2 min read
Add as a preferred source on Google

Blackstone’s Jon Gray Says AI Demand Is Outrunning Supply

Gray cited a projection that five hyperscalers will spend $820 billion in 2026; Griffin warned that excess GPU capacity could send computing prices sharply lower.

An industrial turbine stands beside an unfinished data center campus / TokenPost.ai (mono)
An industrial turbine stands beside an unfinished data center campus / TokenPost.ai (mono)

Blackstone President Jon Gray says AI demand is outrunning infrastructure supply, while Citadel CEO Ken Griffin warns that excess computing capacity could push prices lower and weaken returns.

Gray cited combined capital spending by five hyperscalers at $415 billion in 2025, rising to a projected $820 billion in 2026. He put the projected total at about 2.5% of U.S. gross domestic product. The figure captures planned investment, rather than a direct measure of how much computing capacity is already serving customers. Gray’s argument is that rising budgets alone cannot meet demand if the equipment and sites needed to expand supply remain scarce.

He has pointed to shortages of electricity, turbines and memory chips as limits on how quickly new capacity can be built. Gray has also acknowledged that the industry could overbuild. Those comments describe two risks: construction may lag demand, or new projects may leave providers with more capacity than customers use.

Blackstone has a direct position in the expansion. The firm has outlined data-center leasing plans for 2026 and is among the financial firms working with NVIDIA on AI infrastructure financing platforms. NVIDIA and Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR signed memorandums of understanding Aug. 10 to pursue platforms intended to mobilize more than $500 billion in third-party capital over time. The stated target is spread over time, with no single-year amount specified; it is not a cash commitment by the firms.

Griffin addressed the other side of the debate at a New York conference Tuesday, Oct. 20. He said there are “almost certainly bubbles within this moment in time” and compared graphics processing units with airline seats. An empty seat may still earn revenue if sold at a discount, he said, but too much supply can force prices down. His concern is that computing capacity could expand faster than the number of customers willing to pay for it.

Gray is pointing to investment and supply bottlenecks; Griffin is highlighting the risk that added capacity will erode pricing. The central question is whether demand from paying customers will grow fast enough to absorb the infrastructure being built.

John Kim

John Kim reports on the digital-asset business for TokenPost. Send corrections or tips to info@tokenpost.com.

Loading…