Speaking in Singapore, Broadcom CEO Hock Tan and Intel CEO Lip-Bu Tan warn that access to capital and rising borrowing costs could become AI's main bottleneck
Lee Ying Shan /CNBC:
Context & Ripple Effects
Broadcom had framed AI chips as an offset to its cyclical businesses, projecting $10 billion in 2024 AI-chip revenue after AI revenue quadrupled in its first fiscal quarter. The emphasis has shifted from demand for chips to the financing capacity behind the infrastructure buildout.
For Intel, capital availability is already an operational concern: its board reportedly stalled Lip-Bu Tan's efforts to raise capital and pursue an AI acquisition in 2025. Tan's appearance alongside Hock Tan ties that company-specific constraint to a wider AI investment question.
First-order effects
- AI labs and infrastructure buyers using borrowed money face a higher effective cost for expanding compute capacity as borrowing costs rise.
- Broadcom and Intel's customers must weigh financing terms alongside chip availability when planning AI infrastructure purchases.
Second-order effects
- Chip suppliers selling into large AI buildouts face customers that may phase orders, favor projects with clearer returns, or seek alternative financing structures.
- Intel's capital constraints become more consequential if AI competition increasingly depends on funding capacity as well as product execution.
Third-order effects
- If capital access becomes a binding constraint, AI infrastructure spending will favor companies and projects able to demonstrate durable returns rather than simply secure hardware.
- The AI buildout is moving toward a more explicit debate over speculative investment and lab economics, making the cost of capital a competitive variable across the compute supply chain.
The trend: AI's bottleneck is broadening from physical compute supply to the financial capacity required to fund and earn returns on infrastructure.