- Lead. US semiconductor stocks sold off sharply on Monday as investors grappled with a structural concern that has dogged the sector for months: whether the circular logic of AI-funded AI infrastructure spending can hold if hyperscalers begin trimming capital expenditure plans.
- Fact. Sandisk fell 11%, AMD dropped 5.2%, Nvidia shed 5%, SK Hynix retreated 7.5%, and Micron fell 2.3%, with the moves arriving ahead of earnings from Amazon and Apple and the Federal Reserve’s July 28–29 meeting.
- Stake. The S&P 500 fell to roughly 7,400 — losing 0.17% — while AI-adjacent names bore the brunt of selling, raising questions about whether the sector’s valuations have adequately priced the risk that hyperscaler capital spending growth plateaus in the second half of 2026.
Fresh selling pressure hit semiconductor stocks on July 28 as concerns about the sustainability of AI data-centre investment resurfaced ahead of a dense earnings week and the Federal Open Market Committee’s rate decision. The declines were concentrated in names most directly exposed to large-scale AI infrastructure spending, with Sandisk’s 11% drop the steepest single-session loss among major chip names.
The specific concern driving the move centres on what analysts describe as circular AI financing: hyperscalers — Microsoft, Google, Amazon and Meta — are investing heavily in AI infrastructure partly because it enables AI services revenue, which in turn finances further infrastructure. If any of the major hyperscalers signals a pullback in capital expenditure guidance, the feedback loop reverses. Recent trading in Nvidia shares has reflected that anxiety: the stock fell even after Alphabet raised its capex guidance, suggesting investors are pricing in a scenario where increased overall AI spending does not translate to proportional chip demand because hyperscalers are developing more custom silicon in-house.
Custom Silicon as the Structural Threat
Beyond the near-term capex debate, Monday’s selloff reflected a longer-running structural concern. Google’s TPUs, Amazon’s Trainium and Inferentia chips, and Meta’s MTIA inference accelerator are all in active deployment and represent a gradual reduction in each hyperscaler’s reliance on off-the-shelf Nvidia GPUs. Analysts at multiple banks have raised the possibility that this trend, rather than a slowdown in overall AI investment, poses the more durable challenge to Nvidia’s dominance of the AI accelerator market.
AMD’s 5.2% decline is notable in that context: the company has positioned MI300-series chips as a lower-cost Nvidia alternative, yet Monday’s move suggests markets are treating the entire merchant silicon ecosystem as vulnerable to the same custom-chip displacement risk. For more on prior AI-driven chip market swings, Alphabet and Tesla triggered a comparable Nasdaq selloff earlier this month on similar AI capex concerns.
Context: Earnings and the Fed
The timing of Monday’s declines is significant. Amazon and Apple are both scheduled to report quarterly earnings this week, and their AI infrastructure spending disclosures will serve as a direct data point on whether hyperscaler budgets are expanding, flattening or contracting relative to earlier guidance. The Fed meeting on July 28–29 adds a macro overlay: most economists expect the FOMC to hold its benchmark rate in the 3.50–3.75% range for the fifth consecutive meeting, but investors are watching Fed Chair Kevin Warsh’s language for any signal about a September move in either direction. In an environment where rate-hike risk is not off the table, higher-multiple growth names — including chip stocks trading on expectations of sustained AI demand growth — face a compounding valuation headwind if the Fed holds the line on restrictive policy.