Three hawkish dissents against an unchanged policy rate is an unusual configuration, and the long end of the Treasury curve took the hint: yields rallied after the decision while the Fed's target sat still. The tightening arrived through the curve rather than the funds rate, and equity investors applied it exactly where the arithmetic says it belongs — to the cash flows that arrive furthest out. Eight of the ten largest US names closed lower, and the two that did not were a payments network and a search franchise. Chair Warsh's reiterated commitment to price stability, paired with President Trump's comment that the US will be "hitting Iran hard," removed both the policy put and the geopolitical calm in a single afternoon.
The Data Point
The most instructive number of the session was not a decline. Microsoft held its AI capital expenditure forecast unchanged — one of the first data center operators to decline to raise the number — and the stock traded up to $401.24, a print $7.84 above its own open. It closed at $391.46, down $1.89 on the day and $9.78 below the high, finishing in the lower third of the $388.75–$401.24 range. Spending discipline is precisely what the skeptics of AI infrastructure returns have been demanding, and the market granted it a two-hour hearing before rejecting it.
| Symbol | Close | Change | Day range | 52-week range |
|---|---|---|---|---|
| AAPL | $338.40 | -1.68 (-0.49%) | $337.38–$344.56 | $201.50–$344.56 |
| AMD | $429.16 | -25.46 (-5.60%) | $424.15–$456.98 | $149.22–$584.73 |
| V | $369.00 | +2.41 (+0.66%) | $357.79–$373.97 | $293.89–$373.97 |
| JPM | $345.02 | -12.29 (-3.44%) | $343.89–$357.32 | $279.10–$359.25 |
| TSLA | $298.44 | -9.00 (-2.93%) | $297.39–$309.00 | $297.39–$498.83 |
| META | $587.74 | -5.67 (-0.96%) | $582.29–$599.96 | $520.26–$796.25 |
| AMZN | $226.68 | -4.18 (-1.81%) | $226.16–$232.82 | $196.00–$278.56 |
| GOOGL | $336.55 | +2.84 (+0.85%) | $331.63–$342.49 | $187.82–$408.61 |
| NVDA | $190.31 | -6.70 (-3.40%) | $190.03–$197.06 | $164.07–$236.54 |
| MSFT | $391.46 | -1.89 (-0.48%) | $388.75–$401.24 | $349.20–$555.45 |
The valuation implication is uncomfortable for both camps. If capex restraint is read as evidence that demand visibility has shortened rather than as margin protection, then holding the line is not a bullish revision — it is an information event about the order book. That is the interpretation the close supports, and it is consistent with what happened one layer down the supply chain.
The Numbers
Advanced Micro Devices lost $25.46 to close at $429.16, a 5.6% decline and the worst performance in the group, settling in the lower third of a $424.15–$456.98 range. The stock has now taken out the $476.93 level identified in Tuesday's note on the repricing of AI capex exposure by $47.77 per share. NVIDIA closed at $190.31, effectively on its session low of $190.03, on 25.3M shares — by far the heaviest turnover in the complex. The move coincided with a second consecutive rout in Seoul, where the KOSPI fell nearly 6% after an almost 11% drop the prior session as investors dumped semiconductor exposure.
The offsetting flow is measurable in per-share terms. Visa opened $8.35 below Tuesday's close — a 2.28% gap down — then printed a new 52-week high at $373.97 and finished at $369.00, $10.76 above its open and in the upper third of the day's range. Apple also set a fresh 52-week high intraday at $344.56 before closing $1.68 lower at $338.40. Both are trading within touching distance of their peaks while the AI hardware names carry substantial multiple compression already booked.
AMD and NVIDIA closed 26.61% and 19.54% below their respective 52-week highs. Visa and Apple closed 1.33% and 1.79% below theirs.
That spread — roughly 25 percentage points of relative drawdown between the two cohorts — is the clearest available proxy for how much duration risk has already been priced out of each. The payments and devices names are being valued on cash flows that arrive this year. The accelerator names are being valued on cash flows that arrive in 2029.
The Counterargument
Two facts complicate the duration read, and neither should be waved away. First, JPMorgan fell $12.29 to $345.02, a 3.44% decline that left it near the session low, despite a steepening curve that ordinarily supports net interest margin expectations. A clean rate-rotation would have found the banks a bid; instead, the renewed exchange of attacks between the US and Iran appears to have contaminated the signal with straightforward risk aversion.
Second, Alphabet closed up 0.85% at $336.55, the strongest performer on the board, after raising its own capex projection again last week. If capex discipline were the operative variable, Alphabet should have been penalized and Microsoft rewarded; the opposite occurred. This also settles an open item: the call for Alphabet to close below $314.91 expired unfilled, and the stock now sits $21.64 above that level. What changed is that Alphabet's spending increase is being received as investment against a monetizable base, while Microsoft's flat line reads as caution. That is a judgment about demand, not about balance sheets.
The bull case on the semiconductor complex is also not trivial. AMD remains 187.6% above its 52-week low of $149.22, and nothing in the session touched reported earnings — this is multiple compression working through a discount-rate change. Tesla's new 52-week low at $297.39 tells the same story from the consumer side rather than the enterprise side.
Verdict
Stretched. Recovering the 52-week high requires $163.99 per share from Microsoft, or 41.89% appreciation, and $155.57 per share from AMD, or 36.25% — figures that a higher long-end discount rate makes materially harder to underwrite. The argument is most directly tested in Visa: a close above $373.97 within five sessions confirms that capital is rotating toward near-dated cash flows, while a Microsoft close below its session low of $388.75 over the same window confirms the other side of the trade. If Microsoft instead reclaims $401.24 on a closing basis while Visa stalls beneath its high, the duration framework is wrong and the session was simply geopolitical risk aversion. On the evidence available, the AI capex complex remains stretched.