To The Moon

The indices: the selling stayed inside technology, so the Nasdaq fell while the Dow closed higher

The three indices split on Thursday: the $NASDAQ(.IXIC)$ fell 1.25 per cent to 27,193.34, the steepest of them; the $S&P 500(.SPX)$ fell 0.47 per cent to 7,765.36; and the $Dow Jones(.DJI)$ rose 0.10 per cent to 51,231.60. The 10-year Treasury yield came back from 5.28 per cent to 5.23 per cent on the same day.

Rates fell and equities fell with them, which is not the chain of cause and effect of recent days: what was repriced was the assumption about demand along the AI chain, and the discount rate had nothing to do with it. That the Dow could close green says the selling stayed inside technology rather than spreading into a market-wide move out of risk. Two paths follow. Should the falls over the next few days stay ringed around AI-related names, this is a repricing of a single story. Should cyclicals and financials start weakening too, what the market is revising is not the AI arithmetic but its appetite for risk altogether. The thing to watch is whether the Dow keeps separating from the index.

OpenAI: annualised revenue of about US$50 billion, some US$20 billion below what had been expected

The trigger for the selling was a report on OpenAI's revenue: annualised revenue of about US$50 billion, roughly US$20 billion less than the market had assumed. The names selling compute followed it down: Nvidia fell 2.94 per cent to US$230.48, Broadcom 4.35 per cent to US$360.14, AMD 3.90 per cent to US$620.68 and Nebius 7.35 per cent to US$219.71. Oracle fell 5.48 per cent to US$135.69, and it had debt news of its own that day, which the last section covers.

Why should one private company's revenue figure knock down that whole row? Because a large part of what that row has sold ultimately has to be paid for by buyers like OpenAI — what it takes in decides how much compute it can contract for. What is priced along this chain is the scale of those buyers' purchases over the next several years, not what has already been delivered this year. Two paths follow. Should a firmer revenue definition or an official statement close the gap, this fall was sentiment magnified. Should purchase plans actually be cut, what has to be recalculated is next year's orders along the whole chain. The thing to watch is whether the order disclosures from these sellers change.

Storage: Samsung posted a record profit, and it still did not hold Micron up

Samsung's record quarterly profit did not hold storage up on Thursday: $Micron Technology(MU)$ fell 4.79 per cent to US$1,035.84, $SanDisk Corp.(SNDK)$ 4.90 per cent to US$1,609.46 and $SK hynix(SKHY)$ 4.35 per cent to US$170.50.

The reading in public reports is that while the absolute profit is a record, the slope of the growth is flattening, and the pace of memory price increases is slowing as well. Both point at the same question: what the market buys is not how high any one quarter's profit is but how many more quarters the price increases have left. Two paths follow. Should the next set of results show memory prices and shipments still rising together, this flattening is an illusion created by a higher base. Should prices stall first, the durability priced in today has to be pulled back. The thing to watch is how memory prices and shipments move next quarter.

Intel: the Terafab story turned around, and the shares still fell with the group

$Intel(INTC)$ fell 5.34 per cent to US$107.08. The news was, if anything, favourable: Intel's chief executive Lip-Bu Tan said the company would keep working on Terafab, and Tesla's chief executive confirmed that TSMC might only sublease part of the fab rather than own it.

That runs opposite to Monday's news that Musk had confirmed talks with TSMC, which the market then read as squeezing Intel's position; on this telling Intel has not been displaced. The shares still fell with the rest of semiconductors, and its own favourable news did nothing to stop that. Two paths follow. Should Terafab produce a formal arrangement with capacity and a term attached, Intel's position is settled. Should there be only statements from both sides, it stays a line that swings with the news. The thing to watch is whether there is a formal disclosure.

AI debt: the bond market started repricing what it costs to borrow and buy chips

There was one more thing the same day, over in the bond market: public reports say large-scale AI-related borrowing is hitting technology-sector debt, with traders repricing that risk; among the reasons given for Oracle's fall, alongside the OpenAI revenue, was the debt it has added to buy AI chips.

These are two sides of the same thing. Borrow to buy chips, and the repayment has to come from the revenue those chips generate; when the buyer's revenue figure is revised down, what creditors have to ask is whether the money comes back, and that shows up in bond prices first. Two paths follow. Should the spreads on this debt steady, the fall on the equity side was sentiment. Should spreads keep widening, funding costs press directly on the size of the next rounds of purchasing. The thing to watch is where spreads on this technology debt go.

This is personal analysis and not investment advice.

💬 【Talking Point】

Intel's own news on Thursday was favourable and the shares still fell 5.34 per cent with the group. When does a company's own story stop mattering?

💰 【Bounty】

Drop your view in the comments and there are Tiger Coins in it for you! 🎁

🔔 Better shared than saved — tag a friend and split the coins!