The first three-plus years of the artificial intelligence build-out have been paid for through record amounts of equity and debt issued by the world's leading tech companies, some of whom are spending so much of their existing capital that they've turned cash flow negative.
Nvidia CEO Jensen Huang just revealed what he expects to be the next phase of financing, backed not by corporate balance sheets, but by Wall Street's top power brokers.
In an interview with CNBC on Monday, Huang called his plan a "big concept," unveiling it on camera alongside leaders from Goldman Sachs, BlackRock, Blackstone, KKR, Apollo and Brookfield. Together, those firms say they're willing to raise $500 billion, and potentially more, for the construction and build-out of new AI factories, as chipmakers and hyperscalers race to meet seemingly endless demand.
Huang and his big-money partners, one by one, described what they view as a fundamental shift in the tech industry: AI infrastructure has become a new asset class.
"These systems are not like our PCs, not like our phones," Huang told CNBC's Becky Quick. "These are revenue-generating assets now. They're productive, they're long-lived, they're fungible, they're flexible."
The discussion was thin on specifics as far as the types of borrowers that will emerge, what interest rates will look like, where the facilities will be constructed and when it will all kick off. Their joint press release said the companies had signed memos of understanding, with no reference to any contracts.
The details matter. Almost 11 months ago, Nvidia announced a partnership to invest up to $100 billion in OpenAI as part of a plan to build out data centers requiring a combined 10 gigawatts of power. That investment never materialized, but Nvidia contributed $30 billion to the record-breaking funding round that OpenAI closed earlier this year.
Monday's announcement struck a different tone, with the companies collectively pushing the message that money won't be the problem as the AI build-out hits what McKinsey expects will be $7 trillion in global outlays by the end of the decade.
So far this year, Alphabet, Amazon, Meta, Microsoft and Oracle have raised well over $150 billion combined by selling debt and equity to build data centers and fund the development of new AI models and support the explosion of AI agents. Intel just announced a $15 billion stock offering, then upsized it to $20 billion.
Financial firms are now gearing up to jump into the market in a different way, as executives like Goldman Sachs CEO David Solomon and KKR's Waldemar Szlezak see AI equipment attaining familiar money-making characteristics.
"You're starting to see, in a sense, you know, asset-based financing against this infrastructure build-out," Solomon said on the CNBC panel. "That's not surprising because these are real assets. They have real value."
Instead of seeing supercomputers as devices that customers buy and use — the argument goes — these systems, filled with Nvidia's graphics processing units that can cost $3 million per rack, look like profitable investments. Huang says the systems can be improved through his company's CUDA software, and their lifespans extended, leading to better economics.
"You can think about it as a revenue stream, and you can securitize it or effectively divide that risk and sell it to investors who want to participate anywhere in that stack," said Szlezak, KKR's head of digital infrastructure.
When Wall Street starts getting noticeably excited about securitizing physical assets, a natural question emerges: What could go wrong?
One of the hallmarks of the financial crisis of 2007 to 2009 was the packaging of subprime mortgages into bundled securities that were then sold to investors as another way to make money from the housing boom. When mortgage defaults started going up, the whole system began to unwind.