Intel and Broadcom CEOs Warn Capital Shortage May Become AI's Biggest Bottleneck
CEOs of Intel and Broadcom warned at a conference in Singapore that capital acquisition may become the biggest bottleneck in AI development, surpassing shortages in semiconductors, memory, or power.
Broadcom CEO Hock Tan stated at the DBS Asia Future Forward Conference that he believes the biggest constraint is capital; the company is overcoming shortages in wafers and memory, but financing will become a larger obstacle, while Anthropic and OpenAI have demonstrated that the latest models can generate significant revenue.
Intel CEO Lip-Bu Tan similarly pointed out that the challenges ahead are capital and capital costs, as investors want to see returns on investment in the long term; he mentioned that a series of bottlenecks have emerged and been overcome over the past two years, with memory shortages possibly easing after a year, wafer shortages potentially alleviating by 2028 or 2029, and power supply also likely to improve.
The industry is rapidly building infrastructure to support more complex models, leading to increased capital demand; Broadcom has previously established financing tools with Apollo, Blackstone, and others to support customer chip procurement and has committed large loans to Anthropic.
Hock Tan had previously viewed power-ready sites as a less certain bottleneck but has now prioritized capital; both companies are experiencing rapid growth in AI-related demand, with Intel's data center and AI revenue recently seeing significant increases, and Broadcom predicting that AI chip revenue will more than double in the coming years.
Market Mechanism: Capital expenditures for AI infrastructure are driving up financing demand, while rising capital costs put pressure on large-scale deployments; chip and power bottlenecks may ease with capacity expansion, while capital constraints depend on customer revenue proof and financing channels, with leading cloud and model companies benefiting from revenue growth, while capital-intensive expansion projects face higher thresholds.
Source: Public Information
ABAB AI Insight
Hock Tan has long led Broadcom, supporting the expansion of hyperscale and model companies through customized AI chips and customer financing arrangements; Lip-Bu Tan took over as CEO of Intel in recent years, focusing the company on data center CPUs and AI-related demand while adjusting capacity and capital expenditures to match orders.
On the capital front, Broadcom has established financing tools in collaboration with private equity and financial institutions to provide chip procurement support to customers, expanding funding from traditional cash to debt and structured financing; Intel, on the other hand, is increasing factory capacity through stock issuance and internal capital expenditures to respond to CPU demand driven by AI inference.
Similar cases include early data center expansions where both power and capital became constraints, and cloud computing companies supporting infrastructure through large-scale financing; the current phase of AI development is transitioning from chip and power shortages to capital availability becoming the main constraint.
Essentially, this relates to capital concentration and industrial chain restructuring. The rapid rise in demand for AI models and infrastructure has made capital acquisition a major limitation, surpassing physical capacity; leading companies maintain expansion through revenue proof and external financing, while capital costs and return verification determine subsequent capacity and deployment pace.
ABAB News · Cognitive Law
Physical bottlenecks can be overcome by capacity, while capital bottlenecks depend on return proof.
Large-scale infrastructure expansion is ultimately constrained by financing capability.
Revenue verification is a prerequisite for continuous capital acquisition.