Nvidia is putting as much as $3.2 billion behind Corning, a company better known for glass, fiber and the tough cover material used on iPhones. The pairing looks unusual only until the data-center bottleneck comes into view: faster GPUs need faster, cooler ways to move data inside racks and between racks.
Moving photons is five to 20 times more energy-efficient than moving electrons.
That was how Corning CEO Wendell Weeks framed the deal. Nvidia can invest up to $3.2 billion through Corning warrants and pre-funded warrants, while also making separate multibillion-dollar prepayments to support new manufacturing capacity.
Three plants for optical products
Corning plans to build three optical-products plants in North Carolina and Texas. Once running, the sites are expected to expand Corning's U.S. optical manufacturing capacity tenfold and create at least 3,000 jobs. Corning shares rose 12% on the day of the news, after a gain of more than 300% over the previous year, while Nvidia climbed 6%.
The market reaction reflects a structural shift. Blackwell and Rubin systems push computation higher, but Rubin Ultra racks already require internal bandwidth measured in TB/s. At tens of thousands of GPUs, copper cables struggle with signal loss after a few meters, energy use and the sheer space taken by thick, heavy cabling.
Optical interconnects address those problems, which is why the move from copper switches to optical links has become industry consensus for AI data centers. The shortage is capacity. Nvidia makes chips, not fiber, and Corning's traditional telecom and consumer-electronics customers move on slower demand cycles than AI infrastructure now requires.
A supply-chain message
Jensen Huang described AI as the largest infrastructure expansion of the era and a once-in-a-century chance to reindustrialize U.S. manufacturing and supply chains. For Wall Street, the message is that Nvidia is becoming an integrator of the AI infrastructure stack, from GPUs to optics and power. For Washington, it is a story about U.S. factories, jobs and domestic capacity in a tariff-heavy political environment.
The Corning move fits a broader pattern: fiber capacity, ServiceNow's Action Fabric for enterprise agents, GTC 2026 ecosystem deals, CUDA's developer moat and the IREN AI cloud contract. Nvidia is no longer just the strongest GPU seller; it wants leverage over every critical node of AI infrastructure.
Corning wins, Nvidia keeps optionality
For Corning, the deal turns a glassmaker founded in 1851 into an AI infrastructure stock. But Nvidia's $3.2 billion is structured as warrants, not direct cash equity, giving it flexibility to decide how much to exercise as Corning performs and the AI data-center buildout unfolds.
If construction slows, or if a disruptive optical technology such as silicon photonics changes the market, Nvidia has not committed itself to spending the full amount. Corning's new plants are expected to start construction in 2027, with jobs arriving over the following years. The copper-to-fiber rebuild inside AI data centers has only begun.
Sources: Nvidia to invest $3.2B in Corning to build next-generation AI optical infrastructure in the U.S. (Tech Startups); NVIDIA and Corning Announce Long-Term Partnership To Strengthen U.S. Manufacturing for AI Infrastructure (Corning official); Nvidia funds construction of Corning plants, CocoLoop, in addition to equity investment (Yahoo Finance)