Nvidia and Corning announced a multiyear commercial and technology partnership that gives the chipmaker the right to invest up to $3.2 billion in the glass-and-fiber maker, funding a build-out of US optical connectivity manufacturing aimed squarely at AI data centers. The deal was unveiled May 6 and rippled through markets, sending Corning shares roughly 12% higher to a record close while Nvidia gained almost 6%.
Inside the structure
Nvidia's commitment is split into two pieces: a $500 million pre-funded share purchase and warrants potentially worth another $2.7 billion if fully exercised. The investment underwrites Corning's plan to build three new advanced manufacturing facilities in North Carolina and Texas, expand US optical connectivity capacity tenfold, and grow domestic fiber output by more than 50%. Corning expects the project to create more than 3,000 jobs.
The partnership extends beyond capital. The two companies will collaborate on co-developing next-generation optical components for AI infrastructure, including the components that knit together the racks of GPUs powering frontier model training and inference.
Why optics now
The deal is the clearest signal yet that Nvidia is preparing to swap copper interconnects for optical glass fiber inside its rack-scale AI systems. The shift, known as co-packaged optics, moves photonic transceivers directly onto the same package as switching silicon, cutting power draw and unlocking higher bandwidth between GPUs. Jensen Huang has described the transition as essential to scaling AI training and inference clusters past current density limits.
That math has hardened in 2026. As hyperscalers race to stand up multi-gigawatt training campuses, copper's reach inside the rack has become a bottleneck, and the industry has been telegraphing a move to optics for the next generation of AI servers. Locking in a domestic supply line for the specialized fiber and connectors that build-out requires has become a strategic priority.
A bet on US manufacturing
The announcement also lands as Washington pressures hyperscalers and chipmakers to onshore more of the AI supply chain. Recent months have seen multibillion-dollar US factory commitments from Nvidia, TSMC, Samsung, and others, and the Corning deal slots neatly into that pattern by anchoring a critical piece of the AI hardware stack — optical interconnect — inside the United States.
For Corning, the agreement caps a remarkable run. The stock has roughly doubled year-to-date in 2026 as investors rerated the company on its exposure to AI infrastructure, and the Nvidia tie-up converts that thesis into a concrete revenue and capacity plan.
Implications
If Corning hits its capacity targets, the partnership eases one of the less-discussed constraints on AI scaling: the supply of high-grade optical fiber and connectors that link tens of thousands of GPUs into a single training fabric. It also tightens Nvidia's grip on the AI hardware stack — moving the company further beyond GPUs into the networking layer that determines how efficiently those GPUs can be used.
The risks are familiar. Warrant-based commitments depend on continued demand, and the new factories will take years to come online. But the direction is unmistakable: the next phase of the AI build-out will run on light, and Nvidia just locked in a US-based pipeline to make that happen.
