When Quantum Tech Meets the AI Bottleneck: PsiQuantum's Optical Pivot

PALO ALTO, CA – October 28, 2026 – The modern artificial intelligence boom is running into a physical wall. While the processors driving large language models have grown exponentially faster, the networks connecting them are choking. Moving petabytes of data across massive clusters of GPUs requires immense power, generates crippling heat, and introduces latency that leaves million-dollar chips sitting idle. Today, a solution to this classical computing crisis has emerged from an unlikely source: quantum engineering.

PsiQuantum, a company that has spent the last decade building utility-scale, fault-tolerant photonic quantum computers, has officially unveiled its Optical Circuit Switching (OCS) technology. By repurposing its proprietary, low-loss silicon photonics platform originally designed for quantum qubits, the company is stepping directly into the multi-billion-dollar AI infrastructure market.

The announcement marks a fascinating convergence of two distinct technological eras. It demonstrates how the extreme engineering required for tomorrow’s quantum machines is uniquely positioned to solve the most pressing bottlenecks in today’s AI data centers.

Breaking the AI Scaling Wall

To understand the significance of PsiQuantum’s entry into optical switching, one must look at the architecture of a modern AI data center. Training advanced AI models generates massive amounts of "east-west" traffic—constant, high-volume communication between thousands of GPUs.

Traditionally, this traffic is managed by Electrical Packet Switches (EPS). However, routing optical data through electrical switches requires power-hungry optical-electrical-optical (OEO) conversions. At the scale of modern AI, these conversions consume thousands of watts per chassis, add critical microseconds of latency, and demand extensive cooling infrastructure. Latency issues have more than doubled since 2024, becoming the most widespread operational challenge in enterprise AI.

Optical Circuit Switching bypasses this bottleneck entirely by keeping the data in the form of light. OCS uses microscopic mechanisms to physically or electro-optically direct light from one fiber to another. Because the data never converts to electricity, OCS is entirely agnostic to data rates and wavelengths. A data center can upgrade its endpoint transceivers from 400G to 1.6T without ever replacing the core optical switch.

While hyperscalers like Google have pioneered OCS using Micro-Electro-Mechanical Systems (MEMS) mirrors, traditional systems often suffer from slow reconfiguration times—sometimes taking seconds to remap network pathways. PsiQuantum’s silicon photonics approach shatters these limitations.

Testing of PsiQuantum’s 8x8-radix demonstration vehicle revealed a typical insertion loss of just 1.2 dB and 40 dB of cross talk. More importantly, the hardware demonstrated a reconfiguration time of less than 3 milliseconds without the need for signal amplification. In the world of optical networking, a 1.2 dB insertion loss is exceptionally low, practically eliminating the need for thermal-heavy optical amplifiers that degrade signal integrity. Meanwhile, the sub-3 millisecond switching speed provides the dynamic flexibility required for highly variable AI workloads.

Monetizing Deep Tech in the AI Gold Rush

For PsiQuantum, the move into AI data center networking is a masterclass in monetizing deep tech. Building a fault-tolerant quantum computer is a notoriously long-horizon endeavor. By pivoting a subset of its technology toward an immediate, desperate market need, the company secures near-term revenue and solidifies its strategic relevance.

"This opportunity exists because we chose to solve the hardest scaling problems at the beginning of our roadmap to utility-scale quantum computing," said Victor Peng, Chief Executive Officer of PsiQuantum. "As AI computing pushes infrastructure to its limits, we now have an opportunity to put our state-of-the-art technology to work today while continuing toward our ultimate goal: building and deploying the world's first useful, fault-tolerant quantum computers."

The company is already sharing samples of its 8x8 demonstration vehicle with select prospective customers, with commercial products slated for announcement in the coming months.

"AI is driving extraordinary growth in the scale and complexity of computing infrastructure, and optics will play an increasingly important role in enabling that growth," said Rob Soderbery, Executive Vice President at PsiQuantum. "PsiQuantum has spent nearly a decade developing silicon photonics against some of the most demanding requirements imaginable. What we are demonstrating today is that this platform can deliver exceptional performance for optical circuit switching for AI computing."

Interestingly, PsiQuantum is keeping its most advanced materials in reserve. The company has successfully developed 300mm wafer-scale integration of barium titanate (BTO), an electro-optic material that allows light to be modulated with vastly less voltage and energy. While PsiQuantum uses 64x64 BTO switches in its quantum systems, the initial AI demonstration vehicle intentionally omits this layer, prioritizing monolithic integration on standard silicon. This preserves a clear upgrade path for future, higher-tier commercial products as data center demands inevitably escalate.

Strategic Silicon and the New Industrial Policy

Beyond the commercial implications, PsiQuantum’s optical switching technology has become a focal point of U.S. industrial policy. The geopolitical race for AI and quantum dominance relies entirely on the physical hardware that underpins it, and the U.S. government is moving aggressively to onshore these critical supply chains.

In September 2026, PsiQuantum finalized definitive documentation with the U.S. Department of Commerce for a $100 million award under the CHIPS and Science Act. This funding is explicitly targeted at accelerating domestic semiconductor process development and advancing the manufacturability of critical photonic components, including high-performance optical switches.

The federal investment directly supports the 300mm wafer production of BTO at PsiQuantum’s Santa Clara facility, as well as automated chip-to-fiber packaging lines in Milpitas, California. This is not an isolated government engagement. Over the past few years, the company has secured a $125 million agreement with DARPA and a conditional $150 million loan commitment from the Department of Defense’s Office of Strategic Capital.

By backing PsiQuantum, federal agencies are hitting two birds with one stone. They are subsidizing the foundational hardware necessary to maintain American leadership in quantum computing, while simultaneously securing a domestic supply of the optical networking components that will keep U.S. artificial intelligence infrastructure running at peak efficiency.

The Symbiosis of AI and Quantum

The optical circuit switch market is projected to surpass $12 billion by 2035, driven almost entirely by the scale-up requirements of massive GPU clusters. As hyperscalers look to eliminate power-hungry electrical conversions and cut cooling costs, the demand for low-loss, high-speed optical switching will only intensify.

PsiQuantum’s expansion into classical AI infrastructure highlights a profound symbiosis between today’s computing paradigm and tomorrow’s. AI and quantum computing are complementary forces. More efficient AI infrastructure, enabled by advanced optical switching, will accelerate the complex simulations and control systems required to operate quantum machines. In return, the relentless engineering required to manipulate individual photons for quantum qubits has yielded the exact networking hardware needed to save the modern AI data center from choking on its own data.

As the tech industry scrambles to rebuild its physical infrastructure to support the weight of artificial intelligence, the solutions are requiring us to look beyond traditional silicon. The future of classical computing, it seems, is being forged in the crucibles of quantum research.