The Growing Demand for Photonics in Data Centers: A Structural Shift in AI Infrastructure

May 10, 2026 – In the race to build bigger AI systems, one constraint is becoming impossible to ignore: moving data fast enough without blowing out power budgets. That is why photonics—once a specialized corner of the tech supply chain—is rapidly becoming central to the future of data centers.

Silicon photonics, optical interconnects, and advanced lasers are no longer optional upgrades. They are increasingly the technologies that determine whether an AI data center can scale efficiently, control energy use, and keep performance from stalling.

Market Growth and Financial Metrics

The numbers underline how quickly this shift is taking hold. The global silicon photonics market reached $2.62 billion in 2025 and is projected to climb to $34.34 billion by 2035, a 29.6% CAGR from 2026 through 2035. Data centers and high-performance computing remain the largest end markets.

Within that broader category, the optical transceiver market reached $14.7 billion in 2025 and is expected to grow to $17.15 billion in 2026, with projections of $46.12 billion by 2034. The fastest growth is concentrated in high-speed transceivers built for AI workloads. The 400G optical transceiver market was valued at $1.13 billion in 2025, is projected to reach $1.27 billion in 2026, and could expand to $3.61 billion by 2035.

The broader optical interconnect market for AI data centers reached $3.75 billion in 2025 and is forecast to grow to $18.36 billion by 2033, far outpacing more traditional data center component categories.

The Power and Performance Imperative

The reason is straightforward. Data centers now consume 1% to 2% of global electricity, and AI workloads account for about 20% of power use in modern facilities. In large AI clusters, traditional copper-based interconnects are becoming a serious bottleneck. The cables and transceivers linking thousands of GPUs can represent roughly half of total network cost and more than half of power consumption.

Photonics offers a way around that wall. Optical approaches can cut AI pod power consumption by up to 40% while keeping interconnect latency below 10 milliseconds. They can also reduce data center floor space requirements by about 20%, an increasingly important advantage as operators push for greater density.

That helps explain the rapid shift from 400G to 800G and 1.6T connectivity. Deployments of 400G and 800G systems jumped 40% in 2023, and the upgrade cycle is still building through 2026.

Strategic Investments and Market Catalysts

A major turning point came on March 2, 2026, when NVIDIA announced a $4 billion strategic investment in photonics companies. The company invested $2 billion each in Lumentum Holdings and Coherent, while also making multi-year, multi-billion-dollar purchase commitments for advanced lasers and optical components. NVIDIA also participated in a $500 million funding round for Ayar Labs, which is developing optical chiplets.

The strategy is clear: secure supply, lock in manufacturing capacity, and gain an edge in the next generation of AI infrastructure. As CEO Jensen Huang put it, “Together with Lumentum, NVIDIA is advancing the world's most sophisticated silicon photonics to build the next generation of gigawatt-scale AI factories.”

Stock Performance and Investor Implications

Investors have noticed. The photonics sector has produced some of the strongest year-to-date returns in the U.S. market as of spring 2026. Lumentum has emerged as a major beneficiary of NVIDIA’s backing and the broader buildout in AI infrastructure. Corning gained 73% in 2026 alone, while analysts expect $8.1 billion in fiscal 2026 optical-segment sales.

Bank of America analyst Tal Liani expects the optical transport market to grow 10% in both 2026 and 2027, helped by what many in the industry now call an optics supercycle. Tight supply has also strengthened pricing power. Among the top five cloud providers, optical spending as a share of total capital expenditure is expected to rise from 2.7% in 2025 to 3.1% in 2026.

Forward-Looking Implications

Photonics is no longer just supporting infrastructure. It is becoming a core determinant of how far and how fast AI can scale. As models grow larger, the challenge is not just computing power—it is moving enormous volumes of data efficiently between chips, racks, and systems.

That makes photonics one of the clearest structural investment themes in AI infrastructure. For retail investors, the opportunity extends across the optical interconnect value chain, from components and materials to integrated modules and systems. If AI is the story investors are chasing, photonics is increasingly the part of the stack making that story possible.