Chip stocks have slumped in recent weeks on fears that AI demand may be cooling, but a chorus of industry executives says the opposite is true. Supply constraints, not waning appetite, are shaping the market, with one CEO describing demand as "almost unlimited" and another saying his products are sold out through 2030.
Meta's sale of excess AI compute and Samsung's falling stock despite a projected profit surge have stoked volatility. But in interviews with CNBC, leaders from across the semiconductor and data center ecosystem pushed back against the idea that demand is softening.
Executives see no ceiling on AI compute demand
Pat Gelsinger, the former Intel CEO who now serves as a general partner at Playground Global, said energy availability is the only factor capping growth. "I somewhat think of AI demand as almost unlimited," he said. "Because how much economic value do you get for increased intelligence? Almost infinite across every industry imaginable."
Marc Boroditsky, chief revenue officer at Nebius, a company building Nvidia GPU-powered data centers, said, "What we're experiencing in terms of demand is extraordinary. There's much more demand than we're able to fulfil, and that's been our experience for some time now."
Andrew Feldman, CEO of chip startup Cerebras Systems, called the Meta and xAI capacity sales a "unique" situation. "For the industry as a whole, the demand for compute far outstrips available capacity, and we're short on data centers. I think we're short on, as an industry, many of the inputs to compute," he said.
Sungyun Park, CEO of South Korean chip startup Rebellions, which is backed by Samsung and SK Hynix, said the AI infrastructure momentum remains "huge" and rejected the notion that hyperscalers are overinvesting. Lumentum, which sells optical connectivity products for data centers, said its current product line is sold out for the next five years. CEO Michael Hurlston told CNBC, "We're trying to build up our capacity as much as we possibly can to fulfil a demand that we see out five years at this point."
For professionals working in AI for IT & Development, these supply signals mean that the infrastructure buildout will continue to drive demand for skills in data center architecture, GPU provisioning, and network optimization for years to come.
Enterprise spending shifts from "tokenmaxxing" to ROI
As businesses face the cost of frontier AI models, spending habits are changing. Boroditsky said a phase of "tokenmaxxing" - where enterprises encouraged heavy AI usage regardless of outcome - is giving way to more disciplined evaluation. "The CFO bringing the hammer down and slowing spend should actually be looking for value or valuemaxxing," he said, adding that AI should create value that justifies the cost.
Feldman noted that workloads will bifurcate: advanced problems will use frontier models, while simpler tasks move to less expensive alternatives. "I think it's probably the case that you don't need a giant bus to go to the grocery store," he said.
This rationalization mirrors previous tech cycles and is a key theme in AI for Executives & Strategy, where leaders must decide when to invest in premium AI tools and when open-source models provide sufficient performance.
Why this matters for IT and development professionals
The supply constraints in AI hardware and infrastructure mean data center capacity and specialized chips will be tight for the foreseeable future. IT and development teams should plan for longer lead times on GPU clusters and build flexibility into their architectures to accommodate both high-end and open-weight models.
On the software side, the move toward ROI-driven AI spending will increase pressure to measure and demonstrate tangible outcomes from AI projects. Developers and engineers who can optimize model selection, fine-tune open-source models for specific tasks, and tie AI performance to business metrics will be in high demand.
Your membership also unlocks:
