Research firm SemiAnalysis says the next bottleneck for AI is financing, not GPUs or data centers, and that credit markets will be the primary funding source for AI infrastructure. SemiAnalysis projects annual global AI capex will exceed $2.0 trillio

2026-07-08

Research firm SemiAnalysis says the next bottleneck for AI is financing, not GPUs or data centers, and that credit markets will be the primary funding source for AI infrastructure. SemiAnalysis projects annual global AI capex will exceed $2.0 trillion by 2028 and cumulative AI capex from 2024–2029 will reach about $11.1 trillion. The firm says NVIDIA is providing minimum-revenue guarantees to AI compute operators such as Neocloud to make GPUs financeable as infrastructure. Based on this, SemiAnalysis forecasts outstanding AI-related debt could top $7.0 trillion by 2029, making it the second-largest asset-backed debt market after U.S. mortgage-backed securities (around $13.0 trillion). NVIDIA’s described back-up plan would use its AA/Aa2 investment-grade credit to underwrite minimum revenue for operators, lowering bank lending risk and unlocking multi-trillion dollars of AI infrastructure financing; SemiAnalysis likens NVIDIA to the “central bank” of AI.