Nvidia has now committed more than $40 billion in equity deals in 2026 alone — and $30 billion of that went to a single company that has never turned an annual profit. The chipmaker isn’t just selling GPUs anymore. It’s financing the entire AI supply chain from fiber optics to cloud compute, and making sure every dollar it invests flows right back into Nvidia hardware purchases. Wall Street calls it a growth strategy. Critics call it something else entirely: a $40 billion circular subsidy dressed up as venture capital.
The Numbers Tell a Very Specific Story
CNBC and TechCrunch confirmed this week that Nvidia has topped $40 billion in equity commitments for 2026, a pace that would’ve been unthinkable even two years ago when Jensen Huang was still primarily a chip executive, not a dealmaker. The anchor investment: $30 billion into OpenAI, the company that consumes more Nvidia H200 and Blackwell chips than perhaps any other organization on Earth.
But the OpenAI check, massive as it is, obscures the more interesting pattern. Nvidia also announced seven multi-billion-dollar deals in publicly traded companies this year:
Up to $3.2 billion in Corning, the glass manufacturer that supplies fiber optics for AI data center connectivity. $2.1 billion in IREN, a data center operator. $2 billion in CoreWeave, the GPU cloud startup that just IPO’d on the back of Nvidia’s chips. And $2 billion in Nebius Group, an AI cloud company that designs and deploys infrastructure built — you guessed it — entirely on Nvidia hardware.
This isn’t diversification. This is vertical integration funded by equity checks instead of acquisitions.
The Circularity Problem Nobody Wants to Say Out Loud
Here’s the part that should make you pause. Nvidia is investing billions of dollars into companies that are, by definition, Nvidia’s own customers. CoreWeave buys Nvidia GPUs. IREN runs Nvidia hardware in its data centers. Nebius deploys Nvidia chips. OpenAI has multi-year purchase agreements for Nvidia silicon worth tens of billions.
The capital loop works like this: Nvidia writes a $2 billion equity check to Company X. Company X uses that capital (and more) to buy $5 billion worth of Nvidia chips. Nvidia books the GPU revenue, its stock goes up, and the equity stake Nvidia holds in Company X also appreciates because Company X now has more compute capacity and higher revenue. Everyone’s balance sheet looks better. Everyone’s stock goes up. The money goes in a circle.
Critics have been quick to point out that this resembles the kind of vendor financing that blew up spectacularly during the telecom bubble of the early 2000s, when Cisco, Lucent, and Nortel lent money to their own customers to buy networking equipment. When those customers went bust, the loans defaulted, and the equipment makers cratered.
Nvidia’s counterargument is simple: AI demand is real. These aren’t speculative telecom buildouts. Every hyperscaler on Earth is racing to deploy more compute, inference demand is growing faster than training demand, and Nvidia’s chips remain the only game in town for serious AI workloads. The equity investments aren’t loans — they’re ownership stakes in companies riding the same wave Nvidia is riding.
Why Jensen Huang Is Playing Venture Capitalist
The strategic logic is actually cleaner than the critics acknowledge, even if the optics are messy. Nvidia’s chip business is a near-monopoly, but monopolies have a shelf life. AMD is gaining ground. Custom silicon from Google (TPUs), Amazon (Trainium), and Microsoft (Maia) is getting better every quarter. Startups like Cerebras and Groq are carving out niches in inference.
Nvidia’s $40 billion in equity bets is an insurance policy. By owning stakes across the entire AI infrastructure stack — from the glass in the fiber cables (Corning) to the data centers that house the chips (IREN) to the cloud platforms that rent the compute (CoreWeave, Nebius) to the AI labs that consume the most compute (OpenAI) — Nvidia is making itself structurally indispensable, not just technologically dominant.
If a customer considers switching from Nvidia GPUs to AMD MI400s or Google TPUs, that customer now has an Nvidia board observer, an Nvidia equity stake, and deep financial ties that make switching costly in ways that go far beyond technical migration. It’s the same playbook Intel ran in the PC era — except Nvidia is executing it at 10x the scale and 100x the speed.
The $30 Billion OpenAI Bet Is the Riskiest Part
Three-quarters of Nvidia’s 2026 equity capital went to a single company that is losing billions of dollars per year and has never posted an annual profit. OpenAI’s $25 billion revenue run rate is impressive, but its compute costs are astronomical, its workforce is expensive, and its path to sustained profitability depends on enterprise adoption rates that are still unproven at scale.
If OpenAI stumbles — if the IPO disappoints, if enterprise customers balk at pricing, if a competitor undercuts them with open-source models — Nvidia’s $30 billion stake could lose a third of its value overnight. And because Nvidia’s stock price is partially propped up by the perceived value of its investment portfolio, a write-down on OpenAI would ripple through Nvidia’s own market cap.
That’s the real risk of the circular model: when one domino falls, the whole loop reverberates.
What This Means for the Rest of the Industry
For AMD, the message is stark. Lisa Su can build faster chips, but she can’t match Nvidia’s financial lock-in. When your competitor is simultaneously your customer’s chip supplier, equity investor, and infrastructure partner, you’re not competing on specs anymore — you’re competing against an ecosystem. AMD’s MI400 chips could benchmark 20% faster than Blackwell Ultra and it still might not matter if the customer has $2 billion in Nvidia money on its balance sheet.
For startups, the math is even harder. Nvidia’s equity investments come with preferred access to chips, co-design partnerships, and the implicit understanding that your cloud platform will be built on Nvidia hardware. That’s not corruption — it’s capitalism operating exactly as designed. But it does mean the AI infrastructure market is consolidating around one company’s financial web in a way that makes it increasingly difficult for independents to compete.
For regulators, this should be a case study. Nvidia doesn’t technically own any of these companies. It doesn’t control their boards. It isn’t acquiring them. But through $40 billion in equity stakes, it has built a financial gravity well that shapes the decisions of every major AI infrastructure player on the planet. Whether that qualifies as anticompetitive behavior depends on how narrowly you define the word.
The Verdict
Nvidia isn’t just selling picks and shovels anymore. It’s financing the entire gold rush and taking an ownership stake in every mine. The $40 billion in equity commitments — up from essentially zero three years ago — represents the fastest transformation of a chip company into a financial conglomerate in tech history. The circular dynamics are real and the concentration risk is genuine. But as long as AI demand keeps growing, the loop keeps spinning, and everyone inside it keeps getting richer. The question isn’t whether Nvidia’s strategy is smart. It obviously is. The question is what happens when the music stops — and whether $40 billion in cross-investments amplifies the crash the same way it amplified the boom.
