Chamath Flags AI Compute Crunch: 8 Stocks To Watch

Investor Chamath Palihapitiya is calling out exploding compute demand and rising token prices. Yet the market may not be rewarding the companies positioned to solve the shortage.

• NVDA stock is moving ahead of earnings. See the price action here. 

Responding to a post from Liz Thomas noting that average LLM token costs have climbed to $2.12 per million tokens, up 12% in one week and 65% since the end of February, Palihapitiya wrote: "We are massively power-constrained. This market, if functioning properly like every market before it, should be moving in the exact opposite direction."

Palihapitiya is making the point that AI usage is scaling faster than available compute capacity. 

Agentic AI, where models do not just answer prompts but independently execute tasks, call tools, run workflows and iterate through multi-step reasoning, can consume substantially more inference than traditional chatbot-style usage.

If the cost of running LLMs is rising, it suggests demand for inference capacity is outstripping supply.

That kind of price pressure should draw capital toward the bottleneck. 

Which Stocks Could Benefit?

If compute is the bottleneck, the likely winners are the companies that own, build or optimize AI capacity. 

That starts with Nvidia Corp. (NASDAQ:NVDA), whose GPUs remain central to AI training and inference, and extends to Advanced Micro Devices Inc. (NASDAQ:AMD) as customers seek alternative accelerator supply. 

Broadcom Inc. (NASDAQ:AVGO) and Marvell could also benefit from demand for custom AI chips and networking silicon.

The second layer is infrastructure: CoreWeave Inc. (NASDAQ:CRWV), along with hyperscalers like Alphabet (NASDAQ:GOOG) (NASDAQ:GOOGL) and Oracle Corp. (NYSE:ORCL), can monetize scarce GPU cloud capacity. 

Further down the supply chain, Taiwan Semiconductor Manufacturing Co. (NASDAQ:TSM) benefits from advanced-node and CoWoS packaging demand, while Micron Technology Inc. (NASDAQ:MU) is levered to high-bandwidth memory shortages.

In short, rising token costs point investors toward the AI infrastructure stack: GPUs, custom chips, cloud compute, advanced packaging, HBM and networking.

Photo: Shutterstock