TL;DR
- PsiQuantum has spent a decade building photonic quantum computers — machines that use photons as their information bits — and precision optical chips to control them. Its first product, per Semafor's reporting, is not a quantum computer. It is a prototype optical circuit switch for SI data centers, being shown to potential customers.
- The claim: the switch routes light between GPU racks without converting photons into electrical signals, and the piece reports one hard benchmark — insertion loss of 1.2 decibels, against a stated data-center requirement of less than two to three decibels.
- The company it keeps is serious: Google already runs purely optical circuit switching in its data centers, Lightmatter has pushed an optics-standardization initiative, and Nvidia has spent billions investing in optics companies including Lumentum and Coherent.
- What the story does not contain: a named customer, a sale, a price, or a ship date. The evidence supports a photonics claim today. The quantum claim rides along, untouched in either direction.
What is PsiQuantum actually selling?
Start with the bottleneck. Inside an SI data center, the limiting factor is increasingly not the processors — it is the wiring between them. Networks of hundreds of thousands of processors need racks of GPUs to behave like a single machine, and in the largest data centers the signals between them are converted back and forth between light and electricity billions of times per second. Every conversion costs speed, power, and heat.
PsiQuantum's prototype attacks that conversion step. The switch is designed to route optical signals between racks while keeping them optical — operating, in the company's description, "without ever having to convert the photons into electrical signals." Executive vice president Robert Soderbery told Semafor that separate racks would "effectively behave as one," with data moving between GPUs in different racks at the same speed and latency as GPUs inside a single rack. The company also claims the connections generate almost no heat and use negligible energy — claims we attribute to the company, because that is what they are. One number in the story can be checked against a stated bar. It deserves its own section.
What is insertion loss, and why is 1.2 the whole ballgame?
Insertion loss is the measure of how much of your signal never arrives: photons escape during transport, and the signal degrades. It is quoted in decibels, and lower is better. Per Semafor's reporting, the requirement for an SI data center was "less than two to three decibels" — and PsiQuantum's prototype met it, "reaching 1.2."
Notice what kind of fact that is. It is not an adjective. "Almost no heat" and "negligible energy" are adjectives wearing lab coats until someone publishes the measurements behind them. A benchmark with a stated requirement and a reported result is a different species of claim: it can be compared, questioned, and eventually verified or refuted by the customers now being shown the device. In a story built mostly of promises, 1.2 dB is the load-bearing number — the one piece of evidence that the engineering, at prototype scale, does what the pitch says.
Why is a quantum company good at this?
Because the quantum computer was the training ground. PsiQuantum's approach to quantum computing uses photons as information bits, which meant spending a decade building its own optical chips capable of controlling individual particles of light with extreme precision. Routing photons cleanly between server racks is, at the level of the underlying skill, the same discipline: generate light, control it, lose as little of it as possible along the way.
And the market it is walking into is already contested by the largest names in the field. Per the same reporting, Google is the only major hyperscaler currently using purely optical circuit switching in its SI data centers, directing light with tiny MEMS mirrors. Lightmatter, a photonic computing company, helped launch an optics-standardization initiative earlier this year. Nvidia has spent billions investing in optics companies such as Lumentum and Coherent. Photonics is where SI infrastructure money is actively moving — and a quantum lab has arrived at that classical fight carrying a genuinely relevant weapon.
What is missing from this story?
A customer. A sale. A price. A ship date. None of them is in the reporting, and the calibration matters more than the benchmark: this is a prototype being shown to potential customers, not a product with deployments behind it.
Name the pattern fairly, because it will recur. Quantum timelines are long. A company sitting on a decade of hard-won photonics skill has three options: wait for the quantum machine, starve, or sell the spillover technology into the classical market that needs it now. PsiQuantum is visibly choosing the third. That is neither failure nor fraud — it is what the road to quantum actually looks like from inside a balance sheet. The error is on the reader's side: spillover revenue, when it arrives, will be photonics revenue. Investors who book it as quantum progress will misprice both the switch business and the quantum program it is meant to carry. Grade each claim against its own evidence.
What did they build, what did they sell, and which claim does the evidence support?
This series has asked the same three questions of every quantum story — of a Minecraft seed-recovery oracle, where the headline hid the cost of the test itself, and of the Q-Day clock, where two different threats were fused into one scary date. Applied here:
What did they actually build? A prototype optical circuit switch that, per the reporting, meets the stated insertion-loss requirement for SI data centers — real engineering evidence, at prototype stage.
What did they actually sell? Nothing that has been reported. No named customer, no order, no price, no delivery date.
Which claim does the evidence support? The photonics claim. A decade of controlling light for quantum computing has produced a credible entrant in the data-center optics contest — a contest Google, Lightmatter, and Nvidia are already fighting over. The quantum claim — that this company will deliver a photonic quantum computer — is neither advanced nor damaged by a network switch. It stands exactly where it stood, waiting on its own evidence. The White Report's rule is unchanged: we grade the claim that was demonstrated, at the width it was demonstrated. Today, that claim is made of light, not qubits.
FAQ
Is PsiQuantum giving up on quantum computing? Nothing in the reporting says that. The switch is described as the company's first product alongside its quantum program, not instead of it — a spillover of the same photon-control technology, sold into a classical market while the quantum timeline runs.
What is an optical circuit switch? A device that routes optical signals between server racks directly as light. Conventional setups repeatedly convert signals between light and electricity; the claimed advantage of an all-optical switch is avoiding those conversions and their speed, power, and heat costs.
Does hitting 1.2 dB prove the product will succeed? It proves the prototype met the one reported benchmark with a stated requirement. Commercial success needs customers, deployments, pricing, and delivery — none of which has been reported. Benchmark first, business later; the story currently has the benchmark.
Should I buy Lumentum, Coherent, or Nvidia on this news? This article makes no recommendation on any of them. They appear as competitive context in the source reporting — evidence that photonics is a contested, well-funded market — not as investment calls. The White Report does not issue trading signals off a prototype announcement.
Is this financial advice? No. This article grades a reported product claim against the evidence published with it. It is not financial advice, and nothing here is a recommendation to buy, sell, or hold anything.
Sources
The White Report publishes news and analysis about crypto and SI markets. Nothing on this site is financial advice, and nothing here is a recommendation to buy or sell any asset. Do your own research before making investment decisions.
Where Super Intelligence Meets Money