PsiQuantum Launches Optical Circuit Switch for AI Infrastructure

PsiQuantum recently revealed its new Optical Circuit Switching technology, which uses a silicon photonics platform to address scaling issues in artificial intelligence data centers. This advancement targets the high demands of networking between massive compute pools. The company successfully tested an 8x8 demonstration vehicle that shows high efficiency and speed without needing optical amplification.

Advanced Silicon Photonics for Data Centers

The new technology utilizes a silicon photonics platform that the company developed over nearly ten years. This foundation was originally built for the rigorous requirements of utility-scale quantum computing. By applying these standards to AI networking, the company provides a solution for the rapid growth of compute infrastructure. Current AI systems are expanding at a rate that often exceeds the capacity of traditional network connections.

Data movement is a significant bottleneck for modern AI scaling. Moving information between systems efficiently is now a requirement for continued industry growth. Optics offer a faster and more power-efficient method for these connections. The demonstration vehicle features an 8x8-radix design that focuses on low-loss performance. This approach ensures that signals remain strong as they travel through the switching fabric.

Internal Testing and Performance Metrics

Laboratory tests conducted by the company show that the switch performs better than initial design goals. The hardware achieved a typical insertion loss of 1.2 dB. This low loss level is critical because it allows data to move without the heat and power consumption of extra amplifiers. High-performance networking requires maintaining signal integrity across complex paths, and these results indicate a stable platform for high-speed data transfers.

The switch also demonstrated a crosstalk performance of 40 dB. This metric is vital for preventing interference between different data streams in a dense data center environment. Furthermore, the reconfiguration time was measured at less than 3 milliseconds. Rapid switching allows the network to adapt to changing workloads in real-time. These technical specifications suggest that the platform can handle the intense traffic patterns found in large-scale machine learning tasks.

Transitioning to Commercial Production

The company is now distributing samples of the 8x8 demonstration vehicle to a select group of potential customers. This phase allows partners to evaluate the technology within their own environments. Commercial products based on this platform are expected to be announced in the coming months. The shift toward commercialization marks a new chapter for the organization as it enters the AI hardware market.

The move into AI infrastructure does not replace the primary mission of building fault-tolerant quantum computers. Instead, the two fields support each other. More efficient AI infrastructure helps in the design and control of quantum systems. Meanwhile, the extreme engineering required for quantum mechanics provides tools that improve standard data center operations. This dual focus allows the company to generate value from its research while pursuing long-term goals.

Synergies Between Quantum and AI Hardware

The engineering requirements for quantum computing are often much stricter than those for AI networking. By solving the most difficult scaling problems first, the company created a platform that naturally fits the needs of modern data centers. Chief Executive Officer Victor Peng noted that the decision to tackle hard problems early in their roadmap created this current opportunity. The technology is ready to work in AI environments today while quantum development continues.

Optical circuit switching is becoming an essential part of the conversation around AI growth. Executive Vice President Rob Soderbery highlighted that the complexity of computing infrastructure is growing at an extraordinary pace. Optics will play a larger role in supporting this expansion. The silicon photonics platform was tested against some of the most demanding requirements imaginable during its development for quantum applications.

Integration of New Materials

In addition to silicon photonics, the company developed integration techniques for barium titanate (BTO). This work was performed at the 300mm wafer scale, which is the standard for high-volume semiconductor manufacturing. The company already uses a 64x64 BTO switch in its internal quantum systems. While BTO is not a requirement for the current 8x8 demonstration vehicle, its existence provides a clear upgrade path for future products.

Barium titanate allows for even faster switching speeds and more advanced optical functions. Having this material integration ready at a commercial scale ensures that the company can stay ahead of increasing performance demands. The current silicon photonics platform provides exceptional performance for today’s needs, while the BTO technology sits in reserve for the next generation of high-speed switches. This layered approach to hardware development protects the company against technical obsolescence.

Strategic Support and Manufacturing

The importance of these breakthroughs has gained national attention. In September 2026, the company signed definitive documentation with the U.S. Department of Commerce. This agreement is designed to accelerate research and development efforts. It also focuses on improving domestic manufacturing capabilities for critical optical components. The partnership underscores the strategic value of high-performance optical switches in the broader technology landscape.

Building these components in the United States helps secure the supply chain for vital AI and quantum infrastructure. The Department of Commerce involvement suggests that these technologies are viewed as essential for future economic and technological leadership. By focusing on both performance and manufacturability, the company is positioning itself as a key supplier for the next era of computing. This support helps transition laboratory breakthroughs into reliable products for the global market.

Future Outlook for Optical Networking

The introduction of the optical circuit switch comes at a time when the industry is searching for ways to reduce power consumption. Electronic switches often generate significant heat and require massive amounts of electricity to move data at high speeds. Optical switches move light instead of electrons, which can drastically lower the energy footprint of a data center. This efficiency is a major selling point for operators trying to manage rising utility costs.

As AI models continue to grow in size, the need for low-latency communication between GPUs and memory becomes even more pressing. The speed of light provides the ultimate limit for data transmission, and optical switching brings the network closer to that limit. By removing the need for repeated conversions between electrical and optical signals, the company’s technology simplifies the data path. This simplification results in both better performance and a more reliable system architecture.

Expanding the Product Roadmap

The company plans to release a full line of commercial products following the success of the 8x8 demonstration. These products will likely target different scales of data center operations, from individual racks to massive warehouse-scale facilities. The flexibility of the silicon photonics platform allows for various configurations to meet specific customer needs. This versatility is important in a market where different AI architectures require unique networking topologies.

The ongoing development of quantum computers will continue to feed innovation into the AI product line. As the quantum systems scale up to more qubits, the switching requirements will become even more intense. The lessons learned from those extreme environments will lead to updates for the AI-focused switches. This creates a cycle of innovation where each breakthrough in one field provides a direct benefit to the other.

Impact on the Global AI Market

The availability of high-performance optical switches could change how AI clusters are designed. Currently, many clusters are limited by the physical distance that electrical signals can travel without degrading. Optics allow for larger, more distributed clusters that still function as a single unit. This capability enables the training of much larger models than were previously possible. It also allows for more flexible data center layouts, which can improve cooling efficiency and space utilization.

The tech industry’s move toward optics is not just a trend but a physical necessity. As speeds increase, copper wiring faces insurmountable physical hurdles. Silicon photonics represents the most viable path forward for the next decade of networking. The announcement of this new switch places the company at the center of a major shift in how the world’s most powerful computers are built and connected. With government support and a clear technical lead, the organization is prepared to influence the direction of both AI and quantum computing.