Broadcom's CEO Says AI Investors Are Gambling Like They're at a Casino, and Capital Could Be the Industry's Real Limit

Broadcom CEO Hock Tan used an appearance at the DBS Asia Future Forward Conference in Singapore on October 14 to compare much of today's AI investing behavior to gambling, arguing that many investors "have no clue what they are investing in" and are setting what he called "very extreme expectations" for how the technology will pay off. The comparison stood out partly because it came from someone whose own company has been one of the AI boom's biggest beneficiaries, not a skeptic standing outside it.

Broadcom, which designs custom AI chips for major cloud computing operators, has booked roughly $68 billion in AI semiconductor revenue since late 2023. In its fiscal third quarter, which ended August 2, AI chip revenue alone reached $16.7 billion, more than tripling from a year earlier. That track record gives Tan's warning extra weight: he's describing froth in a market his own business has ridden higher than almost anyone.

His deeper concern is about whether the economics underneath the AI boom actually work. Tan pointed to OpenAI's reported $50 billion in annualized revenue, a huge number on its own, but noted the company is expected to spend around $25 billion this year on training costs alone, a gap that raises real questions about when, or whether, AI labs turn a profit at that scale. He went further, suggesting that the industry's biggest growth constraint going forward may not be demand for AI but the availability of capital needed to keep funding it.

Tan also took aim at how companies outside the AI industry itself are spending on the technology, saying many are "wasting money" experimenting with AI tools without a clear return to show for it. He cited a McKinsey survey in which 80% of respondents said AI had boosted individual productivity, but only 37% said it had actually contributed to their organization's earnings, a gap he frames as evidence that enterprise AI spending is running well ahead of measurable payoff.