AI company executives pushed back on concerns about slowing demand this week, even as chip stocks swung sharply and enterprises began scrutinizing the return on their AI investments more closely.
Chip stocks endured a turbulent week amid questions about whether the artificial intelligence infrastructure buildout had become overextended, but several AI company executives pushed back on the notion that demand is slowing — even as they acknowledged enterprises are applying greater cost discipline to their AI spending.
"I somewhat think of AI demand as almost unlimited," Pat Gelsinger, the former Intel CEO and now general partner at Playground Global, told CNBC on Wednesday, adding that energy availability is "the only real limiter." "Because how much economic value do you get for increased intelligence? Almost infinite across every industry imaginable," Gelsinger said.
The remarks came against a backdrop of renewed volatility in semiconductor stocks. The VanEck Semiconductor ETF fell nearly 4% on Tuesday after Samsung's quarterly results disappointed investors and reports emerged that China's DeepSeek is developing its own AI chip. Micron dropped 4.7% on the same day.
The sell-off was compounded by Meta Platforms' announcement that it would sell excess AI computing capacity to outside customers — a move that, while boosting Meta's own stock, prompted investors to question whether broader overcapacity had taken hold across the industry.
Elon Musk's xAI also rented out its excess capacity earlier this year, adding to those concerns.
Executives at companies deeper in the AI supply chain said the reality on the ground looks different. Marc Boroditsky, chief revenue officer at Nebius — a data center operator building infrastructure using Nvidia's GPUs — told CNBC on Thursday that demand far exceeds what his company can supply. "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," Boroditsky said.
Andrew Feldman, CEO of Cerebras Systems, characterized the Meta and xAI situations as outliers. "For the industry as a whole, the demand for compute far outstrips available capacity, and we're short on data centers," Feldman told CNBC on Wednesday. Cerebras went public earlier this year and is among a group of semiconductor startups attempting to challenge Nvidia in the data center market.
Lumentum, which supplies photonics and optical products used in data center connectivity, said its products are sold out for the next five years. "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," CEO Michael Hurlston told CNBC on Wednesday. Lumentum's stock has risen roughly 600% over the past 12 months.
South Korean chip startup Rebellions, backed by Samsung and SK Hynix, reported similar conditions. "AI infrastructure momentum [is] still huge," Sungyun Park, Rebellions' CEO, told CNBC on Wednesday.
The debate over demand has a second front: enterprise spending behavior. Executives described a shift away from so-called "tokenmaxxing" — a period in which companies encouraged employees to use AI tools without regard for cost — toward what Nebius' Boroditsky called "valuemaxxing," or ensuring AI spending generates a measurable return.
"The CFO bringing the hammer down and slowing spend should actually be looking for value or valuemaxxing," Boroditsky said. "We're seeing a shift now to more rationalization. We've seen it with every tech cycle, and that rationalization will definitely continue the demand."
Frontier models from companies like OpenAI and Anthropic have faced particular scrutiny relative to lower-cost open-source alternatives from DeepSeek and Alibaba. Cerebras' Feldman suggested the market will naturally stratify, with different model types handling different workloads. "I think it's probably the case that you don't need a giant bus to go to the grocery store," he said.
Despite the week's swings, broader market indexes held gains. The Nasdaq rose 1.74% for the week and the S&P 500 gained 1.23%, with both finishing higher in four of the past five weeks. SK Hynix's U.S. market debut on Friday — opening at $170, roughly 14% above its $149 offering price — provided a further signal of sustained investor appetite for AI-adjacent memory plays.
Whether enterprise cost discipline translates into a meaningful deceleration in infrastructure spending, or merely redirects it toward higher-efficiency deployments, is the central question the industry will face heading into the second half of 2026.
TechEchelon Staff bylines are produced collectively by the newsroom for short, breaking, and wire-style coverage. Longer-form reporting is published under the responsible reporter's name.
More from the Staff →Meta has pulled an AI feature from Instagram that let users generate images referencing public accounts' photos without notifying those account holders, citing feedback that it "missed the mark."
Apple filed a federal lawsuit Friday against OpenAI, alleging the AI lab systematically stole trade secrets through former employees and job candidates in order to build its own consumer hardware, naming OpenAI's hardware chief and IO Products as defendants.
Stifel upgraded Shopify to buy on Friday and raised its price target to $150 from $110, arguing that agentic AI commerce will accelerate the platform's share gains even as the technology's financial impact remains in early stages.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.