Broadcom CEO likens AI investors to casino gamblers, says capital is biggest constraint

Broadcom chief executive Hock Tan said at the DBS Asia Future Forward conference in Singapore on Oct 12 that capital, not chip supply or electricity, has become the biggest constraint on AI expansion. He warned that many AI investors lack understanding of the technologies they are betting on yet expect high returns, comparing their behavior to casino gamblers.

Tan pointed to OpenAI as an example, noting that the company reached $50 billion in annualized revenue in September but remains loss-making because of high model training costs. Revenue growth, he said, does not mean the business model has been proven, and scale alone is not enough to show frontier AI models can achieve long-term profitability.

He said capital has replaced chips, storage and power as the main limiting factor for AI. While wafer shortages may ease between 2028 and 2029 and electricity supply could become less tight as new infrastructure comes online, higher financing costs and investor return expectations will raise the bar for AI infrastructure investment.

Tan expects supply-demand imbalance in computing capacity to persist until 2029 or even 2030, but said demand could slow if adoption of advanced AI models falls short of expectations or model progress hits a bottleneck. Intel CEO Lip-Bu Tan shared similar views, saying capital costs are a core challenge. The withdrawal of AI data center company Firmus Grid's IPO also underscored investor caution over AI infrastructure valuations and future returns.