Wall Street has spent the last week absorbing Nvidia's (NVDA) plan to arrange half a trillion dollars of other people's money, and the sharpest objection came from someone who was explicitly trying to argue the other side. Asked on the All-In podcast published Aug. 14 where the arrangement could break, David Sacks, while still an overall bull on AI and similar, is acutely aware of the risks of the massive, multi-trillion-dollar AI buildout, saying, "The biggest risk to me is not on the demand side," he said. "The biggest risk is that you get a glut of compute and you get an overbuild. And in the same way that we had dark fiber after the dotcom crash, if you had dark GPUs, that'd be a disaster for everyone."
On Aug. 10, Nvidia said it had signed memorandums of understanding with Apollo (APO), BlackRock (BLK), Blackstone (BX), Brookfield (BN), Goldman Sachs (GS), and KKR (KKR) to establish independent compute financing platforms to "mobilize over $500 billion of third-party capital for the buildout of AI infrastructure over time." The release adds that the partnerships "remain subject to execution of the final agreements." Meaning, that figure is an aspiration, not raised money, not committed money, and not Nvidia revenue. Several outlets blurred the distinction.
More Top Stocks Daily: Go behind Wall Street’s hottest headlines with Barchart’s Active Investor newsletter.
Read more: Unlocking $2.1T in Energy Potential. See Why Investors Are Watching ‘FASF’.
Sacks co-founded Craft Ventures and co-hosts the podcast. He served as the White House AI and crypto czar until March 2026 and now co-chairs the President's Council of Advisors on Science and Technology, a council whose initial membership includes Jensen Huang. Readers should weigh a warning about Nvidia's structure with that proximity in mind, in both directions.
The failure mode he names is not the one most people reach for. The bear case on AI infrastructure is usually that demand disappoints. Sacks is describing something else: demand is real, and too many people race to serve it at once. The telecom buildout of the late 1990s left behind years of fiber that was laid but never lit, and the companies that financed it did not survive to see it used. An oversupply that collapsed rental rates would damage every link in the chain: the neoclouds that borrowed, the lenders that underwrote them, and the vendor standing behind the residual values.
Then he argued himself partway out of it, which is the more interesting half. Sacks said that "in a weird way all the political headwinds I think insure against that outcome, because it is so hard to build data centers," pointing to what he called a "moral panic" over the buildout. His conclusion was that those same headwinds "will almost guarantee that there's not an oversupply relative to the exponentially growing demand. So in a weird way, you're protected against that." It is an unusual position: the local opposition, permitting friction and political hostility that developers treat as the industry's greatest obstacle, is, on his reading, the thing preventing its worst financial outcome.
The bull framing he was pushing against came from investor Gavin Baker on the same episode, who reached for a different institution entirely. "In some ways, Nvidia is kind of becoming the central bank of AI, the Federal Reserve of AI," Baker said. "And in the same way, the Federal Reserve, like a lot of monetary policy, it works through private banks." That earns its keep on transmission; Nvidia is not writing the checks, the six firms are, and breaks on the obvious point that Nvidia is also the vendor. A central bank does not sell the asset it stands behind.
What makes the paper financeable is a guarantee, and where it comes from matters. Nvidia's release contains no residual-value language at all. Huang described the mechanism separately, capping the guarantee at roughly 25% of a transaction's residual value, project by project, with credit decisions staying with the capital providers. That cap determines how much of the glut risk Sacks describes lands back on Nvidia rather than on Blackstone or Goldman.
All of it rests on one checkable figure: how long a GPU earns. CoreWeave's (CRWV) most recent annual report puts the useful life of technology equipment, the line containing its GPUs, at six years. On its Aug. 10 earnings call, chief executive Mike Intrator said it had signed a contract for Nvidia A100 chips running into 2029, nine years after Ampere launched and three beyond that schedule, adding that "pricing for prior generation SKUs is at or above where it was years ago." If older silicon keeps earning that long, the glut is far away. If real economic life is four years, it is not.
The sell side is treating the arrangement less as risk transfer than as a new revenue line. Morgan Stanley's Joseph Moore kept an “Overweight” rating and a $288 target, arguing third-party capital "should arguably alleviate circularity concerns" and help "unlock annuity revenue streams." That is a different debate from the one Barchart covered when Bank of America argued Nvidia stock was cheap, and it sits awkwardly beside Huang's insistence that there is no AI bubble because AI is profitable. If the economics were unambiguous, no guarantee would be needed.
The thing to watch is disclosure. Nvidia's most recent quarterly filing shows one relevant obligation: a facility lease guarantee with maximum gross exposure of $3.5 billion, partly escrowed, classified as a credit derivative, and called immaterial. No residual-value commitment appears in any filing yet because the announcement postdates the last one. The first that could carry it lands after the company reports on Aug. 26. Until then, Nvidia's actual exposure to the scenario Sacks described is something investors are taking on discretion.
On the date of publication, Caleb Naysmith did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. For more information please view the Barchart Disclosure Policy here.
More news from Barchart
Warren Buffett’s Successor Bought Back 19 Times More Berkshire Stock Than Last Quarter, and the Filing Says It Was Abel’s Decision, Not Buffett’s David Sacks Says the Biggest Risk to the AI Buildout Isn’t Demand, ‘If You Had Dark GPUs, That’d Be a Disaster for Everyone’ Famous Hedge Fund Tiger Global Opened a New Position in This 1 AI Stock This Stock’s Already Up Nearly 600% Over the Past Year