AI Chip Demand to Outlast 2028, but Capital Access Looms as New Bottleneck

Key points

• Hock Tan of Broadcom stated that financing could be a bigger obstacle to AI expansion than semiconductor, memory, or power shortages.
• Intel CEO Lip-Bu Tan echoed concerns, noting that rising capital costs and the need for clear ROI are major challenges for the sector.
• Both executives maintained that demand for AI computing infrastructure currently exceeds supply, with the imbalance potentially lasting until 2030.
• Tan predicted that wafer shortages may ease by 2028 or 2029, and power constraints could improve as new infrastructure comes online.
• The warnings come amid a sharp sell-off in global chip stocks, driven by Tan’s comments and reports of progress by Chinese memory maker CXMT.

Background

These warnings follow a period of massive debt financing in the AI sector, including a record $60 billion debt package syndicated to fund Anthropic’s chip leases from Broadcom in early October 2026. That deal highlighted the growing reliance on debt to fund AI infrastructure, even as questions about long-term profitability persisted. The current market volatility reflects ongoing tensions between high capital expenditure and uncertain returns.

How outlets are covering it

Broadcom’s Hock Tan emphasized the speculative nature of current AI investing, comparing it to a 'casino' and noting that many investors lack understanding of the underlying business models. He pointed out that OpenAI remains unprofitable despite high revenues. In contrast, Intel’s Lip-Bu Tan focused more on the structural challenge of capital costs and the need for demonstrable ROI. Both agreed that near-term demand is strong, but diverged in emphasis: Tan (Broadcom) highlighted the risk of demand slowing if AI capabilities plateau, while Tan (Intel) stressed the importance of efficiency gains. The Guardian highlighted the market’s reaction to Chinese competition from CXMT, while CNBC focused on the financing bottleneck.

Why it matters

The shift in bottleneck from hardware to capital suggests that the AI boom’s sustainability depends on financial markets’ willingness to fund expensive infrastructure with uncertain returns. If capital becomes too expensive or scarce, it could slow the deployment of AI models, affecting companies from chipmakers to AI labs. The volatility in chip stocks underscores the market’s sensitivity to both competitive threats from China and financing risks.

What to watch

Investors will watch for signs of easing in capital costs and whether AI companies can demonstrate tangible ROI. The performance of Chinese memory makers like CXMT will also be critical, as their progress could affect the competitive landscape for global giants like Samsung and Micron. The next few quarters will test whether the AI industry can sustain its growth without a major financing crunch.