AI Selloff Reprices Buildout Pace, Not Demand: BlackRock

Nvidia H100 rental prices have risen roughly 35% year to date, and the three largest cloud providers grew revenue 56% year over year in the third quarter. Those two figures sit at the center of BlackRock's read on the post-midterm AI selloff: markets are repricing how fast infrastructure can be built, not whether the demand exists.

The note, written by Jay Jacobs and Kristy Akullian, CFA, argues that the election outcome changed expectations about permitting and development timelines rather than the commercial case for AI compute. Demand for computing power continues to outstrip supply, and the investment opportunity remains compelling in the firm's view.

What did the midterms actually change?

After the midterms, markets moved quickly to price in a sharp slowdown in AI development. BlackRock sees that reaction as premature, because much of the underlying policy picture remains unclear.

The specific worry is procedural. A shift in political power raises the odds that permitting and development decisions in certain markets slow down, which would make AI infrastructure take longer and cost more to build. Much of the feared restriction rests on campaign proposals rather than enacted policy.

Greater scrutiny may delay projects, particularly those scheduled beyond 2027, which would pressure near-term earnings and valuations for the businesses most exposed to rapid capacity expansion. The takeaway in plain terms: the risk is timing and cost, not a collapse in AI demand.

How much capacity is at risk?

The firm's numbers separate committed builds from speculative ones. Much of the capacity expected in 2027 is already committed, which limits how much near-term supply can slip.

Looking further out, the 2028 global capacity estimate excluding China has fallen from 66 gigawatts to 62 gigawatts. BlackRock judges that the missing 4 gigawatts is more likely delayed than permanently lost.

Geography matters here. Texas accounts for approximately 30% of planned U.S. projects in 2028, so permitting friction in a single state carries outsized weight for the national buildout. In short, the 2028 revision is a scheduling problem concentrated in a few jurisdictions, not a demand problem.

What do the demand signals show?

The demand side of the ledger is unambiguous, according to the note. Compute demand continues to outstrip supply across the board.

Three figures anchor that claim:

• H100 rental prices rose approximately 35% year to date.
• The three largest cloud providers grew revenues 56% year over year in Q3.
• The 2027 capacity pipeline is largely committed, indicating buyers have already locked in near-term supply.

Rising rental prices are the clearest signal. When the price of renting a flagship accelerator climbs by a third in under a year, supply is not keeping pace with demand regardless of what equity markets are pricing. BlackRock treats this spread between price signals and stock prices as the core mispricing in the selloff.

Was the selloff purely fundamental?

Not entirely. BlackRock notes that the post-election decline also coincided with a substantial hedge fund positioning unwind. In other words, part of the drop reflects leveraged investors exiting crowded trades, not a fresh assessment of AI economics.

That mechanical selling can amplify a narrative. The firm implies the market conflated forced deleveraging with a deterioration in fundamentals, when the demand data pointed the other way. The takeaway: positioning, not only policy anxiety, drove the size of the move.

What happens next?

BlackRock maintains that the long-term opportunity in AI-related investments remains significant as demand for infrastructure and capacity expands. The buildout could stretch longer than markets previously assumed, with election outcomes influencing permitting timelines in certain regions, but the supply deficit has not closed.

The near-term watch items are concrete: whether restrictions move from campaign language to enacted policy, whether projects beyond 2027 slip their schedules, and how the concentrated Texas pipeline navigates scrutiny. If committed 2027 capacity delivers on time and rental prices hold, the firm expects the demand story to reassert itself over the pricing of delay.