(Alexander_Evgenyevich/Shutterstock)
GenAI is exposing a critical flaw in modern computing. While GPU performance continues to grow, the systems built to support these chips are struggling to keep pace. The key challenge now is data transfer. Shuttling huge volumes of data between processors has become the bottleneck, holding back the full potential of large-scale AI.
For decades, chip design advanced under Moore’s Law, with transistor counts doubling every couple of years. That trend fueled massive leaps in compute power. But as AI workloads become more complex and distributed, raw power alone no longer solves the problem.
One company stepping up to meet this challenge is OpenLight. The Santa Clara-based startup has raised $34 million in Series A funding to help scale photonic systems built for next-gen infrastructure. According to the company, its chips are custom-designed in-house and manufactured by foundry partners.
OpenLight says its photonic systems are optimized to move data more efficiently, easing the strain on networks overwhelmed by GenAI’s appetite. It positions integrated photonics as a practical alternative to the traditional interconnects that are now buckling under modern demands.
The funding round was co-led by Xora Innovation and Capricorn Investment Group, alongside Mayfield, HPE’s Juniper Networks, Lam Capital, New Legacy Ventures, and K2 Access.
At its core, silicon photonics offers a new way to connect chips, by using light instead of electricity. The upside? Faster speeds and lower power. As AI and high-performance computing stretch the limits of existing systems, this optical approach is gaining serious traction.
Spun out of Synopsys, OpenLight has carved out a niche in the design of photonic application-specific integrated circuits, or PASICs. Its chips combine lasers, modulators, amplifiers, and detectors on a single die and are backed by more than 360 patents. The company aims to embed this open photonics platform across industries including telecom, healthcare, quantum, and AI.
The fresh capital is fueling expansion. OpenLight is growing its library of components, ramping up R&D, and helping clients move from initial design into deployment. CEO Adam Carter said the investment lets them move faster and bring “groundbreaking products to market.”
To Carter, these investors are more than just capital. Some of them bring expertise in semiconductors and photonics, giving OpenLight a strategic edge as customer needs evolve. The company is not chasing generic solutions. It’s betting that more organizations will demand tailored photonic systems.
That bet mirrors broader trends in compute architecture. Carter believes heterogeneous integrated photonics will “transform the way data is processed and transmitted.” The real shift, he argues, is not just about speed but rethinking how performance, efficiency, and scale come together across entire systems.
(Alexander_Evgenyevich/Shutterstock)
Evidence of the problem is already here. A study last year by Xscape Photonics showed most AI developers use only about 25% of GPU capacity because networking can’t keep up. Those idle chips help explain why photonics, moving data with light, has drawn such keen interest.
OpenLight’s offering includes a process design kit validated at Tower Semiconductor. It gives engineers ready-made building blocks like lasers and modulators built on silicon and indium phosphide. Over 20 companies are already using it, indicating broader adoption.
To show what’s possible, OpenLight has started rolling out sample designs. These include photonic circuits running at 1.6 and 3.2 terabits per second, and a 400-gigabit modulator. To the customers, these demos are like blueprints that can help move from concept to deployment.
The question now is whether photonics can scale. Phil Inagaki of Xora Innovation said the sector is “on track for exponential growth” but warned that manufacturing capacity will be a major hurdle in the years ahead. He sees OpenLight’s design kit as a way to speed adoption.
Capricorn’s Dipender Saluja sees the need most clearly in AI data centers, where pressure is mounting fast. He said “optical connectivity has become critical for next-generation architectures” and argued that OpenLight’s platform meets the long-standing trifecta of performance, reliability, and cost. Solving all three, he believes, could finally bring photonics into the mainstream.
GenAI is speeding everything up, from upgrade cycles to infrastructure demands. OpenLight is aiming to position itself right at the center of that shift. It expects to enter volume production in 2025, with royalties kicking in the year after. To reach that point, the company still needs to scale manufacturing, align with evolving standards like co-packaged optics, and strengthen its supply chain. It also faces strong competition from Ayar Labs, Lightmatter, and Celestial AI, who are all pursuing similar goals.
Related Items
IEEE Study Leverages Silicon Photonics for Scalable and Sustainable AI Hardware