Nvidia Insists It Can Keep Printing Money to Fund the AI Boom

Nvidia NVDA 8.74%increase; up pointing triangle was closing in on a landmark agreement to help OpenAI secure one of the world’s largest planned data-center projects when Colette Kress, the chips firm’s chief financial officer, pumped the brakes.

Concerned about negative investor reaction to a Wall Street Journal report that Nvidia was in talks to financially guarantee a project expected to cost more than $500 billion, Kress pushed for a scaled-back deal, according to people familiar with the matter. In the end, Nvidia agreed to backstop less than half of the original amount discussed.

It was a tactical retreat—but on the broader strategy of using its balance sheet to bolster the AI boom, Nvidia shows no signs of giving ground.

On Wednesday, the world’s only $5 trillion company reported its second-quarter earnings, which included the blockbuster mic drop that the company expects its revenue to grow much faster than Wall Street analysts had been expecting. Shares climbed in after-hours trading en route to a $442 billion gain Thursday.

Nvidia executives seized on the positive vibes created by the revenue news to offer a full-throated, affirmative argument for the company’s growing use of financial guarantees, equity investments and other mechanisms to help its customers buy artificial-intelligence chips in ever greater quantities.

Kress name-checked some of Nvidia’s biggest recent actions, including the data-center project, located on federal land in Ohio; an initiative with six major Wall Street investment firms to provide $500 billion in financing for chip purchases; a deal to provide “selective credit enhancement” to enable an unnamed AI lab to buy 2 gigawatts of computing power; and credit support for neo-cloud companies that agree to share some revenue with Nvidia. (Some of the revenue-sharing agreements have been paused, the Journal reported.)

Kress also said that Nvidia has invested nearly $50 billion in frontier AI labs.

“We recognize the scale of this support, and we know some will call this circular financing. We see it differently,” she said.

How does Nvidia see it? Kress laid it out.

OpenAI, Anthropic and other so-called frontier labs, she said, are “growing faster than what their balance sheets and credit profiles can support.” They are juicing revenues at unprecedented rates, but the growth can only continue as long as they can secure ever-greater amounts of computing power, which they will have trouble doing on their own as long as they lack the ability to borrow huge sums of money at competitive rates.

Until they can do so, Kress said, “Nvidia is needed to help power this flywheel.”

In contrast to the 5% share-price slide that followed news of Nvidia’s planned Ohio backstop, investors greeted Kress’s Wednesday remarks with an 8.7% jump—good for the second-largest one-day market cap gain in history.

John Belton, a portfolio manager at Gabelli Funds, said the financing initiatives are geared toward accelerating the AI data center build-out and reducing borrowing costs for smaller cloud companies, but called them “polarizing for the stock” because of ongoing market concerns about circular financing. Still, helping customers afford data centers probably makes good sense, he said.

“It is unlikely that this debate will be resolved anytime soon,” Belton said. “But for now, they should support revenue and earnings upside both near- and medium-term.”

The ultimate expression of Nvidia’s flywheel strategy might be the data-center project in Portsmouth, Ohio. At its planned size of 10 gigawatts, it would be the largest such facility in the world—with Nvidia providing financial support to each of the major players involved, in some cases at multiple levels.

To help developer SoftBank’s SB Energy raise debt to build the campus, Nvidia agreed to backstop a portion of the completed value of the site, up to $105 billion. In return, it has the exclusive right to provide chips for the first half of the project. Nvidia is expected to provide financing for the chips OpenAI will need to buy, with purchases expected to total $350 billion for the full build-out.

The day of that announcement Chief Executive Jensen Huang put it plainly in an X post.

“Is this circular financing?” he wrote, before answering, “No.”

Huang compared the financial guarantee to the same strategy Nvidia applies to managing its supply chain. “We secure critical inputs when we have visibility into customer demand and when doing so enables long-term productive capacity,” he said.

With Anthropic planning an IPO as soon as September and OpenAI expected to follow next year, the need for credit backstops on their chip buys should lessen over time. Immediately after SpaceX went public, it was able to raise $25 billion in a bond sale.

Other bottlenecks remain. Looming large among them is electricity. There, too, Nvidia is deploying its balance sheet to make sure customers have somewhere to plug in servers.

Nvidia has picked up its pace of investment in this area, with up to billions invested across several different firms. Earlier this year, Nvidia invested $2 billion in power developer Lancium, which gives it a 20% stake in the firm. Nvidia has the option to increase its stake with another $1 billion of investment.

Last week, Nvidia also invested in power developer Cloverleaf for an undisclosed amount in the hundreds of millions. Nvidia’s investment with SB Energy, the developer of the Ohio project, starts at $1.5 billion.

Nvidia’s incentive is clear: The more power developers it works with, the more data centers that are specially built for its chips versus its competitors. In its announcement, Cloverleaf said its partnership with Nvidia meant that the firm would tailor its design decisions to Nvidia’s architecture.