The indices: three of them, three directions
On Tuesday the $NASDAQ(.IXIC)$ Composite closed 0.45 per cent higher at 27,244.28, up 122.18 points on the day and a second straight close at a record; the $Dow Jones(.DJI)$ closed 0.36 per cent lower at 51,863.69, down 185.14 points, with $Cisco(CSCO)$, $JPMorgan Chase(JPM)$ and American Express leading the fallers at 4.50 per cent, 3.35 per cent and 2.62 per cent lower respectively; and the S&P 500 finished at 7,764.64, just 0.06 points below the 7,764.70 it closed at the previous session.
That list of Dow fallers deserves a sentence of its own: one networking company and two banks, none of them on the line that carried the Nasdaq to its record. So this is not a day to read as one market; it is at least two lines running separately. As for why the $S&P 500(.SPX)$ stopped exactly where it started, you would need the day's advancers and decliners and the contribution by sector to say, and a single closing level does not tell you. Two paths follow from here. Should the Dow go on being dragged down by financials while the Nasdaq goes on being carried up by the AI chain, the two have to be read apart and the index level will show nothing for a long time. Should non-financial names start turning up among the Dow's fallers, that is positions being cut across the board. The thing to watch is whether the Dow's fallers stay concentrated in financials and networking.
Memory: last the day before, first on Tuesday
The biggest gains were in memory: $SanDisk Corp.(SNDK)$ closed 6.82 per cent higher at US$1,887.04, $Micron Technology(MU)$ 5.00 per cent higher at US$1,096.16 and $SK hynix(SKHY)$ 3.45 per cent higher at US$195.37. The previous session had looked nothing like that, with SanDisk closing 1.41 per cent lower and the other two gaining far less than the chip names did that day.
But nobody was selling chips to buy memory. $NVIDIA(NVDA)$ closed 0.66 per cent higher, $Advanced Micro Devices(AMD)$ 1.34 per cent higher, $Broadcom(AVGO)$ 0.52 per cent higher, $Intel(INTC)$ 1.71 per cent higher and $Marvell Technology(MRVL)$ 1.93 per cent higher — not one name on that chain finished lower. So what changed was not the direction but which end gained more. The same split applies going forward. Should memory keep outrunning the chip names while those still close higher, the market is adding to the same chain rather than switching sides. Should the chip names start closing lower while memory carries on rising, that is a real switch. What to watch is whether the chip names go on closing higher.
SanDisk: there are contracts under this one
What sits under SanDisk's move is volume already signed for. At its investor day on 13 August it said it had signed new business model agreements with eight customers, covering about 50 per cent of its output in financial year 2027 and about two-thirds in financial year 2028; public reporting puts the weighted length of those agreements at more than four years. Rosenblatt put a target price of US$2,400 on SanDisk this week with a buy rating, and the arithmetic is stated plainly: US$240 of earnings per share in financial year 2028, times ten. Whether that figure is right is a separate question, but it is not extrapolated from a price curve.
Locking in volume cuts both ways. Should memory prices keep climbing, that pre-sold output turns into profit handed over, with most of the benefit going to the buyers. Should prices turn, it is the wall standing in front of the company: while everyone else is still following the spot quote, SanDisk's revenue is already booked into financial year 2028. The thing to watch is whether more long-term agreements get announced — the more sellers willing to sign them, the less confident those sellers are that prices hold.
Micron: guidance and a contract are not the same thing
Micron has to answer sooner: it reports its fourth quarter of financial year 2026 after the close on 30 September, having guided to revenue of US$50 billion give or take US$1 billion, non-GAAP earnings per share of US$31 give or take US$1, and a gross margin of 86 per cent. The options market is pricing a move of about 10 per cent around that report.
Guidance is not the same thing as those SanDisk agreements: guidance is what a company says about its next quarter, while an agreement is volume already signed for, so although the two rose together on Tuesday, what each stands on is not equally solid. That report on 30 September is the thing to watch. Should Micron lift both its gross margin and its guidance for the coming quarter, this memory move stops resting on spot prices alone. Should the margin turn out to have peaked, or the guidance come in merely flat, the market has to think again about whether it is buying contracted capacity or a price that is still rising.
The short side: Burry added, and the price ignored him
The other side showed up as well: Michael Burry, the investor known for betting against companies, disclosed on Tuesday that he had increased his shorts in Micron, Nebius, $iShares Semiconductor ETF(SOXX)$ and $Palantir Technologies Inc.(PLTR)$, without giving the size of any of them. $NEBIUS(NBIS)$ still closed 1.43 per cent higher at US$236.12.
Those missing sizes matter more than the list of names: without them there is no telling whether these sit as hedges alongside long positions or stand on their own as directional bets. Should they be hedges, they do not contradict the day's gains and both things can be true at once. Should they be directional, they are the other face of what SanDisk's agreements are betting on — one side that prices hold, the other that they do not. What to watch is again that report on 30 September, which answers half of the question.
Monday's winners: nobody paid up for the position again
The names coming down were the ones that had run hardest the day before: $Meta Platforms, Inc.(META)$ closed 0.63 per cent lower at US$736.60, $Alphabet(GOOG)$ 0.99 per cent lower and $Microsoft(MSFT)$ 0.72 per cent lower. Alphabet fell further than Meta, though, and three large technology names drifting down together looks more like a broad giving-back than like the market changing its mind about any one of them.
The Muse story did carry on that day: reporting has Shopify rising on a deal it struck with Muse. Yet the company that built Muse was falling too, by an amount that sits inside the same pullback as Alphabet and Microsoft, neither of which has anything to do with it, so it is hard to argue the market was redistributing the benefit here. It is more accurate to say that the price Monday set off a position in an app store was not overturned the next day — it simply stopped being paid up for. This line splits two ways as well. Should Muse hold its position and the list of companies announcing integrations keep growing, that price gets confirmed slowly. Should the position slip, the part of the price that came from it has to be given back. The thing to watch is whether Muse holds the number one spot among free apps in the US App Store, and which companies announce that they are plugging in.
The two days side by side: the price a ranking sets, the price terms set
What priced an entire processor chain on Monday was where one app sits in an app store; what priced memory on Tuesday was capacity signed through financial year 2028, and a set of results due in eight days. Over both days the buying was in the same chain, and the chip names rose on Tuesday too, so what changed was not the direction but what the market was willing to pay for: on Monday a position that can change hands within days, on Tuesday something with a date, with terms, that you can look up. Those 0.06 points on the S&P 500 say nothing about where the money went; they only say that this change happened on a layer the index cannot show.
The above is personal analysis, not investment advice.
💬 【Talking Point】
$Micron Technology(MU)$ reports on 30 September, and the options market is pricing a move of about 10 per cent in either direction. What would have to be in that report for the memory trade to stop needing the next price rise?
💰 【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!
Comments
For Micron, the September 30 earnings will be important. I want to see whether margins and guidance can support the current memory-cycle optimism. A strong report could reinforce the thesis, while weaker guidance would make me more cautious.
I am still bullish on semiconductors over the longer term, but I prefer to accumulate gradually during pullbacks rather than chase sharp moves. For me, the key is whether earnings and AI demand continue catching up with the expectations already priced into these stocks.
@TigerStars @Tiger_comments @TigerClub @Marktomarket
I’d be watching free cash flow, capex discipline and how much of future production is already committed. If Micron can grow earnings without having to keep pushing prices higher, that could be a meaningful sign that the current memory cycle has more staying power. 👀
Why?
AI agents need more computing power.
More AI usage means more demand for GPUs, CPUs, memory and data centers.
Companies like NVDA, AMD, INTC and MU can benefit from this long-term trend.
AI agents are still developing, so demand could continue growing for years.
B is not wrong. Companies like Uber, Airbnb and Schwab have strong businesses. But AI could change how customers access these services, creating some uncertainty.
Bottom line:
For a 5–10+ year investment, I prefer AI infrastructure.
But after a big rally, I would buy gradually, not chase the price.