NVIDIA deploys $40B equity bets

Nvidia is doing something bigger than venture investing. It is using its stock and balance sheet to shape the physical AI buildout around its chips. That means fiber plants, cloud capacity, networking, and data-center power — not just software bets. The news this week is that those commitments have now pushed past $40 billion in 2026, helped by new deals with Corning and IREN. ### Why is Nvidia buying pieces of other companies? Because the bottleneck in AI is no longer just the GPU. A modern AI cluster needs optics, networking, power, land, cooling, and operators that can actually deploy all of it at scale. Nvidia sells the core compute, but if any one of those other pieces runs short, the whole system slows down. So Nvidia is increasingly funding the missing pieces itself. That is the logic running through these deals. (cnbc.com) ### What happened this week? On May 6, Nvidia and Corning announced a multiyear partnership to expand U.S. optical manufacturing for AI infrastructure. Corning said it will raise U.S. optical connectivity capacity 10x, lift fiber production by more than 50%, build three new plants in North Carolina and Texas, and create more than 3,000 jobs. Nvidia also got the right to invest up to $3.2 billion in Corning through warrants. (corning.com) ### Why does Corning matter so much? Because AI factories are becoming communication problems as much as compute problems. Thousands of GPUs have to move data between one another insanely fast, and copper starts to look clumsy at that scale. Nvidia has been pushing co-packaged optics as the next step — basically moving from metal wiring to light where the traffic gets too heavy. Corning makes the glass and optical components that let that happen. (corning.com) ### What is the IREN deal really about? IREN gives Nvidia something different — ready-made infrastructure. On May 7, the companies said they would work together to deploy up to 5 gigawatts of AI infrastructure. IREN issued Nvidia a five-year right to buy up to 30 million shares at $70 each, which could amount to a $2.1 billion investment. Separately, IREN said it signed a five-year $3.4 billion managed GPU cloud deal for Nvidia’s internal AI and research workloads in Childress, Texas. (cnbc.com) ### Where does Marvell fit? Marvell sits on the networking and custom-chip side. On March 31, Nvidia and Marvell announced a strategic partnership around NVLink Fusion and silicon photonics, and Nvidia invested $2 billion in Marvell. That deal is about making Nvidia’s ecosystem more attractive even when customers want semi-custom systems instead of buying every component directly from Nvidia. In other words, Nvidia is widening the tent without giving up control of the architecture. (investor.nvidia.com) ### Is the $40 billion number all cash today? No — and that is the catch. A lot of these commitments are rights, warrants, or staged investments tied to deployment milestones. The OpenAI partnership announced in September 2025 is the clearest example: Nvidia said it intends to invest up to $100 billion progressively as each gigawatt of Nvidia systems gets deployed. So the headline number is best read as strategic firepower, not money already gone. (investor.marvell.com) ### So what is Nvidia really building? Basically, a privately financed industrial stack for AI. Instead of waiting for suppliers and customers to expand on their own, Nvidia is helping fund the factories, fiber, and cloud operators that will buy and run Nvidia gear. That makes the company look less like a chip vendor and more like the planner of an entire computing supply chain. (investor.nvidia.com) ### Bottom line? The simple version is that Nvidia has decided the best way to secure future demand is to bankroll the ecosystem around itself. If AI keeps scaling the way Jensen Huang expects, these equity bets will look less like side deals and more like the blueprint. (cnbc.com)