SINGAPORE — The chief executives of Intel and Broadcom, whose chips sit inside a large share of the world's AI infrastructure, delivered an unusually candid warning on Monday: the industry's binding constraint may not be semiconductors, memory or electricity, but money.
Broadcom CEO Hock Tan told attendees at the DBS Asia Future Forward Conference that financing could become a bigger obstacle to AI expansion than any physical shortage, CNBC reported from the event. For more context on this story, see our ongoing AI trends.
"[What] I think might be the biggest constraint is capital," Tan said.
Intel CEO Lip-Bu Tan, speaking at the same conference, pointed to rising financing costs and to unresolved questions about returns. "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, according to CNBC. The two executives share a surname but run rival chipmakers.
That two of the most powerful figures in semiconductors chose a fintech-adjacent stage in Asia to talk about financing rather than silicon says something about where the AI race stands. The physical supply chain has spent two years proving it can be unblocked. Whether the capital behind it stays equally pliable is the newer and less tested question — and it is now being raised by the very executives whose order books are fullest.
A Supply Chain That Keeps Finding — and Clearing — Bottlenecks
Hock Tan's argument rested on the industry's recent track record. Over the past two years, the AI buildout has hit a series of shortages — advanced packaging, high-bandwidth memory, wafer capacity, grid power — and overcome each one as production caught up.
"[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."
He expected wafer shortages to ease by 2028 or 2029, with power availability improving as new generation and transmission infrastructure comes online. Capital, in his framing, does not clear the same way: it depends on borrowing costs and investor patience rather than factory output.
Demand Still Outstrips Supply
Despite the financing worries, both executives remained emphatically bullish on near-term demand for computing infrastructure.
"There is more demand for compute infrastructure than we can supply," Hock Tan said, adding that the imbalance could persist through 2029 and potentially into 2030.
Lip-Bu Tan likewise said he did not expect a significant slowdown over the next two to three years, citing growing adoption of AI applications and the corresponding appetite for computing power.
Both men framed the longer term as conditional. Broadcom's Tan cautioned that demand could slow if adoption of advanced AI models disappoints or if improvements in model capabilities begin to plateau. Intel's Tan emphasized that companies will increasingly need to demonstrate tangible returns from AI investments through improved efficiency and new business models.
Why Capital Is Different From Chips
Shortages of chips, memory and power have a self-correcting quality: high prices attract capacity, and capacity eventually arrives. Financing works differently. As CNBC noted in its report, rising borrowing costs threaten to complicate funding for technology companies investing heavily in AI infrastructure — higher bond yields make data center construction and chip purchases more expensive, putting pressure on the economics of projects that may take years to generate returns.
The warnings carry particular weight because of who is delivering them. Broadcom has been among the largest beneficiaries of the AI spending cycle through its custom accelerator and networking businesses, giving Hock Tan a direct view into hyperscaler order books. Intel, under Lip-Bu Tan, has been working to position the company as a serious compute and foundry supplier for the AI era.
The ROI Question Hangs Over the Buildout
Neither executive said demand was in question. The concern was the second-order effect: the next phase of the buildout depends on capital markets staying willing to fund assets whose payback periods stretch out for years, and that willingness depends on evidence that today's investments are earning tomorrow's returns.
That evidence has so far been mixed across the industry, with cloud providers reporting strong AI-driven revenue growth even as questions persist about how broadly the productivity gains will spread beyond the largest technology companies. Lip-Bu Tan's formulation — that people "want to see the ROI" in the longer run — captured the tension: the buildout is racing ahead of the proof.
What to Watch
The Singapore remarks arrive in the middle of earnings season, when chipmakers and cloud providers will report fresh data on AI demand and capital spending. If borrowing costs stay elevated, expect the financing question to move from conference panels to balance sheets: which projects get funded, which get delayed, and whether the industry's historic appetite for compute can keep outrunning its cost of capital.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
