Broadcom chief executive Hock Tan and Intel chief executive Lip-Bu Tan both named capital as the likeliest bottleneck for the AI buildout, speaking at the DBS Asia Future Forward Conference in Singapore on Monday.
Hock Tan said financing could grow into a bigger obstacle than semiconductors, memory or electricity. I think might be the biggest constraint is capital, he said, while acknowledging that supply bottlenecks have surfaced repeatedly over the past two years and been worked through each time.
He expects some of the current pressure to ease. Wafer shortages could loosen by 2028 or 2029, and power availability should improve as new generation and transmission comes online. Memory, he suggested, may look very different a year from now.
Lip-Bu Tan focused on cost rather than availability. With the cost of capital rising, he said the long-run question is whether the companies spending heavily on AI will produce enough return to justify it, and that investors will want to see that ROI.
Neither executive expected near-term demand to soften. Hock Tan said there is more demand for compute infrastructure than the industry can supply, an imbalance he expects to persist through 2029 and possibly into 2030. Intel’s Tan said a significant slowdown looks unlikely over the next two to three years.
The warnings land as rising bond yields make data center construction more expensive to finance. Projects that take years to generate revenue are increasingly sensitive to borrowing costs, which puts pressure on the economics of capacity now being ordered.
