PsiQuantum is building its first product. It isn’t a quantum computer, but it could become a superpower for AI data centers.

The company has been working for a decade on building quantum computers that use photons as “bits” of information and has built its own optical chips that control light particles with extreme precision. That ability to manipulate light could turn out to be a marketable tool in the next phase of the AI race, where data centers, bottlenecked by the limits of copper wiring, are switching to using fiber optics instead.

PsiQuantum says it has been showing potential customers a prototype of a new “optical circuit switch” that could be used to route optical signals from GPUs across data centers much more efficiently.

Customers were most focused on PsiQuantum’s ability to beat a benchmark known as “insertion loss,” measured in decibels, that determines how many photons escape during transport, potentially leading to a degraded signal. To work in an AI datacenter, insertion loss needed to be less than two to three decibels, PsiQuantum said. Its prototype barreled through that requirement, reaching 1.2.

The AI boom wouldn’t have happened without the ability to network hundreds of thousands of AI processors together. But all that networking is also the data center’s weakest link. While GPUs can crunch numbers at blazing speeds, they are bottlenecked by the comparatively slow copper that carries the output of that math from GPU to GPU.

That’s especially true during the training process of increasingly large AI models that require the most coordination between large numbers of GPUs.

The next generation of data center racks are more often using optical cables to carry the data between GPUs. But optical information must be converted into an electrical signal in order to be routed to its destination, then converted back into light, before it’s converted once again when it reaches its destination.

All those conversions happening billions of times per second in the largest data centers require optical switches that generate heat, energy and bulk.

PsiQuantum says its new device can do that without ever having to convert the photons into electrical signals. It could be used to create speed-of-light connections between server racks in AI data centers that generate almost no heat and consume negligible energy.

The only major hyperscaler currently using purely optical circuit switching in an AI data center today is Google, which built a system that uses tiny mirrors, known as MEMS, to direct the light signal.

Robert Soderbery, executive vice president of PsiQuantum, said in an interview with Semafor that separate data center racks will effectively behave as one, using PsiQuantum’s device. Data could flow between GPUs in separate server racks at the same speed and latency as they would between GPUs interconnected inside a single rack. That would be a huge advantage because today server racks are reaching the limit of compute density, size and energy efficiency.

PsiQuantum isn’t alone. The battle to convert data center networking to photonics has been raging since the ChatGPT moment kicked off the AI race. Earlier this year, photonic computing company Lightmatter helped launch an initiative to standardize optics in the data center space. Nvidia has been spending billions on investments in optics computing companies like Lumentum and Coherent.

For now, the GPU reigns supreme but data centers are seeing a lot more of the light.