San Francisco, March 4, 2026, 04:22 PST
- Nvidia plans to put $2 billion apiece into Lumentum and Coherent through long-term optics deals.
- Among the agreements: multibillion-dollar buys and rights to capacity, all for laser and optical networking gear aimed at powering AI data centers.
- Wall Street has its sights set on Nvidia’s March GTC conference, seeking fresh details on the company’s roadmap—and how it plans to stack up against rivals.
Nvidia Corp plans to pump $2 billion apiece into photonics suppliers Lumentum and Coherent, betting the move will accelerate connections inside its AI data-center platforms. The news sent Lumentum up roughly 5% and Coherent climbing 9% on Monday.
This latest step zeroes in on an issue adjacent to the GPU. Photonics, which relies on light—think lasers and transceivers—to shuttle data more quickly and efficiently than old-school electrical links, comes into play as AI workloads move from training to inference, the routine process of running models to produce outputs.
Nvidia’s recent supplier moves point to a broader supply-chain crunch, not just in chips. The company is securing future capacity for advanced lasers and optical networking gear—key components for scaling up AI data centers—by putting up purchase commitments and providing direct funding, according to Investopedia.
Coherent described the new deal with Nvidia as nonexclusive and spanning several years, with Nvidia making a multibillion-dollar purchase commitment tied to guaranteed access and capacity for advanced laser and optical networking products. “Computing has fundamentally changed,” Nvidia CEO Jensen Huang said in the release. Coherent CEO Jim Anderson pointed to the partnership as a deepening of a relationship that goes back two decades. Coherent
Nvidia has made a comparable deal with Lumentum—also nonexclusive—and plans to invest $2 billion toward research, future capacity, and operations as Lumentum sets up a new U.S. fabrication plant. CEO Huang called out the ambition: “gigawatt-scale AI factories,” those sprawling data centers built to train and deploy AI models. NVIDIA Newsroom
Nvidia is gearing up to showcase its expanding systems ambitions at the GPU Technology Conference (GTC) in San Jose, scheduled for March 16-19. The company anticipates a crowd topping 30,000 attendees representing over 190 countries. CEO Jensen Huang is slated for a March 16 keynote.
Some analysts link the uptick in optics spending to deeper questions about Nvidia’s long-term competitive position. On Tuesday, Morgan Stanley made Nvidia its top semiconductor pick and said the valuation looked like “a surprisingly good entry point.” They argued GTC might calm nerves over market-share, but also pointed out AMD and Broadcom are expanding more quickly this year—though from much smaller platforms. Investopedia
Just days after Nvidia posted record quarterly revenue of $68.1 billion and projected first-quarter revenue hitting roughly $78 billion, the company remains firmly in the AI spending spotlight.
Still, Nvidia can’t just snap its fingers and make photonics happen. Should AI infrastructure budgets pull back, or clients lean further into developing their own chips and networking, Nvidia could find itself investing in capacity that doesn’t quickly turn into shipped products or profits.
Nvidia is shelling out to expand the supply routes for gear connecting chips, racks, and entire data centers. The question now: will those payments bring more concrete delivery schedules and less congestion, especially with GTC approaching?
