Silicon Photonics Nvidia Invests 4 Billion US Dollars in US Photonics Production

Related Vendors

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

Copper cables are reaching their physical limits at 1.6 TByte/s. With an investment of USD 4 billion, the company is aiming to avoid a looming bottleneck in the AI infrastructure and create the basis for energy-efficient gigawatt factories.

In the world of high-speed electronics, we are currently hitting a physical wall. Anyone designing circuit boards for data centers, high-resolution imaging processes in medical technology or precise optical measurement systems today is familiar with the problem: the signal integrity on copper lines collapses drastically at frequencies above 200 GBit/s per lane. Attenuation, thermal load and electromagnetic interference (EMI) complicate the layout.

Nvidia has now announced a four billion dollar deal with heavyweights Lumentum and Coherent, to be split equally between the two companies. What at first glance appears to be purely a business announcement is, however, much more. It is the starting signal for the widespread introduction of silicon photonics into standard hardware architecture.

This will strengthen Nvidia's strategy to secure the silicon photonics supply chains that enable the broadband, energy-efficient optical interconnects that are critical to scaling AI infrastructures. This includes the next generation of switches and networking equipment. In addition to the two billion US dollars in cash, the two companies will receive multi-year supply agreements and access rights to future capacity for laser components and optical products.

Laser Components And Optical Petwork products

Lumentum plans to build a new manufacturing facility in the US to increase production of advanced laser components as demand outstrips current supply in phosphide indium fabs. The investment at Coherent will be used to expand manufacturing of laser and optical networking products in the US, based on a 20-year partnership.

There is a simple physical necessity behind Nvidia's massive billion-euro investment: the AI infrastructure of tomorrow needs more speed with less energy. Conventional copper connections capitulate to the gigantic data volumes of modern data centers. This is where silicon photonics comes in. It is the technological lever to massively increase transfer rates between clusters while maintaining energy efficiency. This is a decisive factor in enabling the scalability of AI applications in the first place. With this investment, Nvidia is not only securing its technological superiority, but also cementing its leading position in the global AI ecosystem through a controlled supply chain.

In this context, Nvidia CEO Jensen Huang is already talking about "AI factories on a gigawatt scale". The memory of the painful supply bottlenecks for memory modules in 2025 runs deep and Huang wants to proactively prevent a similar bottleneck for optical components. The bare figures underline that this is no mere experiment: With network sales of over 31 billion US dollars in the 2026 financial year, connectivity has long since matured into the Group's second mainstay. (heh)

Computing With Light

Photonic Processor NPU 2.0 Aims to Solve the Energy Problem