Particle.news

CEOs Warn Capital Shortage Could Slow AI Buildout

Rising borrowing costs plus concentrated private financing raise the risk that long‑dated data‑center and chip projects will struggle to show profit.

Overview

  • Broadcom chief Hock Tan and Intel chief Lip‑Bu Tan, which spoke at a conference on Monday, said access to affordable long‑term capital is likely to become the main constraint on expanding AI infrastructure.
  • Both executives said demand for AI compute still outstrips supply and that the imbalance could persist through 2029 into 2030, even as earlier physical bottlenecks ease.
  • They described memory shortages as likely to improve within a year while wafer capacity and power constraints should ease by 2028–2029, shifting the problem from hardware to financing.
  • Broadcom has set up a reported $35 billion financing platform with Apollo and Blackstone and there are reported multi‑billion debt discussions for major AI customers like OpenAI and Anthropic, concentrating risk among a few players and lenders.
  • If long payback horizons collide with higher bond yields and investor pressure for quicker returns, projects for data centers and custom chips could be delayed or trimmed, which would slow jobs, local investment and wider AI deployment.