The AI Buildout Just Put $140 Billion on Its Credit Card

Close-up of server racks with red and blue cabling inside a data center

Three Blockbuster Deals in One Week

It took less than a week for the AI industry’s financing model to change in public. According to the Wall Street Journal, Broadcom has spent recent weeks arranging more than $50 billion in financing for the custom AI chip it is developing with OpenAI, approaching lenders including Apollo Global Management and Blackstone. Separately, Oracle is in talks with Apollo and Goldman Sachs to fund a large chip purchase of its own.

Then came the biggest number of all. Reuters reports that SpaceX is planning to issue $30 billion in investment-grade debt and raise another $10 billion in loans to buy chips from Nvidia — in what would rank among the largest debt financings of the AI era. The talks, first reported by the Financial Times, are still early-stage and could fall apart, but Apollo is expected to lead with Pimco among the lenders circling.

Stack the three together and more than $140 billion in AI-chip debt is being negotiated at once — a figure that would have been unthinkable back when the AI boom was funded out of Big Tech’s cash reserves.

The Dangerous Loop

What spooked markets wasn’t just the size of the numbers — it was the shape of the deals. Reuters reports that the cost of insuring SpaceX debt against default jumped to record highs on the news, while the company’s shares and bonds lost ground.

Nigel Green, CEO of deVere Group, warned of “a dangerous loop where Nvidia was bankrolling the very customers who buy its products,” leaving global investors exposed if expected profits fail to materialize. “The AI build out started on cash,” he told Reuters. “It’s increasingly running on credit, and credit changes the risk profile entirely.”

The loop is real: Nvidia is a major shareholder in SpaceX, the company now looking to borrow $40 billion to buy Nvidia’s chips. The chipmaker gets paid either way; the borrowers — and their lenders — carry the risk. As Green put it, this debt is landing in the bond funds and pension pots of savers around the world.

Why This Matters: Borrowed Money Keeps a Calendar

Here is what the deal headlines miss: this borrowing binge is arriving at the exact moment enterprises are struggling to show returns on the AI they’ve already bought. A new WSJ Intelligence study of 801 C-suite executives at large US companies found the average enterprise now runs 183 AI tools — yet 85 percent lack the systems to connect them, three in five say tool sprawl has left strategic execution stagnant, and only 15 percent of AI investments operate as a unified system.

In other words, the industry is borrowing record sums to build more AI capacity while most buyers can’t prove the last round paid off. And the sums coming dwarf this week’s deals: Morgan Stanley estimates AI infrastructure will require $1.5 trillion in external financing by 2028. Bridgewater founder Ray Dalio calls AI a “classic bubble,” arguing there is a difference between “how terrific a technology is and how terrific an investment is.”

The buildout isn’t stopping — Samsung just projected a 783% jump in quarterly operating profit on AI memory demand, a sign the hardware supercycle still has legs. But the era of funding it from cash flow is over. From here, the AI boom runs on borrowed money. And borrowed money, unlike venture capital, keeps a calendar.

Meanwhile in AI Tech: Anthropic is one of the labs this financing wave is meant to serve — and it’s already at the center of the industry’s risk debate. Read our earlier piece on Jamie Dimon’s warnings about AI cyber risk.

Meanwhile in Space: SpaceX isn’t only borrowing this week — it’s also flying astronauts home. Crew-12 undocks from the space station today ahead of a California splashdown.

Written by
Ryan covers artificial intelligence and enterprise tech — from foundation models and AI chips to the business of machine intelligence. He tracks model releases, funding rounds, and the policy moves shaping the AI industry.