Finance News · Monday, 27 July 2026
01 · Briefing · what happened
Nvidia moves to backstop $250bn of OpenAI's buildout - as Wall Street cools on AI bills
A chipmaker offers to stand behind its biggest customer's borrowing, oil skids on an Iran pause, and the Fed meets under pressure to hike.
Key takeaways
- Nvidia is in talks to guarantee about $250 billion of OpenAI's data-center financing - agreeing to cover the debt if OpenAI cannot.
- Oil fell more than 5% after the U.S. paused strikes on Iran, lifting shares and bonds, though the Fed still meets this week under pressure to hike.
- Wall Street has turned wary of heavy AI spending, and the debt behind the buildout is drawing fresh scrutiny.
The biggest money story this weekend was not a market move. It was a promise. Nvidia, the chipmaker at the center of the AI boom, is in talks to guarantee roughly $250 billion of the financing behind OpenAI’s next data center
The money would help OpenAI, the maker of ChatGPT, lease a 10-gigawatt project that a SoftBank energy subsidiary is building in southern Ohio
Why it matters for anyone with a pension or a power bill: $250 billion is larger than the yearly output of many national economies. And a guarantee does not make the bet safer - it changes who carries the loss if the bet goes wrong. That is the thread running through this week’s markets. (More on the mechanism in today’s lesson.)
The market got a calm weekend
Away from the AI story, markets bounced. Share and bond prices rose across Asia on Monday as oil skidded, easing fears about inflation
Cheaper oil feeds through to cheaper petrol and lower inflation pressure, so the dollar eased and bond yields dipped as traders trimmed their bets on a rate rise
The Fed’s hard week
The U.S. Federal Reserve, America’s central bank, meets Tuesday and Wednesday - the second meeting under new chair Kevin Warsh
That pushed the 10-year Treasury yield up more than 30 basis points (0.30 percentage points) since late June, to about 4.68% - near its highest in a decade
Wall Street stops clapping for AI spending
The Nvidia guarantee lands into a market that has turned wary of AI bills. For years, big tech could spend lavishly on AI and be rewarded as long as revenue rose. That deal is breaking down
Underneath sits a debt question. The smaller AI “neocloud” firms that rent out computing power carry startling loads: CoreWeave’s debt is about 739 times its equity, against roughly 18 for Alphabet
Deals and one to watch
Dealmaking stayed busy. Carlyle and Bain Capital are battling to buy a wealth manager in a possible $7 billion deal
The quieter piece worth noting: the U.S. government now holds about $27 billion in corporate stakes, the largest being 433.3 million Intel shares bought at $20.47 each by the Commerce Department
02 · Lesson · why it matters
When the downside is someone else's, the bet gets bigger
A guarantee does not erase a risk. It moves who carries the loss - and once your own downside is covered, reaching for more becomes the safe move.
The promise underneath the headline
Nvidia is reportedly offering to stand behind about $250 billion of the borrowing behind OpenAI’s next data center. Strip away the size and the plan is simple. A lender puts up money. If OpenAI cannot pay it back, Nvidia does. That is a guarantee: a second name on the loan, one strong enough that the lender stops worrying about the first.
The news reads as a story about scale - the biggest chipmaker, the biggest AI firm, a quarter-trillion-dollar number. But the interesting part is not the size. It is what a guarantee does to everyone’s behaviour once it exists.
What a guarantee actually changes
Picture the lender before the guarantee. OpenAI burns cash and turns no profit. Lending to it is a real gamble, so the lender is careful: small loans, high interest, hard questions. The lender is the brake.
Now add Nvidia’s promise. The loan is the same loan, to the same cash-burning company. But the lender’s slice is safe - if OpenAI stumbles, Nvidia pays. So the careful questions fade. The lender lends more, and lends easier, because the part that used to hurt is now someone else’s problem.
Notice what did not change: OpenAI is exactly as risky as it was this morning. What changed is who feels that risk. This is the pattern economists call moral hazard - when you are shielded from the downside of a risk, you take more of it. The shield does not remove the danger. It removes the flinch. And the flinch was doing useful work.
The twist that makes it sharper
There is a second loop here worth naming plainly. Nvidia sells the chips that fill these data centers. By guaranteeing the loans that pay for the buildout, Nvidia is helping fund the demand for its own product. The guarantee that lets OpenAI borrow is also the guarantee that lets OpenAI buy more Nvidia chips.
This is not a scandal, and it need not end badly. Guarantees are ordinary and often sensible - a parent co-signs a lease, a bank backs a trade, a strong firm helps a promising one grow faster than it could alone. The arrangement can genuinely serve everyone at once: OpenAI builds, the lender earns, Nvidia sells. The point is not that someone is cheating. The point is that a promise like this quietly rewires who has a reason to be careful. The party with the money to lose is no longer the one deciding how much to risk.
Who else is standing in this circle
It is tempting to file this as a story about three giants and walk on. But the risk that Nvidia’s promise moves does not leave the world. It settles somewhere, and the somewhere includes people who never saw the deal.
The “Magnificent Seven” tech firms sit near the center of almost every index fund, so the AI bet is already inside most pensions and retirement accounts. The data centers being financed run on the same power grid that sets ordinary electricity bills. The loans behind the buildout flow through the credit and bond funds where savings are parked. When a guarantee lets more money reach a bet that later sours, the loss does not stay tidily with the people who signed. It shows up, thinned and rerouted, in places far from the room - a fund’s return, a bill, a job.
What no single seat can see
Here is the humbling part. Stand in any one seat and the picture looks safe. The lender sees a guaranteed loan and calls it low-risk. Nvidia sees firm orders on the books and calls it a strong quarter. The pension holder sees a rising index and calls it wealth. Each seat is telling the truth about its own view. Each is being reasonable.
The risk has not disappeared. It has moved into the spaces between the seats - the connections no one person is watching. Everyone is looking at their own patch, and their own patch looks fine. That is how a system can be full of careful people and still, together, lean further out than any of them would alone.
None of this tells you whether the AI buildout is wise or foolish; that is genuinely unknown, and the people inside it do not know either. It tells you something smaller and more useful. When you read that someone is guaranteeing someone else’s debt, ask the quiet question underneath: whose flinch just got switched off, and where did the risk go when it did. The honest answer is usually that it did not go away. It just found a seat further from the one who chose it. That is worth holding in mind before anyone, including us, feels too sure they can see the whole board.
03 · Lab · your turn
The Lending Desk
Rehearse how a guarantee changes which risks you are willing to fund, and where the loss lands when it is covered.
04 · Hope · carry this
The market's fresh wariness about who really carries these AI bets is itself a kind of progress: a system slowly learning to ask who bears the risk before the loss, not after.
More from Finance News