Nvidia and Corning Partner on Optical Tech to Scale AI Data Centers

AI data centers are hitting physical walls. Moving massive volumes of information between chips using copper wiring burns power, generates heat and slows performance at the scales now required for frontier models. On May 6, 2026, Nvidia and Corning announced a multiyear partnership to build three dedicated optical-technology factories in North Carolina and Texas that aim to change the equation.

The companies said in a joint release that the facilities will create at least 3,000 jobs and increase Corning's U.S. optical manufacturing capacity tenfold. Financial terms were not disclosed. The move accelerates a shift toward co-packaged optics, technology that replaces copper interconnects with glass fiber carrying data as light. Nvidia CEO Jensen Huang stated the collaboration advances American manufacturing while laying groundwork for infrastructure "where intelligence moves at the speed of light."

Corning has supplied optical fiber for decades; its cables already link racks across data centers operated by major cloud providers. The new deal pushes fiber closer to the processors themselves. At Nvidia's 2025 GTC conference, Huang described co-packaged optics as essential for future AI systems. Industry researcher Vlad Galabov of Omdia told CNBC that placing light conversion next to the chip reduces wasted energy because signals travel millimeters instead of across circuit boards. Power savings estimates cited by Corning executives range from five to 20 times lower than copper equivalents.

The timing aligns with explosive sector demand. One day earlier, AMD reported first-quarter revenue of $10.25 billion, beating estimates, with data-center sales up 57 percent year-over-year to $5.8 billion. AMD's stock jumped 20 percent in premarket trading on that news. Both AMD and Nvidia are racing to sell full rack-scale AI systems priced in the millions; OpenAI and Meta have committed to AMD's upcoming Helios platform while continuing to buy heavily from Nvidia. Memory shortages, advanced packaging constraints and geopolitical tensions in chip supply chains remain pressing issues across the industry.

Nvidia has already invested $4 billion in laser and component makers Coherent and Lumentum, whose technology works with Corning fiber. Competitors Broadcom, Marvell and Intel are developing similar optical solutions. Corning itself secured up to $6 billion from Meta in January to expand a North Carolina plant, a project expected to create 1,000 jobs. The company, historically known for iPhone display glass, now sees optical communications as its fastest-growing segment.

Technical and commercial questions persist. Integrating optics at chip scale demands extreme precision in alignment and thermal management; deployment timelines and exact performance gains have not been detailed beyond executive statements. Whether the partnership delivers the full promised efficiency at hyperscale remains unproven in production environments. Corning was scheduled to hold an investor day on the announcement date, coinciding with its 175th anniversary.

The announcement underscores a broader truth: continued AI progress now depends as much on innovations in interconnects and power efficiency as on raw chip performance. One set of factories will not resolve every bottleneck, yet the collaboration signals that the world's dominant AI hardware maker is betting on light to carry the next wave of intelligence.