Nvidia Tops $40 Billion In AI Equity Bets As Earnings Loom

Nvidia is no longer just selling the picks and shovels of the AI gold rush. It is funding the miners, the rail lines, and the towns that grow up around them. As of this week, the chipmaker has committed more than $40 billion to equity bets in 2026 alone, a pace that dwarfs anything in its history and turns the world’s most valuable company into something stranger than a semiconductor business. It looks more like a central bank for artificial intelligence.

The two latest deals landed on consecutive days. On May 6, Nvidia secured warrants to buy up to $3.2 billion of Corning stock tied to three new optical-fiber factories in North Carolina and Texas. On May 7, it took a five-year option to buy up to $2.1 billion of IREN shares at $70 each, with IREN agreeing to deploy up to 5 gigawatts of Nvidia’s DSX rack designs. Both stocks ripped on the news. Corning closed up roughly 12 percent. IREN had already climbed 813 percent over the past year before the latest pop.

The $40 Billion Number Hides A Bigger One

Strip the headline figure down and the picture sharpens. Nvidia has signed at least seven multibillion-dollar deals with publicly traded companies in 2026 and roughly two dozen private rounds, according to FactSet data cited by CNBC. The single biggest check, $30 billion into OpenAI, closed in February as part of a $110 billion OpenAI funding round at a $730 billion pre-money valuation.

Then there is the Intel trade, which has quietly become one of the most profitable equity bets a US tech company has ever made. Nvidia bought 214.8 million Intel shares at $23.28 in late December 2025, deploying $5 billion. Intel closed near $100 in early May 2026 after more than doubling year to date. That puts the position somewhere north of $21 billion in paper value, a gain of roughly $16 billion in five months on a single bet.

The accounting is what keeps Wall Street awake. Nvidia’s non-marketable equity securities ballooned to $22.25 billion at the end of January 2025, up from $3.39 billion a year earlier. Gains on private and public equity holdings hit $8.92 billion last fiscal year, against $1.03 billion the prior year. Most of that swing came from Intel.

None of this shows up cleanly on a P/E ratio. It shows up in Other income, where it can swing several billion dollars a quarter and still get described as a footnote.

What Jensen Huang Is Actually Building

Read the deal terms together and a pattern emerges. Corning makes the fiber. Marvell, Lumentum, and Coherent build the silicon photonics, with Nvidia having dropped $2 billion into each in March. IREN, CoreWeave, and Nebius operate the data centers. OpenAI, Anthropic, and xAI write the software that needs the chips. Every node in the supply chain is now partly owned by the company that sells the GPUs.

Our investments are focused very squarely, strategically on expanding and deepening our ecosystem reach.

That is how Huang framed it on Nvidia’s last earnings call in February. In April, on a podcast, he was blunter. “There are so many great, amazing foundation model companies, and we try to invest in all of them. We don’t pick winners. We need to support everyone.”

The reason Nvidia needs Corning specifically is engineering, not accounting. The company’s next-generation Rubin systems are running into a hard physical limit: every time copper bandwidth doubles, usable cable length halves. Inside a single rack, copper still works. Between racks, fiber wins. Nvidia’s co-packaged optics program integrates the optical engine directly onto the switch, cutting power per port by a factor of five and pushing fiber closer to the GPU itself.

That is what the Corning factories will feed. The deal locks in supply for a transition that has to happen if Rubin and Rubin Ultra ship on schedule.

Why “Circular Financing” Will Not Go Away

The criticism is straightforward. Nvidia generated $97 billion in free cash flow last fiscal year. It is now using that cash to buy stakes in companies that turn around and buy Nvidia chips. In some cases, those companies then lease compute back to Nvidia. The OpenAI deal alone could account for as much as 13 percent of Nvidia’s projected fiscal 2026 revenue, based on consensus estimates near $272 billion.

Matthew Bryson, an analyst at Wedbush Securities, wrote that the deals fit “squarely into the circular investment theme” but added that they create “a competitive moat” if execution holds. Mizuho’s Jordan Klein split the difference. The component-maker deals are “super smart by the CFO and team and a great use of cash,” Klein wrote in an email. The neocloud bets are different.

It smells like you are pre-funding the purchase of your own GPUs and products.

Klein attributed that line to the IREN, CoreWeave, and Nebius investments specifically. Nvidia put $2 billion into CoreWeave in January and another $2 billion into Nebius around the same window. Both companies’ valuations depend heavily on access to Nvidia hardware that other buyers cannot get.

Michael Burry, the investor who shorted the 2008 housing bubble, has built his loudest position yet around this thesis. In April, on his Cassandra Unchained Substack, Burry disclosed he had added long-dated puts at a $115 strike with Nvidia trading near $188. He compared Nvidia to Cisco circa 2000, which fell roughly 78 percent in the bust and took 25 years to reclaim its peak. Nvidia responded with a seven-page memo to analysts disputing his stock-buyback math, according to Barron’s. Burry’s reply was three sentences long. He was not changing his trade.

Ben Bajarin at Creative Strategies framed the risk plainly to CNBC: “The risk is that if the cycle turns, the market starts questioning how much of the demand was organic versus supported by Nvidia’s own balance sheet.”

The Intel Stake Changes The Math

One investment makes the rest of the portfolio look conservative. Nvidia’s Intel stock purchase closed on December 26, 2025 at $23.28 per share, an FTC-approved private placement of 214.8 million shares. Intel was trading near $36 within days of close. By early May 2026, the stock had pushed close to $100.

That single position has produced more paper profit than Nvidia’s entire fiscal 2025 net investment gain. It also reframes the broader strategy. If even one or two of the seven 2026 public deals deliver Intel-style returns, the headline circularity argument loses some teeth, because the portfolio starts paying for itself out of mark-to-market gains rather than chip orders.

That is the bull case, in one paragraph. The bear case is that Intel was a bet on a struggling fab giant getting a strategic lifeline, not on a circular AI loop. The two stories are not the same trade.

Earnings Will Force The Issue

Nvidia reports first-quarter fiscal 2027 results on May 20, 2026. Management has guided to $78 billion in revenue, an accelerated 77 percent year-over-year growth rate. Wall Street consensus already prices in roughly 79 percent. A meaningful pop probably requires the company to clear 80.

Analysts at Goldman Sachs, Morgan Stanley, and Bernstein have raised price targets into the $200 to $240 range. The forward P/E sits at 23.8, the cheapest among major AI peers. Broadcom trades at 31.3. AMD trades at 53.6. The valuation discount exists for two reasons: continued China export uncertainty and rising scrutiny of exactly the dealmaking pattern this article describes.

Investors will also get a clearer line on the size of Nvidia’s portfolio. The 10-Q filing dropping with earnings will refresh the carrying value of non-marketable equity securities, the unrealized gains on public holdings, and any new concentrations.

A few specific items to watch:

• Investment income line: Whether Other income, net continues to scale at multiples of last year’s $8.9 billion gain.
• Gross margin trajectory: Management has signaled a glide path from 78 percent peak toward a 71 to 72 percent long-term target as Blackwell Ultra ramps. Anything below 70 percent triggers selling.
• Rubin commentary: Color on Vera Rubin shipment timing, including the CPO-equipped switch generation, would clarify how fast the Corning deal monetizes.
• China exposure: The $78 billion guide explicitly excludes China data center compute revenue. Any change to that assumption resets every model on the Street.

The IREN And Corning Deals Up Close

The two announcements that pushed Nvidia past $40 billion this year illustrate the strategy’s split personality.

IREN, the Australian operator formerly known as Iris Energy, started life as a Bitcoin miner. Its 2 gigawatt Sweetwater campus in West Texas was always engineered for high-density compute, with rack densities approaching 200 kilowatts and liquid cooling baked into the design. In November 2025, IREN signed a $9.7 billion GPU cloud deal with Microsoft. Six months later, Nvidia layered a $3.4 billion managed-cloud agreement on top, plus the $2.1 billion warrant. The company reported AI Cloud Services revenue of $33.6 million in fiscal Q3 2026, a small number that is now expected to scale rapidly.

Corning is the opposite story. The company is 175 years old. Its glass shows up in Gorilla Glass smartphone covers, fiber-optic cables, and Pyrex. The Nvidia deal involves three new US factories, at least 3,000 new jobs, a tenfold expansion of US optical-connectivity capacity, and a 50 percent boost to US fiber production. Nvidia gets warrants on up to 15 million shares at $180, plus a $500 million pre-funded warrant on 3 million more.

This is such an extraordinary opportunity because we can use these market dynamics to reinvest, revitalize American manufacturing for the first time in several generations.

Huang said that on May 7 alongside Corning CEO Wendell Weeks. Strip out the politics and the deal does something concrete: it locks domestic supply for the optical components Rubin needs, at a moment when Nvidia is racing to keep its scale-out network ahead of AMD’s MI400 and Broadcom’s custom ASIC roadmap.

What Could Actually Break

The fragile point in the system is not Nvidia. It is the layer below. CoreWeave has roughly $18.8 billion in GPU-collateralized debt and recently saw shares drop as much as 12 percent intraday on a Business Insider report that financing partner Blue Owl Capital had failed to secure $4 billion for a Pennsylvania data center. Nebius traded down in sympathy. Applied Digital, where Nvidia recently trimmed its stake, dropped further.

The neocloud sector trades on a single assumption: that AI compute demand will not just keep growing but keep outrunning what hyperscalers can build internally. If Meta, Google, or Amazon’s custom silicon programs hit their stride, that assumption weakens. Meta’s $48 billion combined commitment to CoreWeave and Nebius, announced in April, suggests the hyperscalers themselves do not yet feel ready to bring everything in-house. But the clock is moving.

For Nvidia, the bigger question is whether the equity portfolio and the chip business start moving in the same direction at the same time. In a true downturn, they would. The same demand collapse that tanks GPU orders would also tank the AI-exposed equities Nvidia holds. The hedge is not a hedge if both sides are the same trade.

Frequently Asked Questions

When does Nvidia report earnings, and what number actually matters?

Nvidia reports Q1 fiscal 2027 results on May 20, 2026, with a conference call at 2 p.m. PT on investor.nvidia.com. The number that moves the stock is not the headline revenue beat but year-over-year growth. Management guided 77 percent. Consensus is closer to 79. To trigger a real rally, the print likely needs to clear 80, plus gross margin holding above 70 percent.

What is “circular financing” in plain English?

It is when a supplier invests in a customer, and the customer then uses that money to buy from the supplier. Critics say Nvidia is doing this with neocloud operators like CoreWeave and IREN. Defenders say Nvidia is buying scarce things it actually needs, including power, data center sites, and fiber capacity. The honest answer is both are partly true. The 13 percent OpenAI revenue concentration is the line analysts watch.

How much has the Intel stake actually made?

Nvidia bought 214.8 million Intel shares at $23.28 in late December 2025, a $5 billion check. Intel traded near $100 in early May 2026. That puts the position above $21 billion, a paper gain of roughly $16 billion in about five months. The position vests on Nvidia’s balance sheet and shows up in unrealized gains, not GAAP revenue. Realized gains would only appear if Nvidia sells.

Will the OpenAI deal still go to $100 billion?

No, at least not on the original terms. The September 2025 letter of intent for $100 billion was tied to OpenAI deploying 10 gigawatts of Nvidia systems. OpenAI moved away from running its own data centers and the deal stalled. Huang said in March 2026 that $100 billion is “not in the cards” and the $30 billion February 2026 round “might be the last” check Nvidia writes before an OpenAI IPO.

Should the average reader care about any of this?

Yes, if you own broad US index funds. Nvidia is roughly 7 percent of the S&P 500. Its $5.2 trillion market cap means a 10 percent move in either direction shifts overall index performance noticeably. The circular-financing debate is not academic. It is a real disagreement about whether AI demand is organic enough to support current valuations across the entire AI supply chain.

The answer probably arrives in pieces, not all at once. May 20 will resolve part of it. Whether IREN, CoreWeave, and Nebius can post organic revenue growth that does not depend on Nvidia capital will resolve more. Until then, Nvidia keeps writing checks, and the market keeps trying to decide whether that is a moat or a mirror.

For broader context on how Intel’s revival ties into this, see our coverage of Apple’s preliminary deal for Intel to fabricate iPhone and Mac chips, and on Nvidia’s hardware side our look at how Nouveau is closing the gap on Nvidia’s R595 workstation drivers.

Disclaimer: This article reports on company strategy, analyst commentary, and market movements and does not constitute investment advice. Equity investments in semiconductor and AI infrastructure companies carry significant risk, including the potential for substantial loss. Readers should consult a licensed financial advisor before making investment decisions. All price targets, valuations, and figures cited are accurate as of publication on May 9, 2026 and are subject to change without notice.