Intel, Broadcom CEOs: capital may be AI's biggest constraint

The two CEOs spoke at a financial conference in Singapore, and both pointed to the same obstacle: capital. Hock Tan called access to capital "might be the biggest constraint" facing the sector, while Lip-Bu Tan said, "if there's any challenge in the front, it's capital and the cost of capital, and then the other part is in the longer run, people want to see the ROI." Their remarks came as demand for some data center chips is exceeding supply.

Hock Tan also addressed the cost of training large language models, focusing on OpenAI, which Broadcom helped design its first custom inference chip for earlier this year. He noted that a sizable portion of AI labs' enterprise revenue comes from pilot projects, comparing them to a child at a video arcade who spends $10 of tokens in 10 minutes and learns nothing. He added that AI labs should acquire more customers who can realize a return on their inference spending. On AI stocks, he said many investors "set very extreme expectations only to find out that it's almost like they are in a casino."

For context on margins, the Financial Times reported last month that Anthropic's gross margins exceed 80% when excluding model training costs and revenue shared with go-to-market partners, and that the company is reportedly on track for a positive adjusted operating income for the second quarter in a row. Lip-Bu Tan also noted that open-source language models from Chinese labs are proving "good enough" for many budget-conscious users and have gained significant traction over the past year. The discussion later shifted to quantum computing, which he predicted could become mainstream within three to five years.