NVIDIA CEO Jensen Huang is betting that the future of artificial intelligence infrastructure depends on lasers, not just chips. In March, Huang announced a $2 billion investment and partnership with optical component maker Lumentum, along with a similar deal with competitor Coherent, to advance silicon photonics technologies that will power the next generation of AI data centers . This move reveals a critical bottleneck in AI infrastructure that most people don't see: the networking that connects thousands of GPUs (graphics processing units) together.

The partnership reflects a fundamental shift in how AI systems are being designed. As data centers grow larger and more complex, traditional copper wiring and electrical switches can't keep up with the speed and power demands of modern AI workloads. Huang's solution involves embedding lasers directly into GPU designs, a practice he calls "extreme co-design" . These optical components allow data to travel at the speed of light through fiber optic cables, dramatically reducing latency and power consumption compared to conventional electrical networking.

What Are the Key Technologies Behind This AI Infrastructure Revolution?

The optical technologies at the heart of this partnership fall into two main categories that will reshape how AI data centers operate:

• Co-packaged Optics (CPO): These are lasers built directly into GPU designs, allowing data to stay in the optical domain without conversion to electrical signals. This reduces power consumption and latency in AI clusters.
• Optical Circuit Switches (OCS): These network devices establish direct, physical paths of light between ports, acting like software-controlled fiber patch panels that bypass traditional electrical switches and reduce latency in data centers.
• Silicon Photonics: The underlying technology that integrates optical components with semiconductor manufacturing, enabling the miniaturization and cost reduction needed for widespread deployment in AI infrastructure.

"AI has reinvented computing and is driving the largest computing infrastructure buildout in history. Together with Lumentum, NVIDIA is advancing the world's most sophisticated silicon photonics to build the next generation of gigawatt-scale AI factories," said Jensen Huang, founder and CEO of NVIDIA.

Jensen Huang, Founder and CEO of NVIDIA

How Is Lumentum Preparing to Scale Production for This Optical Boom?

Lumentum, the primary beneficiary of NVIDIA's investment, is making aggressive moves to meet anticipated demand. The company announced the acquisition of a North Carolina chip manufacturing facility from Qorvo, which it will retrofit to produce indium phosphide-based optical devices, including continuous-wave and ultra-high-power lasers for co-packaged optics . The 240,000-square-foot facility in Greensboro represents a major capacity expansion that Lumentum plans to invest hundreds of millions of dollars into over the coming years.

The scale of Lumentum's ambitions is striking. When the new facility ramps up in early 2028, management projects it will add $5 billion in annual revenue capacity at chip-level margins . The company is officially targeting $2 billion in quarterly revenue, which would represent an $8 billion annualized run rate. To reach this goal, Lumentum laid out a timeline to hit $1.25 billion per quarter within 9 to 12 months, followed by the $2 billion quarterly target in another 9 to 12 months . As the company scales, management expects operating margins to expand to between 38% and 42%.

Demand for these optical components is already outpacing supply. Lumentum reported that it is currently under-shipping customer demand by 25% to 30%, with customers requesting lasers in the billions of units . The company has a $400 million backlog for optical circuit switches alone in the second half of this year, and management announced a brand-new multi-billion-dollar OCS agreement with a major hyperscaler that projects OCS will ramp to a greater than $1 billion run rate in 2027 .

Why Does This Matter for the Future of AI?

The optical infrastructure market is experiencing explosive growth. Lumentum management showed projections indicating the AI optical total addressable market (TAM), or the total potential revenue opportunity, will explode from $18 billion today to over $90 billion by 2030 . This five-fold expansion reflects the massive capital investments being made in AI data centers worldwide.

The shift toward optical networking also addresses a critical pain point in AI infrastructure. As AI models grow larger and require more computational power, the networking that connects GPUs becomes a bottleneck. Traditional electrical switches consume significant power and introduce latency that slows down AI training and inference. By moving to optical systems, data centers can achieve faster data transfer rates while consuming less power, making it possible to build the "gigawatt-scale AI factories" that Huang mentioned .

NVIDIA's $2 billion commitment signals confidence that this technology transition is not just theoretical but essential for the next phase of AI infrastructure development. The partnership includes a multibillion-dollar purchase commitment from NVIDIA and future capacity access rights for advanced laser components, ensuring that Lumentum and Coherent have guaranteed demand as they scale manufacturing .

For investors and industry observers, this move reveals where the real bottlenecks in AI infrastructure lie. While GPU performance captures headlines, the unsexy but critical work of connecting those GPUs together is becoming just as important. Jensen Huang's $2 billion bet suggests that optical networking is not a niche technology for the future, but a fundamental requirement for building the AI infrastructure of the next decade.

" }