The trade that pushed the Nasdaq 100 up roughly 20% and the Philadelphia Semiconductor Index up nearly 60% in the first half of 2026 is now giving some of it back.
Nvidia and Alphabet have each fallen about 11% from their recent peaks, wiping out an estimated $633 billion and $538 billion in market value respectively, while Taiwan Semiconductor Manufacturing Co. has slipped too, though by a smaller margin than its US peers.
None of this erases the scale of the run-up. It’s a correction inside a rally, not a reversal of it, but the size of the pullback is enough to have investors questioning how much further AI valuations can stretch before earnings need to catch up.
What’s actually driving the pullback
No single event triggered this, which is part of why it’s unsettling investors more than a clean bad-news shock would.
Reports that SK Hynix is moderating its planned expansion of high-bandwidth memory production raised questions about whether AI chip supply chains are quietly bracing for softer demand than the market has priced in.
A more hawkish Federal Reserve under new Chair Kevin Warsh has added pressure on richly valued growth stocks generally, chip names included. Underneath both of those, the more persistent theme is a shift in what investors are willing to accept as justification for AI spending.
For much of the past two years, capital expenditure announcements alone were enough to move these stocks higher.
Now, with hyperscalers guiding toward roughly $1 trillion in combined 2027 data center capex, investors are asking harder questions about when that spending actually shows up as profit rather than just top-line growth.
Why the reaction isn’t uniform across the sector
TSMC’s smaller decline reflects a structurally different position than Nvidia’s.
As the foundry manufacturing an estimated 72% of the world’s chips regardless of which AI accelerator architecture ultimately wins, TSMC doesn’t need to bet on any single winner the way a GPU maker does, and it’s still guiding toward more than 30% revenue growth for 2026 on sustained AI-related demand.
Nvidia’s sharper drop is tangled up with a narrower, more direct threat: its biggest customers, including Alphabet and Amazon, are increasingly deploying their own custom AI chips alongside the Nvidia hardware they still buy, raising questions about how much of future AI infrastructure spending flows to Nvidia specifically versus in-house silicon.
That said, Nvidia’s actual market share in server GPUs has barely moved, sitting near 97% at the end of 2025, which is why several analysts are treating this less as a verdict on Nvidia’s business and more as a valuation reset after a stock that returned over 1,100% since late 2022 finally cooled off.
The clearer dividing line forming across the sector isn’t AI winners versus losers, it’s suppliers with structurally hard-to-replace positions holding up better than those facing direct architecture competition from their own customers.
Hyperscaler earnings later this month, and whether their capex guidance still matches the spending numbers that fueled this rally in the first place, will be the next real test of which side of that line each stock actually belongs on.
Source: Official TCS Press Release, "TCS Begins FY27 with Continued Growth; Wins Multiple AI Transformation Deals"
