Nvidia to invest $2b in US laser supplier Lumentum

Nvidia has signed nonexclusive multiyear agreements with Lumentum Holdings Inc. (NASDAQ: LITE), a San Jose-based maker of lasers and optical components.

The deal includes a multibillion-dollar purchase commitment and a US$2 billion NVIDIA investment to support R&D, future capacity, and operations as Lumentum builds a new US fabrication facility.

The agreement also secures future capacity access rights for advanced laser components used in optical interconnects.

Both companies said the expanded collaboration targets optical interconnects and package integration to improve energy efficiency and resiliency in large-scale AI networks.

Jensen Huang, NVIDIA CEO, said AI is driving the largest computing infrastructure buildout in history and that the partnership will advance silicon photonics for next-generation gigawatt-scale AI factories.

Lumentum supplies lasers, modules, and optical subsystems to telecom and data center customers and operates R&D and manufacturing facilities worldwide.

🔗 Source: Nvidia

🧠 Food for thought

Implications, context, and why it matters.

Nvidia’s $2 billion move adds momentum to a business already growing

Lumentum did not need a bailout. Quarterly revenue rose 65% year over year to $665.5 million, with AI and cloud infrastructure customers bringing in over 60% of revenue 1.
The agreement came with a record optical circuit switch backlog above $400 million. Much of it ties to shipments expected in the second half of 2026, plus incremental multi-hundred-million-dollar orders linked to co-packaged optics where optical connections are integrated directly into a chip package, with deliveries guided into 2027 2.
Nvidia’s $2 billion investment plus multibillion-dollar purchase commitments aim to lock in supply and future capacity access rights for advanced laser components used in optical interconnects that link equipment inside large AI data centers 3.

The AI race now depends on the network between chips

This partnership, along with a similar $2 billion deal with competitor Coherent, suggests the tight spot in AI is moving from compute to network connectivity 4.
As AI clusters grow to thousands of GPUs, older networking eats power and slows performance. Analysts now describe AI scaling as a “communication story” 4.
Nvidia is leaning on co-packaged optics. One analysis puts optical engine plus external laser source power at about 4–5W per 800G of bandwidth, versus roughly 16–17W for an 800G DSP transceiver (Digital Signal Processor transceiver, which uses electronics to process high-speed signals), or about a ~70%+ reduction 5.