Monday's chip-stock selloff, triggered by renewed US-Iran conflict over the Strait of Hormuz, hit the AI-infrastructure trade the way it always does when geopolitical risk spikes: hardest, fastest, and first. Intel, AMD, Broadcom and Nvidia all fell in sympathy, and the iShares Semiconductor ETF dropped 3% -- almost exactly the shape of the drawdown that followed February 2026's initial US-Israel strikes on Iran, the first major geopolitical shock of this AI cycle.
That February episode is instructive precisely because chip stocks recovered within weeks, and the underlying capex story never actually paused. Technology executives told CNBC on July 12 -- just one day before Monday's selloff -- that demand for AI chips and data-center capacity remains "almost unlimited," even as they described enterprises shifting toward "valuemaxxing": squeezing more measurable value out of infrastructure they've already committed to, rather than simply adding more raw capacity. That's a maturing market, not a slowing one.
The hard numbers back up the demand-side story better than the stock charts do. PitchBook's H1 2026 data shows US venture funding hit a record $412.7 billion, with AI-focused companies capturing 86% of every dollar deployed -- a concentration level that hasn't meaningfully budged through two separate Iran-linked market shocks this year. Crunchbase's independent H1 tally put global startup funding at a record $510 billion, well above the $440 billion raised in all of 2025, driven overwhelmingly by the same AI infrastructure buildout that equity markets are currently repricing on geopolitical fear.
“That February episode is instructive precisely because chip stocks recovered within weeks, and the underlying capex story never actually paused.”
The gap this creates is the real story: public-market pricing of AI-infrastructure names is proving far more volatile, and far more reactive to headline risk, than the private-market capital actually flowing into the sector. Venture and growth investors writing checks into compute, power and memory infrastructure aren't pulling back on oil spiking toward $79 a barrel -- if anything, elevated energy costs reinforce the thesis that efficient, well-sited AI infrastructure is a scarcer and more valuable asset, not a less valuable one.
For founders and GPs, the read-through is to treat equity-market chip-stock selloffs as a sentiment indicator, not a demand indicator, when evaluating whether now is a good time to raise or deploy into AI infrastructure. The businesses actually selling compute, memory and power into the AI buildout are reporting undiminished order books even as their public comparables get repriced downward on a geopolitical headline.
The bear case: if oil prices stay elevated for a full quarter rather than reverting within weeks, rising energy costs could eventually flow through to genuinely higher data-center operating costs and compress margins across the AI infrastructure stack -- at which point the equity market's fear would be vindicated rather than premature. What to watch next: whether chip stocks recover within the same multi-week window February's shock took, and whether Q3 venture funding data shows any AI concentration slippage from H1's 86% level.