No. of Recommendations: 5
Remember for scale - U.S. GDP is $32 Trillion.
nytimes.com - AI lending nvidia blackrock
The Perils of Wall St.’s Race to Pour Billions More Into A.I.
Investors including BlackRock have committed an eye-popping amount to lend to the artificial intelligence industry. That enthusiasm isn’t shared by everyone.
By Andrew Ross SorkinBernhard WarnerSarah KesslerMichael J. de la MercedNiko GalloglyBrian O’Keefe and Sri Muppidi, The New York Times, Aug. 11, 2026
Nvidia’s announcement that it was teaming up with a half-dozen Wall Street firms to lend $500 billion for the A.I. buildout adds a huge amount of debt to the A.I. economy. It’s a clever move for Nvidia, which will go from being the de facto financial backstop for the industry (for which it has been criticized) to shifting systemic risk to Wall Street investors...
“We need to raise this money as fast as possible,” Larry Fink, the C.E.O. of BlackRock, said on Monday in announcing the initiative. His firm and several other heavyweights — Apollo! Blackstone! Brookfield! Goldman Sachs! KKR! — said that it was imperative to help tech companies amass the resources needed for their A.I. expansion efforts.
The investment firms will dole out the funds via loans, credit and more. (They’ll be available “at attractive rates,” according to Nvidia.)
Other details weren’t available, though the Wall Street firms alluded to “yield-based products” and securitization. Fink said he foresaw “a next future for financial engineering.”
The context: The A.I. boom is already hugely expensive. Morgan Stanley analysts predict that hyperscalers will invest $3.5 trillion into A.I. infrastructure between now and 2028. Model developers are also spending heavily, with Anthropic reportedly agreeing to lease computing capacity from Riot Platforms for $9 billion... [end quote]
A next future for financial engineering! What could possibly go wrong?
Sales of AI services to end-users outside the AI circular ecosystem is orders of magnitude lower than the spending. And that’s even before “distilled” AI (specialized small language models (SLMs)) is widely available to provide 90% of user applications at a fraction of the cost of “frontier” AI.
Not to mention...where are all the data centers going to be built with such fierce local push back against them? A lot are going into Texas - which isn’t exactly swimming in the fresh water needed to operate them.
Wendy
