Nvidia spent $4 billion on photonics to stay ahead of the AI ​​curve

Nvidia announced Monday that it is investing $2 billion each Lumentum and coherentBoth are developing photonics technology for data centers, such as optical transceivers, circuit switches, and lasers, which are used to transmit data at high speeds over long distances. Their technology could improve energy efficiency, data transfer speeds and bandwidth in future AI data centres, having already benefited from its 2020 acquisition of Nvidia. Mellanox Networking Equipment Company To enhance NVLink and increase the amount of data transferred between its GPUs.

For Lumentum, the non-exclusive, multi-year deal includes “a multi-billion dollar purchase commitment and rights to access future capacity for advanced laser components,” as well as support for expanding R&D and manufacturing. The Coherent deal is described similarly, with a “billion-dollar purchase commitment and future access and capacity rights to advanced laser and optical networking products.”

The growth of agentic AI such as Anthropic Claude Cowork And Microsoft Co-pilot duties Maximizes the bandwidth needs of AI data centers to support rapid multitasking. Photonics can offer a solution – Optical fiber is able to support much higher bandwidth and lower latency than copper cables while also using less power.

Nvidia isn’t the only organization interested in photonics, either. Last month, the Defense Advanced Research Projects Agency (DARPA) made the announcement Call for research proposals To improve optical computing, especially with regard to artificial intelligence applications. AMD is also a competitor to Nvidia Acquires silicon photonics startup Enosemi last year, which it said would “accelerate” AMD’s innovation in optics for its AI systems.