Nvidia commits $6.5B to photonics to solve AI energy bottleneck

Nvidia has committed at least $6.5 billion to photonics companies in three months, betting light-based data transmission can break the energy bottleneck threatening AI infrastructure expansion.

Nvidia has committed at least $6.5 billion to photonics companies since March, betting that light-based data transmission can break the energy bottleneck constraining AI infrastructure expansion.

"Photonics represents a way for Nvidia to scale their AI infrastructure without the energy costs that staying with electrical and copper will incur," Alvin Nguyen, senior analyst at Forrester, said.

The investments include $2 billion each into Lumentum Holdings, Coherent and Marvell Technology, plus $500 million to Corning for advanced optical connectivity solutions. Nvidia also participated in Ayar Labs' $500 million Series E funding round. Co-packaged optics can cut interface power consumption to about 9 watts from roughly 30 watts, a 3.5x efficiency gain over earlier designs, according to Nvidia.

The spending spree reflects a strategic recognition that electrical data transfer through copper — the current standard — consumes too much energy to support the next generation of AI factories connecting millions of graphics processing units. Nvidia's next-generation rack-scale solutions will require exponentially more optical connectivity to handle the bandwidth demands of larger AI models, Brian Colello, senior equity analyst at Morningstar, said.

Silicon Photonics Capacity Falls Short

Speaking at the GTC conference in March, Nvidia Chief Executive Officer Jensen Huang said the company was scaling its silicon photonics technology and incorporating it into GPU-to-GPU interconnect systems. "The amount of silicon photonics technology capacity that we need is substantially higher than the world has today," Huang said, adding that Nvidia is working with suppliers to expand manufacturing capacity ahead of demand.

Nvidia has already introduced photonics-enabled networking switches — the Quantum-X for InfiniBand and Spectrum-X for Ethernet — designed to connect AI factories across sites while reducing energy consumption and operating costs.

Mass Production Remains the Hurdle

While the technology is proven, scaling it to commercial volumes presents significant manufacturing challenges. Complex co-packaged optical assemblies require precise alignment of optical and silicon components, and defects during packaging typically cannot be reworked, Nick Patience, AI lead at the Futurum Group, said.

"The technology is sound, production scale is the harder problem," Patience said. He expects large-scale adoption from 2028 onward.

Nvidia is not alone in the push. Advanced Micro Devices joined the Ayar Labs round and acquired startup Enosemi in 2025, while making equity investments in Teramount and Celestial AI. Alphabet and Microsoft's venture arms backed nEye in an $80 million Series C in April.

The market has already priced in the shift. Lumentum shares have risen 134% this year, Coherent is up 96%, Marvell has gained 122% and Corning has climbed 111%. Nvidia shares, trading at roughly 35 times forward earnings, have risen about 60% in 2026 as investors price in the company's expanding AI infrastructure footprint. The photonics transition, if executed on schedule, could save Nvidia billions in energy costs while enabling the next generation of GPU clusters — but the 2028 timeline means near-term revenue contributions from these investments remain limited.

This article is for informational purposes only and does not constitute investment advice.