The Signal
PsiQuantum Demonstrates an 8×8 Optical Switch for AI Networks
PsiQuantum announced an 8×8 optical circuit-switching demonstrator for AI data-center networks on October 28, 2026. The company reports 1.2 dB insertion loss, 40 dB crosstalk and reconfiguration in under 3 milliseconds in its lab tests.
Quantum.world Team · Written with AI · fact-checked ·
“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.”
Victor Peng, Chief Executive Officer of PsiQuantum
An 8×8 switch built for direct optical routes
PsiQuantum announced the optical circuit switch on October 28, presenting an 8×8-radix demonstrator built on the silicon-photonics platform it developed for quantum computing. The company describes the device as strictly non-blocking: any of its eight inputs can connect to any of its eight outputs. It has no moving parts, and routes light from port to port without converting the signal to electricity inside the switch.
The announcement marks PsiQuantum’s first public demonstration of an optical circuit switch for AI infrastructure. Its intended role is to connect equipment in large computing clusters, where data moving between GPUs can constrain performance. The device is a demonstrator, not a disclosed production product; PsiQuantum says it is sharing samples with selected prospective customers and working toward commercial products.
Lab tests report 1.2 dB of insertion loss
PsiQuantum reports typical test results of 1.2 dB insertion loss, 40 dB crosstalk and reconfiguration in less than 3 milliseconds. It says the results exceeded its initial targets and required no optical amplification. The company’s optical-switching page repeats those figures, but the announcement does not give a detailed test procedure or independent measurements.
Insertion loss measures how much optical signal is lost as it passes through the switch; it is not a measure of total network loss. That distinction matters because losses accumulate as light crosses multiple components. PsiQuantum says its silicon-photonics platform was designed to limit component losses, a relevant requirement for optical links where each additional switch can reduce the signal reaching its destination.
Quantum-computing photonics moves into data centers
The switch draws on a platform PsiQuantum says it has spent nearly a decade developing for utility-scale quantum computing. The company argues that its work on low-loss photonics can also address networking demands in AI clusters. In conventional electronic switching, optical signals are converted to electrical form and back at each hop; an optical circuit switch can establish a direct light path instead.
PsiQuantum also disclosed that it has developed barium titanate integration at 300-millimeter wafer scale and uses a 64×64 BTO switch in its quantum systems. BTO is not part of the 8×8 AI demonstrator or the initial products the company says it is pursuing. Keeping those devices distinct matters: the 64×64 switch is described in the announcement as part of the quantum system, not as a larger version of the AI-network demonstrator.
Customer samples precede product details
PsiQuantum has not named the prospective customers receiving samples, provided pricing or published a commercial launch date. It says it expects to announce commercial products in the coming months. The company’s product page also says it is developing larger optical switches, but gives no port count or delivery schedule for them.
The reported figures offer an early performance snapshot, not evidence of deployment at data-center scale. Larger switches would need to preserve acceptable loss as routes cross more components, and the release supplies no field results or customer commitments. For now, PsiQuantum is sharing samples of an 8×8 demonstrator with selected prospective customers.