Nvidia’s role as AI financier faces first big test as neoclouds take market hit

Nvidia’s emergence as financier for the global AI buildout is facing its first major stress test as credit markets tighten for key partners.

After creating the chips that powered the early stages of the generative artificial intelligence era, Nvidia turned to using its financial prowess to fund AI model developers, data center companies and a host of so-called neoclouds. CEO Jensen Huang’s strategy has been to not only build the most advanced technology but to provide customers with investment as well as help them take out loans so his systems can reach many more hands.

That’s meant taking full advantage of Nvidia’s investment grade credit rating and providing backstop guarantees on mega-projects for OpenAI so that buildouts can get funded at lower interest rates.

As of its last quarterly financial filing in July, Nvidia had $36 billion in off-balance-sheet AI cloud agreements, or backstops, for neoclouds, in addition to its guarantee on up to $105 billion for chips, land and power at a proposed 10 gigawatt OpenAI data center in Ohio. The company signed as a tenant for $20 billion in future data center leases, which will later be assigned to third parties.

For the vast majority of these deals, Nvidia won’t have to pay up unless participants in the project default.

“If the demand for compute doesn’t meet these really super lofty expectations, we’re going to have a problem,” said David Trainer, founder and CEO of financial research firm New Constructs, in an interview.

Nvidia declined to comment for this story.

Market conditions may soon test Nvidia’s mettle. Shares of neoclouds CoreWeave, Nebius and IREN, all cash-burning companies in Nvidia’s portfolio, are down at least 30% this month. CoreWeave, the biggest of the group, has seen the spread on its credit default swaps soar past 11%, as investors are forced to pay up for insurance against default.

Nvidia has backed other neoclouds that have run into challenges.

The company is an investor in Australian AI cloud provider Firmus, which abruptly withdrew its IPO plans this month, as well as Nscale out of the U.K. and SoftBank-backed SB Energy. Both of those companies filed to go public in the U.S. in September but haven’t provided updates on their plans in weeks.

Nvidia wrapped up a third straight down week on Wall Street on Friday, with its stock now off 7.6% in October, on pace for its worst month since November of last year.

Huang reiterated his confidence in the market this week, telling CNBC on Thursday that demand for neoclouds is “off the charts.”

“What is missing in the middle is the working capital to fuel that growth,” Huang told CNBC’s Arjun Kharpal on the sidelines of the company’s GTC conference in Berlin. “Because there’s just so much demand across so many sectors of AI, there’s a scarcity of capital.”

Nvidia has taken equity stakes in dozens of companies, including big investments in OpenAI and Anthropic. Its portfolio is worth $99 billion as of the second quarter.

Investors told CNBC that a big question surrounding Nvidia is how the company can use its cash and $5 trillion market cap to get more chips into customers’ hands, and how much risk it will assume in the process.

Nvidia graphics processing units are critical for training and deploying AI models. Revenue in the latest fiscal year hit $216 billion, up roughly eightfold since OpenAI launched ChatGPT in late 2022.

Before the AI boom, Nvidia sold graphics cards that could fit into a standard server rack or PC. Its newest systems are much larger and more expensive. Nvidia’s flagship Vera Rubin product is a seven-foot rack containing 72 GPUs and weighing as much as a pickup truck. It costs more than $7 million, according to a Morgan Stanley estimate.

While Nvidia still dominates the AI chip market, competition is emerging from the likes of Google, which rents out tensor processing units (TPUs) inside its massive cloud offering, and Advanced Micro Devices, which launched its first AI rack called Helios in July.

Cash is Nvidia’s other great advantage. The company generated $70 billion in free cash flow in the first half of this year, while AMD totaled just over $4 billion and Google parent Alphabet went cash flow negative in the second quarter to fund its buildout.

That’s a big edge for Nvidia in building out its ecosystem. The company is now offering a full-service platform complete with software, services, integrated hardware and, for many customers, financing assistance.

Nvidia’s most recent quarterly financial filing in August outlined over $500 billion in off-balance-sheet commitments over the coming years, with a majority maturing through fiscal 2029. That same month, Huang announced a partnership with six large asset managers on a $500 billion financing push designed to treat compute infrastructure as an asset class.

That agreement, which promised financial firms up to $125 billion in residual value guarantees on GPU financings, was in addition to the stakes Nvidia has taken as a limited partner in private equity funds. The company seeded $2 billion in Brookfield’s inaugural AI Infrastructure Partners Fund, which aims to raise $10 billion in total capital, as well as other unconfirmed sums in KKR’s Helix Digital Infrastructure (an $11 billion vehicle) and Valor Equity Partners’ $5.4 billion dedicated compute fund.

“We used to build chips that we sell, and these are technology components that people buy and use,” Huang told CNBC, in announcing the agreement in August. “But now, Nvidia’s AI factory platform is really an investable asset.”

Huang has adopted a marketing mantra to reinforce the unique value of its equipment.

“We’re the most rentable. We’re fungible, and we’re durable,” he said on a podcast with investor Brad Gerstner published last week.

While Nvidia waits for Wall Street firms to pull in investors to expand the GPU loan pool, the chipmaker is already actively fostering record-sized deals in the AI infrastructure space.

Nvidia’s biggest single guarantee is for OpenAI’s forthcoming facility in Pike County, Ohio, which could be the largest data center in the world. That deal requires use of Nvidia hardware for at least 20 years.

The shift from chips to capital carries significant risks should OpenAI run into trouble.

If OpenAI misses payments or faces insolvency, Nvidia can take over the lease, find a new tenant, or ask the landlord, SB Energy, to sell the facility, guaranteeing to pay the difference for the latter two arrangements up to $105 billion if the release or sale falls below a minimum price. Nvidia could also pay SB Energy a lump sum. On the other side, if OpenAI posts its own collateral, or achieves a high enough credit rating, Nvidia’s guarantee ends.

At CoreWeave, Nvidia’s investment includes equity and a commitment to buy up to $6.3 billion of its compute if it can’t find another taker.

Nvidia’s backstops can take many forms, such as guaranteeing a minimum resale price or a minimum hourly offtake price for the chips in the event the lender has to recoup the assets. With Sharon AI and other neoclouds, Nvidia’s deals have come with revenue-sharing agreements.

“We believe these AI cloud partners have strong customer demand and robust sales pipelines but are constrained by the large-scale infrastructure that is required to meet that demand,” Nvidia said in its quarterly filing.

Huang said this week that CoreWeave doesn’t need Nvidia’s backstop and that the neocloud’s cost of capital is decreasing. CoreWeave has roughly $51.6 billion in total debt, including about $35 billion in notes, bonds and loans. Yields on its bonds maturing in 2031 and 2032 topped 15%.

Brian Stutland, portfolio manager at Equity Armor Investments, said the market jitters reflect concerns about the stability of the business.

CoreWeave and other neoclouds “will sign contracts into the future, and that all looks fine and dandy in their earnings reports, but is that contract going to be stable?” Stutland said. “Is it going to break at some point?”

Nvidia says its financial support can help customers build the infrastructure needed to meet strong demand for its chips. More computing capacity could also help AI labs develop new models and services, creating further demand for Nvidia products.

“They’ll do $700 billion in revenue next year,” Deepwater Asset Management founder Gene Munster told CNBC last month. “That affords them the ability to finance some of their customers.”

Nvidia’s credit backstops have mostly kicked in over the past six to 12 months. Its last annual filing, published in February, makes no mention of AI cloud agreements.

Warren Hosseinion, co-founder of AI infrastructure startup Compute Labs, said Nvidia doesn’t want to be in the position of insurer, “but they’re willing to play that role while the market and the industry catches up.”

“They kind of control how things move,” said Hosseinion, whose company tokenizes GPUs in order to open up pools of capital for neocloud operators. “They have to be very careful about how they release their technology at what pace, because a lot of that will dictate how the market moves around it.”

Nvidia is actively looking for insurers to bear some of the risk associated with the potential loss in value of GPUs.

“Insurers are not yet a complete panacea to this problem,” said Tom Hoad, executive director of product innovation at global insurance group Howden. “But they are probably going to turn up and be helpful in the right situations, and therefore open up some commercial transactions.”

According to research from New Constructs, Nvidia’s net present value of off-balance-sheet liabilities in fiscal 2026 totals over $153 billion, an increase of over $100 billion from last year.

Such liabilities have been growing across the entire AI trade, mainly in the form of lease obligations. Ratings agencies like Moody’s have been sounding the alarm. In 2026, guarantees are the fastest-growing type of off-balance-sheet financing arrangements, the agency said.

Moody’s doesn’t count guarantees as debt when rating corporate buyers, but it’s “something we’re constantly monitoring,” said David Gonzales, a vice president at the agency.

“These things are advanced structures that are unique, and so there is no blanket real approach you can take,” he said. Gonzales said it’s likely that chips will retain “values in excess of contractual residual value amounts” in the near future, which would mitigate risk in many of Nvidia’s financing structures.

Still, how much exposure investment grade AI players can handle is “the question we’re evaluating at all times every day,” Gonzales said.

“It’s the center of all of our discussions right now,” he said.

— CNBC’s Seema Mody contributed to this report.  

WATCH: CNBC’s full interview with Nvidia CEO Jensen Huang