Two of Wall Street’s most powerful voices are looking at the same AI boom and seeing very different things. Blackstone President Jon Gray is rejecting the idea that artificial intelligence has become a bubble.
His pushback came just days after Citadel founder Ken Griffin said investors are “almost certainly” in the middle of one.
Gray’s case: too much demand, not enough supply
Gray’s argument rests on a simple imbalance. He has said the appetite for computing power, driven by increasingly capable AI models, is running well ahead of the infrastructure available to meet it.
In a September 4, 2026 interview with Les Echos Capital Finance, Gray said demand for compute exceeds supply by a significant margin. He framed that gap as the key difference between today’s AI buildout and classic speculative manias.
He also pointed to the scale of spending by the biggest cloud and tech companies, known as hyperscalers. In a September 2026 investor presentation, Gray said hyperscaler capital expenditures doubled to $820 billion, a figure equal to approximately 2.5% of US GDP.
Griffin’s counterpoint: watch the GPUs
Griffin sees the same spending and draws a darker conclusion. Speaking at the Future of Global Markets Conference on October 20, 2026, he argued that investors are almost certainly in an AI bubble.
His concern centers on graphics processing units, the specialized chips that power AI training and inference. Griffin flagged the risk that GPU supply could eventually overshoot demand, which could trigger sharp price declines.
Griffin did not paint a picture of total collapse, though. He also noted that established players tend to prove resilient during downturns.
Blackstone’s money is where its mouth is
Gray is not offering a detached opinion. Blackstone has made AI infrastructure a central pillar of its strategy, which gives his comments both credibility and an obvious interest.
In August 2026, Blackstone and five other Wall Street firms committed to deploying more than $500 billion to finance AI infrastructure.
The firm also plans to lease a minimum of six gigawatts of capacity in 2026 across its fast-growing data-center platform.
Blackstone has also reported a 7x increase in spending on large language models among its portfolio companies. That figure matters because it speaks to usage, not just construction.
Why this debate matters beyond Wall Street
The bubble question has become one of the defining arguments in global markets. AI infrastructure now touches chipmakers, utilities, real estate owners, private credit lenders, and the largest tech companies on the planet.
For investors, the useful signals are fairly concrete. Watch whether hyperscaler capex keeps climbing from the $820 billion level Gray cited, or whether budgets start to flatten.
Watch GPU pricing and availability. A shift from scarcity to easy supply would be an early sign that Griffin’s oversupply scenario is taking shape.
Blackstone’s reported 7x jump in LLM spending across its portfolio is the kind of metric that supports Gray’s thesis. If adoption growth slows while construction keeps accelerating, the gap between the two camps narrows quickly.
There is also a credibility dynamic worth noting. Blackstone has billions riding on AI infrastructure, which means Gray’s optimism lines up neatly with his firm’s positioning. Griffin runs a trading-focused firm that can profit from volatility in either direction.
With Blackstone and five other firms pledging more than $500 billion, a growing share of AI infrastructure is being financed outside public markets. That spreads exposure across pension funds, insurers, and other institutional backers who invest through these vehicles.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.