The CEO's of Broadcom and Intel Say the Quiet Part Out Loud about AI and ROI

Broadcom CEO Hock Tan and Intel CEO Lip‑Bu Tan publicly acknowledged major uncertainties around generative AI: the lab-driven business model, how companies will actually monetize AI, and whether the massive infrastructure investments will deliver acceptable ROI. Hock Tan warned that many investors are behaving like gamblers, pointing out that OpenAI’s roughly $50 billion annualized revenue run rate does not prove a sustainable, profitable business once huge operating and capital costs are counted. Both executives stressed that AI builders must find customers who generate real returns rather than users “playing with tokens” in experimental deployments - a comparison Hock Tan likened to kids burning through arcade tokens.

They outlined specific bottlenecks and a timeline for relief - memory, wafer capacity and power constraints could ease as production and infrastructure expand, perhaps by 2028-2030 - but said financing is the biggest near‑term constraint because higher Treasury yields and wider credit spreads have substantially increased the cost of funding the buildout (one example saw a recent 10‑year corporate note trade up to ~7.2% from a ~5.9% coupon at issuance). Demand for compute may exceed supply through 2029-2030 if adoption remains strong, but both stressed that future growth hinges on demonstrable ROI via efficiency gains or new business models, and ended with a candid, humble recognition of how little is yet certain.