Nvidia has made an absolute fortune selling AI chips. Hundreds of billions in profit in under four years. Everybody waned AI, everybody building AI needs enormous amounts of computing power, and Nvidia sells the chips that make most of it possible. Revenue exploded, profits exploded, and the company became one of the most valuable corporations in history.
But some are starting to sound the warning bells on how it may help start the next chapter in economic troubles for our economy.
According to the Wall Street Journal, Nvidia has recently taken on backstops that could leave it responsible for roughly $230 billion in lease obligations and residual-value arrangements.
A $105 billion backstop on an OpenAI data center lease in Ohio. Up to $125 billion in what’s being called “residual-value support” attached to a $500 billion chip financing plan with major Wall Street asset managers, meaning Nvidia is promising the assets securing those loans won’t fall below a set value. A $4.9 billion minimum revenue guarantee for Sharon AI, one of two young Australian cloud companies where Nvidia has agreed to become the customer of last resort if they can’t find anyone to rent their capacity to.
That figure doesn’t include the $6.3 billion deal from last year to buy unsold cloud capacity from CoreWeave. It doesn’t include the $72.5 billion in public and private equity stakes Nvidia held at the end of its last fiscal quarter, many of them in the same companies it’s writing backstops for.
So Nvidia sells chips to a company, owns a piece of that company, guarantees the loans that company used to buy the chips, and agrees to become the customer if the company can’t find one. The money goes in a circle and comes back out looking like revenue.
Nvidia called the Australian arrangements a “new business model.” That’s the phrase that matters. Not a favor for a strategic partner. A model. Something they plan to do again.
Nvidia Doesn’t Just Want to Sell Chips Anymore
Earlier this month, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR designed to mobilize more than $500 billion in third-party financing for AI infrastructure. Nvidia says the financial firms will independently underwrite the projects, while Nvidia may provide residual-value support covering as much as 25 percent of certain opportunities.
Put that into normal English.
A company wants to build an enormous AI facility filled with Nvidia hardware. Wall Street lends against the project and the equipment. Nvidia may then step in and essentially say that, within certain limits, those Nvidia assets will still be worth something if the deal goes bad.
If the planned financing platforms eventually reach $500 billion and Nvidia backs as much as 25 percent, the potential residual-value support reaches $125 billion.
Then there is OpenAI.
Nvidia is supporting portions of a 20-year infrastructure arrangement at the PORTS-Pike Technology Campus in Ohio, where OpenAI is expected to be the tenant. Nvidia says its support is limited to defined portions of lease and power payments and a residual-value commitment rather than the entire cost of the project, with the exposure declining as facilities come online and OpenAI makes payments. The Wall Street Journal puts the potential backstop at around $105 billion.
And these aren’t the only arrangements.
Last year Nvidia signed a $6.3 billion agreement with CoreWeave under which Nvidia is obligated, subject to the contract terms, to purchase residual cloud capacity that CoreWeave cannot sell to other customers through 2032. That isn’t speculation; CoreWeave disclosed the arrangement in an SEC filing.
In Australia, Sharon AI disclosed a six-year Nvidia collaboration worth up to roughly $4.9 billion involving as many as 40,000 Nvidia GPUs. Sharon describes the arrangement as a revenue-sharing and credit-support model designed to allow the company to commit to Nvidia infrastructure while Nvidia receives hardware revenue plus a portion of the cloud revenue generated by that capacity.
Nobody should pretend this is simply Nvidia selling chips anymore.
Nvidia itself is talking about creating an entirely new financing model in which AI computing infrastructure becomes an investable asset class.cern.
We’ve Seen Versions of This Movie Before
There is an obvious historical comparison here, although it needs to be made carefully.
Lucent Technologies was one of the giants of the telecommunications boom in the late 1990s. As telecom companies raced to build networks, equipment manufacturers increasingly helped customers finance enormous purchases. Vendor financing allowed companies without mountains of cash to buy mountains of equipment, which in turn allowed manufacturers to keep reporting enormous sales growth.
That worked beautifully until it didn’t.
Lucent later collapsed along with much of the telecom bubble. The SEC also accused the company of improperly recognizing more than $1.1 billion in revenue during fiscal 2000 through undisclosed side agreements, credits and other incentives. Lucent eventually settled the case.
To be clear, there is no evidence Nvidia is committing Lucent-style accounting fraud, and that isn’t the point of the comparison.
The lesson is what happens when a booming industry reaches the stage where selling the product isn’t enough. The seller begins helping create the financing necessary for additional buyers to keep buying.
That can extend a boom much longer than skeptics expect. It can also make it extremely difficult to determine how much demand exists because customers genuinely need the product and how much exists because the financial machinery has been constructed to make the purchase possible.
And Now Wall Street Wants to Turn GPUs Into an Asset Class
This is where the story gets even more interesting.
Nvidia isn’t hiding what it wants to accomplish. Jensen Huang has argued that Nvidia compute can become a long-lived, revenue-producing infrastructure asset similar to other things Wall Street finances for decades. The company says GPUs are transferable, reusable and supported by its CUDA software ecosystem, which should help preserve their economic value.
Maybe Nvidia is right.
But notice where we are in the cycle. We are no longer merely talking about companies buying computers because they expect to make money using them. We are talking about creating standardized financial products backed by the future value and cash flow of enormous piles of rapidly evolving computer hardware.
Anyone who lived through the housing bubble should at least feel their eyebrow moving.
The argument behind mortgage securitization wasn’t originally insane either. Houses were real assets. People needed places to live. Mortgages produced predictable cash flows. Historical default rates looked manageable. Wall Street found ways to bring enormous amounts of outside capital into the market.
The problem wasn’t that houses weren’t real. The problem was what happened after cheap and abundant financing changed the behavior of everyone participating in the market.
That’s the question worth asking about AI.
This Is Bigger Than Nvidia Stock
You might not own a single share of Nvidia and still have plenty riding on this boom.
America is currently pouring extraordinary amounts of money into data centers, power generation, electrical equipment, transformers, cooling systems, construction machinery, concrete, transmission infrastructure and backup generation. Companies throughout the industrial economy are expanding factories and hiring workers specifically because of the data center buildout.
The five major technology companies tracked by Reuters already have roughly $1.09 trillion in future lease commitments, much of it related to data centers. These are facilities that often haven’t even begun operating yet, meaning huge portions of today’s AI economy are based on assumptions about computing demand years into the future.
At the same time, the federal government just crossed $40 trillion in national debt. Interest expense is exploding, long-term Treasury yields are under pressure, and Washington needs an enormous amount of capital simply to keep financing itself. Reuters noted this week that the AI investment boom is now competing with government borrowing for the same global pool of savings.
That should sound familiar to anybody who has been reading OFFGRIDSurvival over the last couple of years.
We have an economy increasingly dependent on debt while the cost of carrying that debt remains high. Washington needs credit. Corporations need credit. Consumers need credit. The AI industry now needs hundreds of billions more in credit, and Wall Street is racing to manufacture new ways to provide it.
Meanwhile, American households are carrying about $18.8 trillion in debt. Credit card balances hit roughly $1.26 trillion during the second quarter, while auto debt climbed to $1.71 trillion. The Federal Reserve has also found Americans using buy-now-pay-later loans for things as basic as groceries and food delivery.
At some point you have to notice the common denominator. Everybody is borrowing from tomorrow.
Are You Prepared for an Economic Collapse?
If you haven’t seriously looked at your economic preparedness plan, start with our complete guide to preparing for an economic collapse. We cover the warning signs, financial steps, survival supplies, self-reliance skills and preparations that can help protect your family during a prolonged economic crisis.
You should also read:
Read the full article here
