View / Big tech meets Milton Friedman

A cardinal rule of finance is “always spend other people’s money.” It’s why leveraged buyouts include as much debt and as little equity as possible, and why the wealthy fund their lifestyles by borrowing against assets. The incentives, though, can be less than ideal.

Milton Friedman, the legendary economist, laid out a simple matrix. Spend your own money on yourself and you’re disciplined on cost (it’s your money) and clear on value (you know what you want.). That’s an efficient economy, though not a very vibrant one. Spend your own money on someone else and you keep the cost discipline but lose the value signal (the recipient might not want what you bought them). Spend other people’s money on yourself and you lose the plot on costs and sometimes get a subprime mortgage crisis. And finally, spend other people’s money on other people — that’s why people hate the government.

Big Tech is working its way through all four stages, in roughly the order Friedman described. The AI capex supercycle began with tech giants spending their profits on their own infrastructure — chips and data centers for internal use. Then came other people’s money: Intel’s $15 billion stock offering this week follows big capital raises from Alphabet, Meta, Oracle, and Amazon, all deploying shareholder and bondholder capital on their own buildouts.

Now big tech companies are spending other people’s money on other people. Nvidia is assembling a $500 billion pot from a who’s who of Wall Street firms which will, in turn, backstop OpenAI’s lease of a SoftBank data center in Ohio and help startups rent access to Nvidia’s chips from CoreWeave. Broadcom is similarly spending Blackstone’s money to help Anthropic buy chips. Google is spending bondholders’ money to help Fluidstack buy compute from TeraWulf.

The names aren’t important here, but “[Big Tech Company] is spending [Capital Source]’s money to buy [Expensive Gift] for [Small Tech Company]” is a decent way to think about the current state of AI financing. Wall Street capital sources, rather than tech companies, being the source of AI capital is how it should work. But heed Friedman’s warning nobody in that arrangement is “going to be anything like as careful” as those in the first quadrant of spenders. He never met a hyperscaler, but he’s got their number.