The companies building the future are now competing with the government for the money to fund it.
The Summary
- AI infrastructure debt is pressuring long-end Treasury demand, says Goldman Sachs Asset Management's Lindsay Rosner — a real-world collision between Web4 buildout and sovereign borrowing.
- The classic "crowding out" theory — where government borrowing pushes up corporate rates — is back, but this time it's AI companies doing the crowding.
- Treasury yields rise not from policy uncertainty, but from the sheer capital demands of training clusters, data centers, and the physical layer of the agent economy.
The Signal
AI buildout is creating a financing problem nobody saw coming. Goldman's Rosner points to AI debt as a structural weight on long-end Treasuries, meaning the 10-year and 30-year bonds that fund everything from mortgages to corporate expansion. Companies racing to build frontier models and the infrastructure to run them are issuing debt at a scale that competes with sovereign borrowing. The result: higher yields across the curve.
This isn't a temporary spike. As one analyst put it, AI "just touches everything" — from power grid upgrades to chip fab financing to the billions required for next-gen training runs. When Microsoft, Google, and a dozen well-funded startups all need tens of billions simultaneously, they're not just competing with each other. They're competing with the U.S. Treasury for investor dollars.
"The crowding out theory is back, but this time it's AI companies doing the crowding."
Here's what makes this different from past tech booms:
- The capital intensity is orders of magnitude higher — data centers cost billions, not millions
- The timeline is compressed — everyone's racing to AGI simultaneously, not iterating over decades
- The infrastructure is physical and expensive — you can't virtualize a power substation
Rosner also noted she doesn't expect clarity from FOMC minutes on how the Fed will respond. That's the quiet part: monetary policy is still calibrated for the old economy. The Fed has no playbook for what happens when the private sector needs to finance the future at sovereign-debt scale.
The irony is thick. Web4 was supposed to decentralize everything, route around gatekeepers, tokenize assets so capital flows friction-free. Instead, the companies building it are running headfirst into the oldest constraint in finance: there's only so much money, and somebody has to pay interest on it.
The Implication
Watch corporate bond spreads and long-end Treasury auctions. If AI debt keeps pressuring yields, two things happen: funding costs rise for everyone (your mortgage, your startup's Series B, your city's infrastructure bonds), and the companies without balance sheets deep enough to absorb higher rates get priced out of the build. This is how booms consolidate into oligopolies.
For anyone holding long-duration bonds or betting on lower rates, the AI infrastructure wave is a headwind you can't ignore. For builders, it means the cost of capital just became a competitive moat.