Wall Street endorsed Jensen Huang’s “big concept” for AI. What now?

The first three-plus years of building artificial intelligence have been paid for by record amounts of stocks and bonds issued by the world’s biggest technology companies, some of which have spent so much of their existing capital that they have negative cash flow.

Nvidia CEO Jensen Huang said he expects the next phase of funding to be raised by large Wall Street brokers rather than corporate balance sheets.

In an interview with CNBC on Monday, Hwang called his plan a “big concept” and unveiled it in front of cameras with world leaders. goldman sachs, black rock, black stone, KKR, apollo and brookfield. As chipmakers and hyperscalers race to meet seemingly endless demand, the companies say they are willing to lend a combined $500 billion, and in some cases more, to build or expand new AI factories.

One by one, Hwang and his big-money partners explained what they see as a fundamental shift in the technology industry: AI infrastructure has become a new asset class.

“These systems are not like our PCs or cell phones,” Huang told CNBC’s Becky Quick. “These are now income-generating assets: they are highly productive, long-lived, fungible, and flexible.”

There were few concrete discussions about what kind of borrowers would appear, what the interest rates would be, where the facility would be built or when it would start. In a joint press release, the two companies said they had signed a memorandum of understanding, without mentioning any agreement.

Details matter. About 11 months ago, Nvidia announced a partnership to invest up to $100 billion in OpenAI as part of a plan to build data centers that would require a total of 10 gigawatts of power. That investment didn’t materialize, but Nvidia contributed $30 billion to OpenAI’s record funding round that it closed earlier this year.

Monday’s announcement struck a different tone, with both companies collectively pushing the message that money doesn’t matter at a time when McKinsey expects to spend $7 trillion globally by the end of 2010 to ramp up AI.

This year so far, alphabet, Amazon, Meta, microsoft and oracle In total, we raised well over $150 billion through debt and equity sales to build data centers, fund the development of new AI models, and support the explosion of AI agents. intel It just announced a $15 billion initial public offering, which was later increased to $20 billion.

Executives such as Goldman Sachs CEO David Solomon and KKR’s Waldemar Schlezak see AI devices acquiring familiar money-making properties, and financial firms are now preparing to enter the market in different ways.

“In some ways, we’re starting to see asset-based financing for this infrastructure,” Solomon said on a CNBC panel. “That’s not surprising because these are real assets. They have real value.”

Instead of viewing supercomputers as devices that customers buy and use, these systems, powered by Nvidia’s graphics processing units that cost $3 million per rack, look like profitable investments. Huang says his CUDA software allows him to improve the system, extending its life and leading to better economics.

“You can think of this as a revenue stream, and you can securitize it or sell it to investors who want to participate anywhere in that stack, effectively splitting that risk,” said Schlezak, KKR’s head of digital infrastructure.

As Wall Street becomes significantly excited about the securitization of physical assets, natural questions arise. “What could possibly go wrong?”

One of the hallmarks of the 2007-2009 financial crisis was the packaging of subprime loans into bundled securities and selling them to investors as another way to profit from the housing boom. As mortgage defaults began to rise, the entire system began to loosen.

Michael Varley, a prominent short seller who made his fortune betting on subprime mortgages, suggested late last year that companies like Meta, Oracle, Microsoft, Google and Amazon were exaggerating the useful lives of their AI chips and underestimating their depreciation costs.

Although subprime issues were not a topic of conversation on Monday, several lenders acknowledged that there is some risk in AI trading.

“There will be overreach and there will be setbacks,” said Jim Zelter, president of Apollo Global Management, adding that the number of participants in the project allays concentration concerns.

“Some big companies will win,” Solomon said. “There will be large companies whose results are different from what people expected.”

In discussing BlackRock’s role in Monday’s deal, CEO Larry Fink drew a direct comparison to the mortgage market, but referenced the decades-old era of the housing boom and bust.

“This is just the beginning, like when I entered the mortgage-backed securities market in the 1970s,” Fink said. “I see this as the next future of financial engineering.”

All six lenders make their own lending decisions, Huang said in an interview, noting that NVIDIA connects customers with lending partners.

Nvidia said it will have the option to backstop 25% of all loans, making it a more favorable interest rate structure for companies that previously had to rely on their own credit ratings. Borrowers must use the system architecture specified by Nvidia so that “if something were to happen,” another company could take over and operate it, Huang said.

Nvidia still has a lot to work out with its funding partners, but Monday’s rally was a big step in showing the type of funding available to other companies in the ecosystem. Brookfield CEO Bruce Flatt said Hwang has created a format that investors need.

“Jensen is leading this to create structure,” Flatt said. “Because there are hundreds of trillions of dollars in the world.”

WATCH: ‘Fast Money’ traders react to Nvidia partnership