The same banks that financed the 2008 housing bubble are now bankrolling a $600 billion AI boom they openly admit might be built on sand.
The Summary
- Big banks continue funding AI infrastructure despite acknowledging significant financing risks, creating potential systemic instability
- VCs are using circular funding models — investing in AI companies, then giving those same companies capital to buy cloud credits from VC-backed infrastructure providers, creating a $600B shell game
- The pattern mirrors pre-2008 crisis dynamics: distorted market signals, inflated valuations, and institutional players aware of the risks but too deep to pull out
The Signal
Banks are writing checks to fuel the AI boom while simultaneously warning their own risk committees about it. The circular funding structure works like this: venture capital firms invest in AI startups, then provide those startups with additional capital earmarked specifically for buying compute from cloud providers that the same VCs have backed. The money loops through the system, creating the illusion of organic demand.
Bloomberg's reporting exposes a $600 billion market where the money isn't flowing from customers to companies. It's flowing from investors to companies to other investor-backed companies and back again. Real revenue gets harder to distinguish from funded purchases.
"The circular funding model in AI could lead to distorted market signals and financial instability, reminiscent of pre-2008 crisis dynamics."
This isn't small-time angel investing. Major banks are financing this infrastructure buildout despite internal risk assessments flagging the same warning signs that preceded previous tech crashes. They know the risks. They're funding it anyway because everyone else is, and being the first to pull back means missing the upside if this turns out to be real.
The mechanics are familiar to anyone who lived through 2008:
- Asset prices inflated by cheap, plentiful capital
- Risk acknowledged but deemed acceptable because "this time is different"
- Circular dependencies where value relies on continued funding rather than fundamental economics
- Institutional players too invested to exit without triggering the collapse they fear
The pattern threatens market balance because it makes price discovery impossible. When GPU clusters get financed by the same capital that finances the companies using those clusters, what's the real market rate? What's the actual demand? Nobody knows, and that uncertainty is a feature, not a bug, until the music stops.
The Implication
If you're building in AI, understand that your access to compute might not reflect real market conditions. The infrastructure you're relying on exists because of financial engineering, not sustainable economics. Plan for what happens when the circular funding stops and prices find their actual level.
For everyone else, watch where the banks actually put their *own* capital versus where they're facilitating funding. When institutions acknowledge risks but keep writing checks, they're not betting on fundamentals. They're betting they can exit before the correction. The agents being built today might be revolutionary. The financing structure behind them is not.