AI Stocks After the 2026 Midterms: A Delay, Not a Demand Collapse

By Sophie Lindqvist

H100 rental prices rose approximately 35% year to date even as AI stocks sold off after the 2026 midterms — the clearest sign that the market rout reflects fears about slower infrastructure delivery, not fading demand for computing power. That is the core argument from Jay Jacobs and Kristy Akullian, CFA, of BlackRock, who argue the post-election volatility is a reassessment of how quickly AI infrastructure can be built rather than a change in the underlying appetite for it.

The numbers they marshal make the case bluntly. Compute demand continues to outstrip supply. The three largest cloud providers grew revenues 56% year over year in the third quarter. Meanwhile, the 2028 global capacity estimate excluding China has fallen from 66 to 62 gigawatts — a real downgrade, but one that BlackRock frames as delay rather than permanent loss.

Why did AI stocks fall after the midterms?

Markets moved fast after the election. According to BlackRock's analysts, markets were quick to price in a sharp slowdown in development, even as policy implications remain far from clear. The shift in power in Washington has created concerns that the infrastructure needed to meet AI demand may take longer and cost more to build.

The key mechanism here is permitting. "Our view is that the AI buildout could take longer as election outcomes influence permitting and development decisions in certain markets," Jacobs and Akullian write, "but demand for computing power continues to outstrip supply and the investment opportunity remains compelling."

Takeaway: the selloff prices in political friction, not a demand shock.

How much capacity is actually at risk?

The picture is more granular than the headline selloff suggests. BlackRock's quick insights break it down:

• Much of the capacity expected in 2027 is already committed, which limits near-term downside.
• The 2028 global estimate excluding China has fallen from 66 to 62 gigawatts.
• The remaining capacity is more likely delayed than permanently lost, in BlackRock's assessment.
• Texas accounts for approximately 30% of planned U.S. projects in 2028.

That Texas figure matters. If permitting decisions cluster in a few politically consequential states, then election outcomes can have outsized effects on the national buildout timeline — even when national demand is unchanged.

Takeaway: committed 2027 capacity provides a floor, while 2028 faces genuine slippage risk.

What does the pressure mean for earnings and valuations?

BlackRock is direct about who gets squeezed. Greater scrutiny may delay projects, particularly beyond 2027, pressuring near-term earnings and valuations for businesses most exposed to rapid capacity expansion.

There is also an important caveat about what is actually policy versus what is speculation. Restrictions remain uncertain, with much of the concern based on campaign proposals rather than enacted policy. In other words, markets are trading on promises made during campaigns, not on laws that have passed — a distinction that cuts in both directions once actual legislation takes shape.

Takeaway: companies betting hardest on fast capacity growth carry the most repricing risk.

How much of the selloff was positioning, not fundamentals?

Not all of the decline was about AI fundamentals at all. The selloff also coincided with a substantial hedge fund positioning unwind, according to BlackRock. When heavily positioned funds pull back at the same time, price moves amplify beyond what the underlying news would justify. That mechanical detail helps explain why the drawdown looked more violent than the incremental policy news might warrant.

Takeaway: part of the drop was leveraged money exiting, not investors rewriting their AI outlooks.

Why this matters for the broader market

The stakes go beyond a handful of semiconductor and data-center stocks. As BlackRock puts it, "As demand for AI infrastructure and capacity expands, we continue to see a significant long-term opportunity in related investments." The firm — a fiduciary to investors and a leading provider of financial technology managing savings for millions — is effectively telling clients that the investment case survives the election, just on a longer clock.

The framing matters for anyone trying to read the next few quarters. If BlackRock is right, the trade is not about whether AI capacity gets built, but about when and at what cost — and which companies can absorb a longer timeline without breaking their earnings models.

Takeaway: the debate has shifted from demand to delivery.

What comes next

The signals to watch are concrete. Committed 2027 capacity, the trajectory of the 2028 gigawatt estimates, permitting decisions in markets like Texas, and whether campaign proposals harden into enacted restrictions will determine whether the 62-gigawatt figure stabilizes or erodes further. As long as H100 rental prices keep climbing and cloud revenue keeps compounding at rates like 56%, BlackRock's position is that supply — not demand — remains the binding constraint on the AI buildout.