Could beams of light become AI’s next growth engine? NVIDIA is betting US$6.5 billion on photonics to overcome energy constraints and scale future AI infrastructure.
NVIDIA has committed at least US$6.5 billion to companies developing photonics technology since March 2026, signalling a major push towards addressing one of the key infrastructure challenges facing the rapid expansion of artificial intelligence (AI).
Photonics, which uses light rather than electrical signals to transmit data, is increasingly viewed as a potential solution to the growing energy and performance demands of AI systems.
Traditional data transmission relies on electricity travelling through copper connections, a method that consumes significant power and is becoming a limiting factor as AI models grow larger and more complex.
According to a report by CNBC, over the past three months, NVIDIA has announced investments totalling around US$2 billion in photonics-related companies including Lumentum Holdings, Coherent and Marvell Technology.
The company has also pledged US$500 million to Corning to support advanced optical connectivity technologies and participated in optics startup Ayar Labs’ US$500 million Series E funding round.
Industry analysts say photonics could play a critical role in enabling future AI infrastructure. The technology can be used to transfer data between graphics processing units (GPUs), memory systems, networking chips, servers and data centres more efficiently than conventional copper-based systems.
According to NVIDIA chief executive Jensen Huang, demand for silicon photonics is expected to rise sharply as the company expands its AI networking platforms and introduces optical technologies into GPU interconnects. Speaking at NVIDIA’s GTC conference in March, Huang said the capacity required for silicon photonics would exceed current global supply levels, prompting the company to work closely with supply-chain partners to expand production capabilities.
Meanwhile, NVIDIA is not alone in pursuing photonics. Advanced Micro Devices (AMD) has also invested in Ayar Labs and acquired photonics startup Enosemi, while investment arms linked to Alphabet and Microsoft recently backed optical networking company nEye.
Despite growing interest, industry experts caution that large-scale adoption remains several years away. Manufacturing photonic components at commercial scale presents significant technical challenges, particularly in achieving precise alignment between optical and silicon elements.
Analysts expect broader deployment of photonics across AI infrastructure to accelerate towards the end of the decade as production capabilities mature.
