Recent sharp fluctuations in chip stocks have raised market concerns about a potential slowdown in AI demand, but several industry executives have clearly stated these worries are exaggerated.
Former Intel CEO and current Playground General Partner Pat Gelsinger told CNBC this week that AI demand is "almost infinite," with the real bottleneck for industry development being energy supply, not a lack of demand for computing power. Executives from multiple data center and chip companies have also voiced similar opinions, stating that current market demand far outstrips supply capacity, and the notion of a compute glut does not align with their actual business conditions.
Simultaneously, corporate spending patterns on AI are undergoing a quiet shift—moving from previously encouraging employees to use AI tools extensively regardless of cost, towards a more return-on-investment focused "value maximization" strategy. This trend has sparked external doubts about the sustainability of corporate AI spending, but several executives believe this rationalization adjustment will actually help sustain long-term demand.
Roots of the Chip Stock Volatility: Where the Glut Concerns Originate
Several factors are driving the recent chip stock volatility. Meta Platforms, Inc.'s announcement that it would sell its idle AI compute capacity, while boosting its own stock price, sparked market associations with an industry-wide compute surplus. Elon Musk's xAI is also leasing out excess computing power this year.
Furthermore, one of the world's largest memory chip makers, Samsung, this week forecast a significant profit increase, yet its stock price fell—after a cumulative gain of over 360% in the past 12 months, the market began questioning whether its upside potential had become limited.
These market signals combined have led investors to question the sustainability of AI infrastructure investment, triggering a new round of volatility in chip and data center-related stocks.
The Supply Side Perspective: Demand Far Exceeds Capacity, Orders Booked for Five Years
However, executives from several companies directly involved in AI infrastructure construction offer a starkly different assessment.
"The demand we are experiencing is extraordinary. Demand far exceeds our ability to fulfill it, and that has been the case for some time now," Marc Boroditsky, Chief Revenue Officer at Nebius, which is building data centers using NVIDIA GPUs, told CNBC.
Cerebras Systems CEO Andrew Feldman characterized the cases of Meta and xAI selling spare capacity as "exceptions," emphasizing, "For the industry as a whole, the demand for compute far exceeds the available supply. We have a gap in data centers, and we have a gap in many of the inputs needed for compute." Cerebras, one of several semiconductor startups aiming to challenge NVIDIA in the data center market, went public earlier this year.
Rebellions, a Korean chip startup backed by Samsung and SK Hynix, also reported robust demand. "The momentum for AI infrastructure is still strong," Rebellions CEO Sungyun Park stated, believing the actions by Meta and xAI do not indicate over-investment in infrastructure by hyperscale cloud providers.
The situation at photonics and optical networking product supplier Lumentum may be most telling—its CEO Michael Hurlston revealed that its product order backlog is full for the next five years. "We are doing everything we can to expand capacity to meet the five-year demand that we see right now," he said. Lumentum shares have risen approximately 600% over the past 12 months.
The Demand Side Evolution: From 'Usage Maximization' to 'Value Maximization'
Despite clear signals from the supply side, the way enterprises use AI is undergoing a structural shift, which is another source of market concern.
Previously, many companies were in a so-called "tokenmaxxing" (usage maximization) phase, encouraging employees to use AI tools heavily regardless of cost, primarily employing cutting-edge models from OpenAI and Anthropic. However, as the cost advantages of these frontier models compared to open-source alternatives like DeepSeek and Alibaba come under increasing scrutiny, corporate CFOs are beginning to evaluate the return on AI investment more rigorously.
Nebius's Boroditsky termed this new trend "valuemaxxing" (value maximization). "The CFO stepping in to tighten spending is really about looking for value," he said. "We are seeing a more rational shift. Every technology cycle goes through this, and this rationalization actually sustains demand."
Cerebras's Feldman offered another perspective from the angle of model tiering: as companies mature in their AI deployment, tasks of varying complexity will be matched with different levels of models and compute power. "You don't need a bus to go to the grocery store," he said. "Certain workloads will move to a certain type of compute, and simpler workloads will move to other compute." This suggests frontier and open-source models will form a complementary, not simply substitutive, relationship, and overall compute demand will not shrink as a result.
Comments