AI Chip Boom Faces Capital Crunch as Demand Extends to 2030

Rising financing costs could become a major obstacle to the artificial intelligence boom, even as demand for computing infrastructure exceeds available supply, executives from Broadcom and Intel warned at a conference in Singapore on Monday. Their concerns contrast with ASML's outlook for sustained semiconductor demand beyond 2028, as technology companies expand data centers and develop increasingly powerful AI systems.

Broadcom CEO Hock Tan said access to capital could prove more difficult than shortages of chips, memory or electricity. He expects some manufacturing constraints to ease by 2028 or 2029, but demand for computing capacity could continue outstripping supply through 2029 and potentially into 2030, according to CNBC.

The financing challenge comes as AI infrastructure projects require substantial spending on advanced processors, data centers and power systems. Higher borrowing costs could make these investments more expensive, while companies face growing pressure to demonstrate that AI adoption will generate sufficient financial returns.

Tan identified financing as the industry's most significant potential constraint during the DBS Asia Future Forward Conference in Singapore.

"[What] I think might be the biggest constraint is capital," said Hock Tan, Broadcom CEO.

He said the industry had encountered successive supply constraints over the previous two years, but production and infrastructure improvements could gradually address them.

"[In the] last two years we have seen a series of bottlenecks that keep coming out, but then we overcome it," he said. "Memory shortage today might not be that way a year from now."

Tan also said demand for computing infrastructure remains greater than the industry can supply, with the imbalance potentially lasting into the end of the decade.

"There is more demand for compute infrastructure than we can supply," said Hock Tan, Broadcom CEO.

Intel CEO Lip-Bu Tan raised similar concerns about financing expenses and the need to prove that investments deliver measurable economic benefits. He said demand for AI computing was likely to remain strong over the next two to three years, although longer-term growth would depend on adoption and commercial results.

"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," said Lip-Bu Tan, Intel CEO.

He emphasized that businesses would need to show how AI investments improve efficiency and support new business models. Hock Tan also cautioned that demand could weaken if advanced AI models fail to gain adoption or their capabilities stop improving significantly.

ASML CEO Christophe Fouquet expects the semiconductor industry's expansion to continue as technology companies race to build AI infrastructure. ASML manufactures the equipment required to produce advanced chips and is preparing to report earnings this week.

Fouquet described a sharp change in industry conditions, from expectations of a relatively flat market to an urgent effort to satisfy AI-related orders. In an interview conducted in June and published the following week, he said manufacturers were working to address a substantial backlog.

"What you see now is the whole industry trying to catch up on a pretty significant backlog," said Christophe Fouquet, ASML CEO.

Fouquet expects AI infrastructure construction to continue for two to five years, potentially sustaining semiconductor demand beyond 2028. According to a September assessment by JPMorgan, most orders for ASML's advanced chipmaking equipment were already scheduled for delivery in 2028.

ASML said in July that it planned to increase annual production capacity for its current-generation extreme ultraviolet lithography machines by 30% in 2027. The company was also considering another 30% expansion for 2028.

Broadcom's involvement in financing AI infrastructure illustrates the scale of the industry's capital requirements. The company was in preliminary discussions to arrange more than $50 billion in financing to help OpenAI purchase custom AI chips, according to a Wall Street Journal report published the previous week.

Separately, a proposed $60 billion financing package supporting Anthropic's use of Google chips included $42 billion in debt backed by Broadcom, the Financial Times reported. The arrangements highlight how the expansion of AI computing increasingly depends on access to large amounts of financing alongside semiconductor manufacturing capacity.