Let There Be Compute! (KKR, September 2026): the AI buildout needs $6.3T beyond public investment-grade markets

KKR's Credit and Markets team reads the AI buildout through a credit lens: not how much compute gets built, but who gets paid if expectations change. AI infrastructure capex through 2030 is now estimated at $7.6 trillion to $8 trillion. The hyperscalers could raise roughly $1.7 trillion in public investment-grade bonds before hitting index concentration limits, which leaves about $6.3 trillion that has to come from internal cash flow, private investment-grade debt, asset-based finance, infrastructure debt and structured deals. AI-linked credit issuance was about $189 billion in 2025 and has more than doubled year over year, and US tech capex is now 5.8% of GDP. The letter argues that the index looks calm while dispersion underneath it rises, with software credit selling off on AI disruption fears and hyperscaler spreads diverging from the rest of investment grade. For AI founders, this explains where the money financing data centers, chips and power is coming from, why capital structure discipline is tightening, and why lenders are starting to ask harder questions about asset lives and revenue durability.