The demand for AI is 'boundless.' So why are the shares of semiconductor companies declining?

      TL;DRAI executives assert that demand is "almost unlimited," with Pat Gelsinger identifying energy as the primary constraint, while Lumentum states its products are sold out for the next five years. However, chip and data center stocks are still experiencing fluctuations, as the approximately 60% rally in the PHLX chip index has priced in the expectation of flawless execution. For instance, although Samsung projected a significant profit increase, its stock still declined, and Meta’s plan to offload excess computing resources has created mixed interpretations. The discrepancies stem from expectations rather than actual demand.

      Executives involved in the AI surge are steadfast in their belief that demand is nearly limitless, despite the instability in the stocks tied to it, according to CNBC. Pat Gelsinger, previously with Intel and now at Playground Global, expressed this succinctly, viewing AI demand as almost boundless, with energy availability being "the only real limiter."

      The company's order books back up this perspective. Lumentum, which provides optical components for data center connectivity, reports that its inventory is sold out for the next five years.

      So, why the stock volatility?

      The answer lies in valuation; the prices already reflect all this potential. The PHLX chip index has risen around 60% this year, indicating that future flawless execution is already expected. At such valuation levels, even good news may not suffice. Samsung's substantial profit forecast didn't prevent its stock from falling, following a more than 360% rally over the previous year.

      This trend also manifested in other companies, such as Cerebras, which saw a doubling in revenue yet experienced a stock price drop. When expectations are set so high, exceeding them can sometimes be interpreted as a failure.

      Meta heightened investor uncertainty by announcing it would sell off surplus AI computing power. This move could be perceived as either a savvy monetization strategy or as an acknowledgment that the firm overestimated its computing needs.

      The optimistic and pessimistic viewpoints coexist. Proponents, or bulls, have strong financials to support their claims. Unlike the dot-com era, the companies fueling this surge are quite profitable, and the demand signals from suppliers are authentic. SoftBank's Masayoshi Son even argued that labeling AI as a bubble is an insult, framing the development as a generational infrastructure endeavor rather than mere hype.

      On the other hand, skeptics, or bears, do not dispute demand; they contest the pricing, pointing out that market concentration has surpassed levels seen in 2000 and the returns on substantial investments remain uncertain.

      Both perspectives may hold validity. Genuine demand can exist alongside stocks that are overvalued for what that demand can realistically return over any reasonable timeframe.

      The critical limitation that cannot be circumvented is Gelsinger’s caution regarding energy. If energy represents the true restraint, then chips are not the limiting factor, and the industry's valuations depend on infrastructure that lies beyond its control.

      Capital is being funneled into addressing this gap, with Nvidia-backed startups securing funding to tackle data center energy challenges. However, the development of electricity grids, turbines, and securing planning permissions operates on timelines that disregard quarterly earnings.

      This reality underlies the volatility in the sector. The industry may believe in infinite demand—and it may be correct—but the availability of electricity is finite, and stock prices do not reflect that.