightening.

Spending cuts are postponed, while revenues don’t increase accordingly. That means future deficits and borrowing needs could be higher than expected, increasing pressure on the supply of longer-dated Treasuries.

If economic demand remains strong, a relatively looser fiscal stance could also make inflation harder to bring down, reducing market expectations for rate cuts.

Path 3: Fiscal conflict raises the compensation investors require to hold long-term Treasuries.

On Aug. 18, Charles Schwab said that if the Senate also flips, personnel confirmations would face even more resistance.

For long-term government bonds, policy reversals may lead investors to demand more compensation, which is known as the term premium.

The 10-year nominal yield can be broken down into the real yield and inflation compensation. The latter is usually measured by Breakeven, which includes inflation expectations and is also influenced by the risk premium and liquidity.

Micron MU still has two earnings releases left. This time, it’s hard to significantly beat expectations.

More importantly, the market has raised its FY2027 EPS forecast from $90 in March to $156, an increase of about 73%.

Strong AI demand, tight memory supply, and the ramp-up of HBM—everything has already been priced in early.

Second, the market has raised the bar ahead of time. The midpoint of the Q4 revenue guidance is $50 billion; consensus is roughly $50.7B–$51.2B, nearing or even exceeding the upper end of the guidance.

Even if the company delivers $51B in revenue, it’s still difficult to create a surprise.

Third, the growth in the prior quarter relied mainly on higher prices: the average selling price rose by about 60%, while shipment volume increased by only low single digits.

The company has indicated that the pace of price increases will slow noticeably in Q4, and that most of the incremental capacity won’t start contributing until 2027.

Gross margin rose from 74.9% in Q2 to 84.9% in Q3, with Q4 guidance around 86%; profit margin improvement has room but the scope is narrowing.

Support at 1040 is key—if it breaks, look again at 900.

$MU

I’m not very optimistic about the upcoming market trends.

Demand for AI semiconductors remains strong, and stocks also have a lot of price elasticity.

But on the macro level, the discount rate is compressing valuations.

So I think the core of trading right now is to judge when this compression might start to ease.

In the short term, the pressure caused by rising discount rates may outweigh improvements in individual companies’ fundamentals.

Hedge funds and institutional capital will adjust their risk exposure and increase hedging based on interest rates, volatility, and the macro environment.

Even if a company’s fundamentals are excellent, continued upward pressure in U.S. Treasury yields could offset the positive impact from earnings growth by valuation contraction.

First is the midterm election. As the election approaches, control of Congress and the fiscal, tax, and regulatory paths after the election will become variables that the market repeatedly reprices.

If the Democrats win the House, policy bargaining under a divided government could intensify.

The second variable is oil prices.

Fluctuating statements surrounding the situation in Iran have caused oil prices to swing significantly.

If Brent crude keeps trading above $100, energy inflation may be harder to bring down, and it could also increase market concerns about interest rates staying at high levels.

If natural gas and electricity prices rise at the same time, they may further push up operating costs for data centers in some regions.

And last week, the U.S. September composite PMI initial reading rose from 56 to 58.4, the highest in more than five years.

Economic activity, employment, and corporate costs strengthened at the same time.

Last week, the 5-year Treasury yield briefly broke above 5%, and the 10-year yield moved further up to above 5.1%. As of September 25, they were 4.98% and 5.17%, respectively.

So the main pressures facing the bond market recently come from two sides: energy-inflation risk on one side, and economic data that is stronger than expected on the other.

The midterm election also adds uncertainty to the path of subsequent policies.

AI and semiconductor fundamentals remain very strong, but if real yields continue to rise and earnings expectations are not adjusted upward accordingly, valuations will continue to face pressure.

Here, it’s especially important to watch the 10-year real yield. A rise in the nominal yield does not necessarily mean that the real yield and the inflation expectations are rising in sync.

$NVDA seems to be forming a daily price head-and-shoulders top. First, a short position was opened in one tranche, with the stop loss placed at the right-shoulder top at 230.

Midterm elections don’t replace the president—so why can they affect the US stock market?｜US midterm elections and the US stock market ①

What exactly do the midterm elections decide? Will the president be replaced?

No. Midterm elections take place halfway through the president’s four-year term, and the main contest is control of Congress.

Congress consists of the House of Representatives and the Senate.

The House has 435 seats. Representatives serve two-year terms, so all seats are up for election. Seats are allocated by population, with each district electing one representative.

The Senate has 100 seats—two per state. Senators serve six-year terms, and about one-third are up for election every two years.

In addition, some states also elect governors and local officials.

So why does the president care so much?

Because to advance policies, the president often needs Congress’s approval.

For example, lowering corporate taxes requires bills passed by both chambers; many government programs require funding that Congress must approve; and key nominations like cabinet members and federal judges need confirmation by the Senate.

If the president’s party controls both chambers, policies are usually easier to push through. The opposing party only needs to take one chamber to add friction on legislation and appropriations.

But the opposition can’t do whatever it wants either—the president can still veto bills.

So what midterm elections determine is: how much policy the president can advance over the next two years, and how many compromises must be made.

Why does that affect US stocks?

Taxes affect corporate profits, government spending affects industry demand, and regulation influences operating costs.

For investors, beyond who wins, it’s also important to ask: which policies can actually be implemented—and whether stock prices have already priced them in.

$INTC $META

Storing real intentions is starting to enter the second half—there are now signs of a coming froth🧐

《The Information》's latest report says lending institutions are studying new financing instruments.

The plan is that financial capital would first buy up storage capacity, then resell that capacity to the actual customers who need it, profiting in the process.

However, whether the capacity can be freely resold is not yet determined.

Storage is gradually shifting from an industrial input to an asset that can be financed.

That’s a lot like data centers.

In the past, companies rented servers on an as-needed basis. Now, to ensure AI computing power, more and more contracts require upfront payments.

In NBIS’s second quarter, about 70% of newly signed contracts already require customers to pay in advance.

But once capacity starts to be financialized, it can create a kind of intermediary bubble.

《The Information》 even mentioned that executives from chip companies were at an event, where someone proactively tried to sell “HBM that’s already ready in the warehouse,” saying it could be delivered immediately.

That’s a very early signal worth watching.

So next, beyond watching the prices, we also need to see who is buying storage capacity—are they the end customers, or are they financial institutions?

$MU $DRAM $SKHY