On October 12, 2026, at the DBS Asia Future Forward Conference in Singapore, Broadcom CEO Hock Tan and Intel CEO Lip-Bu Tan said capital access and the cost of capital could become the biggest constraint on the AI infrastructure buildout—larger than remaining chip, memory, or power shortages. Both executives said near-term demand still exceeds supply and could do so through 2029 or into 2030. Startup founders and operators raising for AI infrastructure or dependent on hyperscaler capacity should stress-test financing assumptions and customer ROI timelines this week.
Broadcom’s Hock Tan and Intel’s Lip-Bu Tan told a Singapore conference on October 12, 2026, that financing, not chips, may be the next binding constraint on AI expansion. Startup founders, AI-infrastructure operators, and teams that sell into the AI buildout should care because higher capital costs and rising investor scrutiny of returns can slow data-center projects and tighten the market for compute.
- What changed? Two major chip CEOs publicly identified capital and the cost of capital as potentially the largest constraint on AI infrastructure growth.
- When? Remarks delivered October 12, 2026, at the DBS Asia Future Forward Conference in Singapore.
- Who is affected? AI startups raising capital for infrastructure or model training, companies dependent on hyperscaler capacity, and operators selling tools or services into the AI buildout.
- What to do this week? Re-run financing scenarios at current long-term yields; pressure-test customer ROI timelines; and map which of your own capacity or chip commitments could slip if capital becomes tighter.
What did the CEOs say on October 12?
Speaking at the DBS Asia Future Forward Conference in Singapore on Monday, October 12, 2026, Broadcom CEO Hock Tan said: “What I think might be the biggest constraint is capital.” He argued that financing could become a larger obstacle than the semiconductor, memory, or electricity shortages the industry has been managing. Intel CEO Lip-Bu Tan echoed the point, stating that the front-end challenge is “capital and the cost of capital,” and that over a longer horizon investors will want to see returns on the large sums being spent.
Hock Tan added that the industry has repeatedly hit supply bottlenecks over the past two years and then overcome them. He suggested memory shortages could look different a year from now and that wafer constraints might ease by 2028 or 2029, with power availability also improving as new infrastructure comes online. Both executives remained bullish on near-term demand: Hock Tan said there is more demand for compute infrastructure than the industry can supply and that the imbalance could persist through 2029 and potentially into 2030.
How does a capital constraint differ from a chip shortage?
Chip, memory, and power shortages are physical. They can be addressed by expanding fabrication capacity, securing long-term supply agreements, or bringing new generation assets online—processes that take time but have clear engineering paths. A capital constraint is financial. Higher bond yields and a higher cost of equity raise the hurdle rate for data-center projects whose cash flows may be years away. Projects that looked viable at lower yields can stall or be scaled back even if chips are available.
Hock Tan has also been reported likening some AI investors to casino gamblers who do not fully understand the technology they are funding, particularly at the model frontier. That remark sharpens the capital warning into a caution about speculative excess as well as absolute funding availability.
Why are the remarks landing now?
Long-term Treasury yields have been elevated. Markets are pricing the path of Fed policy against sticky inflation and energy-price pressure. At the same time, hyperscalers and infrastructure companies have announced very large multi-year capital plans for AI. The combination of high absolute spending and higher financing costs makes the return question more urgent. Lip-Bu Tan explicitly paired the capital point with the longer-run need for companies to demonstrate ROI through efficiency gains or new business models.
For startups, the implication is indirect but material. If hyperscalers slow or rephase capacity additions, available compute tightens. If infrastructure-focused funds become more selective, the cost and availability of growth capital for AI companies can change quickly.
What should startup and operator teams do this week?
Re-run any infrastructure or large compute-related financing models at current long-term yields rather than last year’s lower rates. Identify which customer or internal projects depend on capacity that is still being financed rather than already contracted. Ask commercial counterparts for updated delivery timelines on reserved instances or dedicated clusters. For companies raising, prepare materials that show a clear path to revenue or cost reduction rather than solely to model capability.
Teams that sell into the AI buildout should also map concentration risk: how much of near-term revenue depends on a small number of hyperscaler or infrastructure customers whose capital budgets could be revised. Diversifying the customer base or shifting more of the offering toward efficiency and ROI tools can reduce exposure to a pure capacity-build cycle.
What should teams watch next?
Watch hyperscaler capital-expenditure guidance in the current earnings season, any further commentary from chip executives on financing vehicles, and movements in long-term yields around the September CPI release and the late-October FOMC meeting. Also monitor whether infrastructure IPOs or large debt raises for AI capacity continue to clear or are delayed. The CEOs’ remarks do not predict an immediate stop in building; they flag that capital, rather than silicon, may set the pace.
FAQ
What did Hock Tan identify as the biggest constraint?
Capital. He said financing could become a larger obstacle than chip, memory, or power shortages.
When and where were the remarks made?
October 12, 2026, at the DBS Asia Future Forward Conference in Singapore.
Did the CEOs say demand is slowing?
No. Both said near-term demand for compute infrastructure still exceeds supply and could do so through 2029 or into 2030.
What did Lip-Bu Tan add?
He pointed to the cost of capital and the longer-run need for companies to show returns on AI investment.
Does this mean AI infrastructure projects will stop?
The executives did not say that. They flagged capital access as a constraint that could become more binding than physical supply shortages.
Last Updated: October 13, 2026.
Related: Startup Department Hub · September 2026 Venture Funding: What a $1.47 Billion Week Reveals About Raising Capital Now · Technology Department Hub
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.