Nvidia bets billions on photonics for AI future

Nvidia has committed at least $6.5 billion into companies developing photonics technology over the past three months, aiming to solve a major bottleneck in the rollout of artificial intelligence [1]. The chipmaker is betting that using light to transmit data will offer a more efficient alternative to current electrical methods, which consume significant energy and are increasingly seen as a barrier to broader AI deployment [1].

Key takeaways

• Nvidia invested $6.5 billion in photonics companies like Lumentum and Marvell to improve data transfer efficiency [1].
• Photonics uses light instead of electricity to move data between GPUs and servers, reducing energy consumption [1].
• Manufacturing complex optical assemblies remains a challenge, with large-scale adoption expected around 2028 [1].
• Nvidia currently controls between 70% and 95% of the AI chip market [2].

Overcoming energy bottlenecks

Since the beginning of March, Nvidia announced $2 billion investments into Lumentum, Coherent, and Marvell, all of which are developing photonics tech [1]. The company also committed $500 million to Corning for advanced optical connectivity solutions and participated in Ayar Labs' $500 million funding round [1]. Following these investments, stocks for Lumentum, Coherent, Marvell, and Corning have risen between 96% and 134% since the start of the year [1].

Photonics utilizes light to move data between graphics processing units (GPUs), memory, and data centers, rather than relying on electrical signals over copper [1]. While copper is currently the standard due to lower cost, analysts argue that Nvidia's next-generation solutions will require increased optical connectivity to handle rising bandwidth from new AI models [1]. CEO Jensen Huang stated in March that the company is scaling its silicon photonics technology and adding it to GPU-to-GPU interconnects, noting that the required capacity is "substantially higher than the world has today" [1].

Securing the AI lead

This aggressive investment strategy comes as Nvidia holds a dominant position in the AI chip market, with estimates suggesting it controls between 70% and 95% of the market for chips used to train and deploy models [2]. The company reported a 78% gross margin in the latest quarter, significantly higher than rivals Intel and AMD [2]. However, Nvidia faces rising competition from chipmakers like AMD and Intel, as well as cloud providers like Amazon and Google, which are developing their own internal processors [2].

The photonics sector is attracting attention beyond Nvidia, with AMD joining the Ayar Labs round and acquiring startup Enosemi, while Alphabet and Microsoft backed nEye [1]. Despite the enthusiasm, industry experts caution that scaling the technology is difficult. Manufacturing yields on complex co-packaged optical assemblies are challenging because precise alignment is unforgiving and errors often cannot be reworked [1].