Broadcom CEO Hock Tan said Monday that capital, not wafer or memory shortages, may become the biggest limit on AI growth, even as demand could outrun supply into 2030.
Key Points:
- Broadcom CEO Hock Tan said capital, not wafer or memory supply, may be the biggest limit on AI growth.
- Tan said demand for compute could exceed supply through 2029 and possibly into 2030.
- Intel CEO Lip-Bu Tan also cited the cost of capital and pressure to show returns on AI spending.
Hock Tan Flags Capital
Speaking on a panel in Singapore on Monday, Tan said the chipmaker is overcoming wafer and memory shortages, but capital will be its biggest constraint as AI development continues.
He added that frontier labs Anthropic and OpenAI have shown they can generate significant revenue from the latest models deployed as AI personal assistants.
The industry has hit one supply bottleneck after another over the past two years and worked through each, Tan told the DBS Asia Future Forward Conference. "Memory shortage today might not be that way a year from now," he said. He suggested wafer shortages could ease by 2028 or 2029, while power supply could improve as new infrastructure comes online.
"There is more demand for compute infrastructure than we can supply," Tan said. He added that the gap could last through 2029 and possibly stretch into 2030.
Also Read: ChatGPT’s Intelligent UI Puts Charts And Calculators In Answers For 1.2B Users
Lip-Bu Tan On Returns
Intel CEO Lip-Bu Tan echoed the concern at the same event, pointing to financing costs and the need for companies to prove their AI spending pays off. "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," he said. That pressure falls on data center, chip and power projects whose returns can take years to arrive.
Hock Tan cautioned that demand could soften if adoption of advanced models disappoints or gains in model capabilities level off. Lip-Bu Tan said companies will increasingly need to show tangible returns through higher efficiency and new business models.
Broadcom AI Financing
A month ago, Tan pointed to a different bottleneck. At a Goldman Sachs technology conference on Sept. 8, he said Broadcom could plan around supplies of wafers, memory and substrates, but power-ready data center sites were harder to predict.
He also described a financing setup with Apollo and Blackstone in which banks and financial partners take on part of the risk of funding compute for AI labs that lack enough cash flow.
The three companies unveiled that platform in June with an initial $35 billion tranche led by Apollo, aimed at more than 20 gigawatts of compute for frontier labs through 2028. Its first deal covers Anthropic's previously announced expansion of more than 1 gigawatt of computing capacity.
Read Next: Can TRON Shield $95B In Stablecoins From Quantum Attacks? Sun Says It’s Ready
