Summary: Nvidia’s equity pledges to AI companies had passed US$40 billion by May 9, 2026, including a US$30 billion OpenAI commitment. The total also covers seven multibillion-dollar public-company stakes and roughly two dozen private rounds, concentrating money at suppliers and data-center operators. That scale makes Nvidia both supplier and investor, which critics describe as circular even while it strengthens demand and ecosystem access.
With nearly eight months of 2026 still to run, Nvidia’s equity pledges to AI companies had already topped $40 billion by May 9, according to a CNBC tally. OpenAI accounts for the largest piece: a $30 billion investment. Reporting on the same figures that day, TechCrunch counts seven announced stakes in publicly traded companies, each worth billions — the glassmaker Corning, at up to $3.2 billion, and the data center operator IREN, at up to $2.1 billion.
This is not just a story for the finance pages. Capital allocation and procurement in AI have merged into a single decision surface: the company that dominates compute supply now holds major stakes in the customers and vendors around it. Nvidia ships GPUs, yes — but it also spends its balance sheet to nail down future demand.
Scale Check: Many a Fund’s Entire Capital, Outlaid in Four Months
TechCrunch’s January 2 retrospective sets the baseline: across all of 2025, Nvidia took part in 67 venture deals with AI startups. FactSet’s data shows how much faster 2026 is moving — with only about a third of the year gone, Nvidia’s name already sits on something like two dozen private-startup funding rounds. Stack the public-company stakes on top and total commitments have passed $40 billion, a sum larger than what many major venture firms raise for an entire fund, and it went out the door in a bit more than four months.
Following the Cash: OpenAI First, Then Glass and Data Centers
The $40 billion breaks down along a few concentrated lines:
- The biggest single commitment: $30 billion into OpenAI
- Seven public-company positions, each in the billions of dollars; the latest are Corning and IREN, at up to $3.2 billion and $2.1 billion respectively
- What remains is spread over about two dozen rounds backing private startups
Neither pick takes much decoding. Corning makes fiber and specialty glass; IREN operates AI data centers — placing one upstream and the other downstream of Nvidia’s compute supply chain. Instead of waiting for the ecosystem to grow on its own, Nvidia is injecting capital straight into these pinch points, keeping the glass, the facilities, and the demand for models inside its own sphere of influence. And when a single balance sheet backs the supplier of materials, the operator of data centers, and the marquee customer at once, every tier of the stack gains a reason to fall in line with Nvidia’s roadmap.
Why Critics Call These Deals Circular
The pattern has earned itself an unflattering label: circular deals. A large share of the companies Nvidia backs are also Nvidia customers, so the money effectively travels a loop — customers fund their Nvidia purchases with Nvidia’s own capital while revenue figures and valuations lift one another. Wedbush analyst Matthew Bryson says the investments fall “squarely into the circular investment theme.” He does concede the upside, though: if the strategy lands, it hands Nvidia a “competitive moat.”
March carried a quieter message. TechCrunch reported on March 4 that CEO Jensen Huang described Nvidia as easing off its OpenAI and Anthropic commitments — an explanation the outlet noted “raises more questions than it answers.” Whatever cooling was under way, May’s running total says it never touched the aggregate outlay: a $40 billion commitment ledger drowns out any pause narrative, and Corning plus IREN — the two newest public-company stakes — only surfaced after that much-discussed pullback was supposedly in motion.
Consequences for the AI Ecosystem
For one, the hunger for AI exposure has yet to crest. The energy that sends public investors crowding into AI chip names — Cerebras’s IPO pricing and launch is one expression — shows up on the institutional side as Nvidia’s widening equity web; two costumes, one capital race.
For another, founders who take Nvidia’s money collect an ecosystem pass along with the check. The investments tend to come packaged with allocated compute and software support, and portfolio companies move through the CUDA ecosystem’s inner circle with less friction. That is a moat in practice — and a genuine reason the terms get signed.
Third, a warning the procurement and architecture side should take personally: the vendor you lean on most now owns pieces of your competitors’ buildouts. Whenever you score technology choices for “neutrality” — networking, cloud, optics, even model providers — put Nvidia’s equity stakes down on the risk register; the company deciding your roadmap profits from the way the market around it consolidates.
Nothing here breaks the law, and a good deal of it may be perfectly rational. Someone has to finance real chip demand, and Nvidia taking equity in that buildout is part of why it happens at all. Rational for Nvidia, however, is not the same thing as neutral for you. The countermeasure is unglamorous but it works: map out where your critical-path suppliers double as shareholders across your market, and bake that exposure into the pricing of every multi-year contract you put your name on.