$NVIDIA(NVDA.US)$Taiwan Semiconductor(TSM.US)

OpenAI's growth fell short of expectations, which may not indicate insufficient AI demand, but could also mean that previous product supply was constrained by computing power.

If there's not enough computing power, OpenAI would face several issues: limited experience for free users, downgrades for Plus or Pro users, slow speeds for complex tasks, high costs for context and deep reasoning, and affected availability and stability of enterprise APIs. This would suppress both user growth and paid conversions. The recently launched version 5.5 is noticeably better, and Codex has also improved. This might indicate that model capabilities, reasoning computing power, and system scheduling capabilities have all improved simultaneously.

So this can be interpreted in two ways:

Pessimistic interpretation: OpenAI missing its target shows that AI user growth and revenue realization are not as fast as the market imagined, and the large computing power contracts might overwhelm cash flow in the future.

Optimistic interpretation: OpenAI previously didn't lack demand, but rather had insufficient supply. Inadequate computing power limited product experience, which in turn suppressed growth below its theoretical potential. Now, with the significantly improved experience of version 5.5, it actually shows that massive capex is translating into better products. In other words, AI capex isn't wasteful; it's filling infrastructure gaps.

Tomorrow's earnings reports from AMZN, GOOG, and MSFT will provide further insight.