Wednesday's Fed meeting went the way I expected, and the market took it worse than I expected. The committee held, but the 2026 median dot rose to 3.8% from 3.4%, and nine of eighteen officials now project a hike this year. Stocks had their worst first-Fed-day for a new Chair since 1994. Accenture then fell 18% on Thursday, its worst day on record, after bookings shrank. SpaceX has now fallen for three straight sessions. This morning, Alphabet, Amazon, and Microsoft are all down 3% to 5%.
Into that tape, Micron reports its fiscal third quarter on Wednesday after the close. I wrote in January that this was the cleanest way to own the AI buildout, and I still believe it. But the question this week is not whether the quarter is good. It is whether a good quarter can hold in a market that has decided to sell AI.
Why the Print Will Be Enormous
Micron guided the quarter to $33.5 billion of revenue. That number is going to be beaten, and probably by a lot. Three reasons:
- HBM is sold out. Management said in March that 2026 high-bandwidth memory supply was fully contracted. Every GPU cluster the hyperscalers are raising equity to build needs it, and there are three suppliers on earth.
- NAND is in a genuine shortage. Sandisk hit a record close above $2,300 this month and is the best-performing stock in the S&P 500 this year for a reason: enterprise SSD contract pricing has been rising every quarter. Micron's NAND business rides the same curve, and I estimate blended NAND pricing rose well over 50% in the quarter.
- Gross margin has room the street is not modeling. When both DRAM and NAND are supply-constrained at the same time, memory margins go places that look like a typo. I think the non-GAAP gross margin prints above 75%, which would be a record by a wide margin.
The guide for the August quarter is what I care about most. If Micron guides above $40 billion with HBM4 ramping, the earnings power for fiscal 2027 is so far above consensus that the stock is cheap on almost any multiple, regardless of what the Nasdaq does this week.
Why I Am Still Nervous
Because this market is not rewarding good news. Broadcom beat and fell 15%. Oracle beat and fell 11%. The buyers of AI paper are exhausted, the Fed just told them rates are going up before they go down, and the May PCE report lands Thursday morning, hours after Micron's call, and is very likely to show core inflation at its highest since 2023. A blowout quarter could be up 15% at the open Thursday and give a chunk of it back by Friday.
There is also the Sandisk problem. When the best stock in the index is up more than 500% for the year, the memory trade is no longer a secret. Positioning is heavy. That does not change the fundamentals, but it changes the path.
What I Am Watching Wednesday
| Revenue, my expectation | > $38B vs $33.5B guide |
| Non-GAAP gross margin, inflection signal | > 75% |
| August-quarter revenue guide, bull case | > $40B |
| HBM commentary that matters | 2027 supply contracted, HBM4 ramp on schedule |
| Reaction I would sell into | > +15% on Thursday with the Nasdaq down |
How to Play It
The print is worth holding through. Adding before it makes less sense, because the tape is fragile and the same fundamentals may be cheaper if the market sells the news. If the stock gaps up more than 15% Thursday morning into a hot PCE print and a Nasdaq that is still falling, that is a spot to trim and look to re-enter lower. If it sells off on a great quarter, that is a spot to add. The thesis is a two-year memory cycle, not a one-day trade.
Also on the calendar: FedEx reports Tuesday, its first print since spinning off Freight, and Nike reports a week from tomorrow. Nike is a reset story I wrote about last summer and it is still resetting.
My Take
Micron will report one of the best quarters in semiconductor history on Wednesday. The market may or may not care, because it is in the middle of unwinding the AI trade on a hawkish Fed and a flood of new AI equity. I am treating the two things separately.
Playbook: hold Micron through the print, trim into a euphoric open, add into a sell-the-news reaction. The memory shortage does not end because the Nasdaq had a bad week.