Key Takeaways

  • The 10-year US Treasury yield jumped 12%, triggering a debate over whether massive capital expenditure for AI infrastructure is driving global interest rates higher.
  • Jordi Hays notes that hyperscaler and AI hardware debt issuance equals roughly 70% of projected US Treasury bond volume through 2026.
  • John Coogan questions whether traditional fixed-income allocators will ever abandon sovereign treasuries, which are backed by government taxation power, to fund private data centers.
  • Software startups launched in a 4% to 6% rate environment are structurally safer than zero-interest-rate SaaS companies that relied on 100x revenue multiples and zero near-term earnings.

The Disagreement

Tech capital expenditures are reaching historic highs, and bond markets are feeling the strain. On TBPN, Jordi Hays and John Coogan debated whether the scramble for computing capacity is pulling yields up across the global economy.

Coogan opened the case for structural rate pressure: “To what degree is the AI buildout driving interest rates upwards? People are going back and forth about this. Interest rates are spiking. Specifically, the 10-year yield has jumped 12%.” He outlined the aggressive thesis: “Global yields will continue to go higher because why would you invest in literally anything other than data centers when they have such a short payback period and you can throw tens of billions at it.”

Hays reinforced the scale of the debt flood. “Hyperscaler and Nvidia like debt issuance as a percentage of like total treasury bond issuance through 2026 has been like 70%,” Hays noted. When private cloud operators and hyperscalers issue hundreds of billions in corporate debt and private credit to secure chips and megawatts, they compete directly for the world's pool of lendable cash.

Yet Coogan pushed back against treating sovereign debt and AI infrastructure loans as interchangeable assets. “If you're a bond investor and you're looking at two different markets, are you really shifting out of government treasuries that are historically the risk-free rate that they have a monopoly on violence? They have the ability to tax their citizens to pay back the debt.”

Who's Right (and When They're Wrong)

Hays is right about the competition for private credit. When Neoclouds and utilities borrow tens of billions at premium spreads, institutional money managers demand higher yields everywhere else. Energy costs and supply bottlenecks add fuel to consumer inflation, keeping central banks on defense.

Coogan is right about the institutional floor. Pension funds, central banks, and sovereign wealth vehicles cannot replace risk-free government paper with private data center debt, no matter how fast GPU clusters pay for themselves. A sovereign default risk profile does not match a server farm's obsolescence risk.

For builders, the real consequence is how high rates reshape startup mechanics. Coogan explains the divide between software generations:

“I do believe that if you're a software company trading at a 100x revenue multiple and you weren't forecasting real earnings for 10 years or something like that like the traditional SAS playbook and interest rates go from zero to 4%. That's a lot more damaging than okay, you were already born in the era of higher interest rates and interest rates go from four to 6%.”

Zero-rate SaaS was designed for an imaginary world where capital had zero cost and earnings could wait ten years. Today's AI founders are forced to price their compute costs on day one, bill customers immediately, and build for real unit economics.

What to Do With This

Audit your startup runway under an assumed 6% risk-free rate through 2026. Calculate your compute payback timeline on every dollar spent on model training or inference hosting. If your gross margins cannot clear your underlying compute debt within twelve months, rework your pricing contracts before seeking your next capital round.