'AI demand is infinite — energy is the only constraint,' industry leaders say, dismissing peak-chip fears

Former Intel CEO Pat Gelsinger and other AI industry executives said demand is in effect unlimited and that energy supply is the only constraint on realizing a fully AI-driven economy.

CNBC reported Sunday that the common message emerged from interviews with Gelsinger and other industry leaders.

Gelsinger resigned from Intel last year and now serves as CEO of technology-focused venture capital firm Playground Global. He told CNBC he believes "AI demand is nearly unlimited" and that energy supply is "the only real constraint." He added that "the economic value available through AI enhancement is nearly infinite across every industry you can imagine."

Mark Boroditsky, chief revenue officer of neocloud company Nebius, said "what we're experiencing on the demand side is extraordinary," adding that there is "far more demand than we can handle — and that's been the case for some time now." The remarks amount to a direct rebuttal of the so-called peak-chip thesis, which holds that computing resources may be heading toward oversupply.

Andrew Feldman, CEO of US AI chip maker Cerebras Systems, called Meta Platforms' recent suggestion that it might lease out surplus AI computing capacity an "unusual case." He said that "across the industry, demand for compute far exceeds available capacity, and data centers are in short supply," adding that "many of the components that go into compute are scarce across the industry."

Park Sung-yoon, CEO of Korean semiconductor startup Rebellions — which counts Samsung and SK Hynix among its investors — said "the momentum for building AI infrastructure is still enormous" and that supply cannot keep pace with demand. On Meta and xAI's recent moves to lease computing power, he said, "Personally, I don't think this is a signal that all hyperscalers are overinvesting in infrastructure."

Lumentum, a manufacturer of optical and photonics components for data centers, said it has already contracted out all the products it expects to produce over the next five years. CEO Michael Hurlston said the company is "trying to ramp capacity as fast as possible to meet demand five years from now." Driven by surging demand, Lumentum's share price has risen roughly sevenfold over the past year.

Even those who emphasized the enormous demand for computing power acknowledged that a note of caution has recently crept into AI-related investment decisions.

Nebius CRO Boroditsky said "CFOs are now starting to cut or scrutinize AI budgets more carefully," but stressed that "this doesn't mean they're abandoning AI investment altogether." He described the shift as a process of "value-maxing" — optimizing return on investment — and said the earlier era of "token-maxing," or unlimited token consumption, "only makes sense when companies are getting a return on that investment."

"We're seeing a transition toward more rational spending — that's something we've always seen in past cycles," he said. "This rationalization will ultimately make AI demand sustainable."

Cerebras CEO Feldman said that in the future, cutting-edge AI models will handle more complex tasks while other models take care of the rest — "just like you don't need a big car to go to the corner store."