Lambda, an AI cloud company that rents out Nvidia’s chips, has completed a $1 billion private short-term debt financing arranged by JPMorgan, according to people familiar with the matter. Bloomberg reported the deal. The money will go toward buying Nvidia AI accelerators that Lambda then rents to customers, including Microsoft, one of the largest buyers of cloud computing in the world.
The financing is the latest in a string of borrowings that have turned the company into one of the most active borrowers in the AI infrastructure market. In May, Lambda closed a $1 billion credit line. This week it completed a $926 million loan to buy GB300 systems, Nvidia’s newest generation of data center hardware. People close to the company say it has also been in discussions about a roughly $3 billion pre-IPO round, which would rank among the largest private financings in the industry this year.
Lambda’s model is simple: borrow money, buy chips, rent them out at a markup. The company is one of a handful of so-called neoclouds, smaller cloud providers that specialize in GPU capacity, which have grown rapidly as demand for AI computing outstrips what the big cloud providers can supply on their own. These companies have carved out a business by offering access to the newest chips faster than the giants, and by renting to customers who cannot get allocation from Amazon, Microsoft or Google.
Microsoft is a notable customer. The software giant has built enormous data center capacity of its own, but it has also signed deals with outside providers to secure additional GPU supply as demand for its AI products strains its infrastructure. For Lambda, a contract with Microsoft is both revenue and a signal to other potential customers, and the company has been open about wanting more such anchor tenants as it expands its fleet.
The debt markets have embraced the AI buildout with unusual enthusiasm. Bloomberg’s tally shows more than $400 billion in AI-related debt financing globally since the start of 2026, with banks underwriting loans for data centers, power plants, chipmakers and cloud providers. Lenders are betting that demand for AI computing will keep growing long enough for borrowers to pay back what they owe, and the willingness to finance the same asset class again and again has become a defining feature of this cycle.
The middleman model has its defenders and its critics. Supporters note that it lets companies with strong customer demand grow without diluting founders, and that the big cloud providers themselves borrow heavily to build infrastructure. Critics point to the debt: a company that borrowed billions at floating rates to buy a depreciating asset is exposed on both sides of its balance sheet. If AI demand growth slows, or if the large providers flood the market with their own capacity, the value of rented chips could fall faster than the loans come due.
Lambda’s numbers illustrate the trade-off. The company was valued at $2.5 billion in a 2024 funding round, and since then demand for GPU capacity has soared, letting it grow revenue quickly on a thin equity base. The debt raises the stakes on execution: every new financing pays for growth, but it also raises the revenue the company must generate just to stay current on its obligations.
The bigger question is whether the debt wave is a sign of a maturing industry or a credit cycle in the making. Banks are lending against assets, chips, that lose value quickly as new generations arrive. A GB300 bought today will be worth a fraction of its price in a few years, and the loan book rests on the assumption that rental income arrives before the depreciation does.
So far, the bet has paid off. GPU cloud providers say their capacity is booked out months in advance, and the largest customers show no sign of slowing their orders. The risk is that the same dynamics that produced the boom, a shortage of supply and a surge of capital, also produce the next glut, when new data centers come online faster than the workloads that justify them. For Lambda, and for every lender in the AI buildout, that is the question the next two years will answer.
The lending itself has become a business within the business. Banks that once treated AI as a niche have built dedicated teams to finance everything from chip purchases to power plants, and the deals keep getting bigger as customers demand more capacity faster. The scale of the borrowing has drawn attention from regulators, who are watching whether the lenders understand the assets behind the loans and the speed at which those assets depreciate.
For Lambda, the path forward is growth first, profitability eventually. The company has said it plans to keep expanding its fleet as long as customers are waiting for capacity, and the debt provides the fuel. The risk is that the same expansion that makes the company valuable also makes it fragile: a slowdown in demand, a delay in chip delivery or a rise in interest rates would each test a business built on borrowed money. The next few quarters will show whether the model holds, and whether the lenders who backed it were right.
