Yesterday's headlines all said the same thing: "Russia starts selling quantum computers." Technically true — for the first time, a Russian quantum computing effort isn't offering "cloud access," it's issuing an invoice and shipping hardware. But behind that headline is a much narrower and much more interesting story than "Russia catches up to Google."
What's actually being sold isn't a personal computer, or even a data-center server — it's a research installation the size of a walk-in freezer, priced like an apartment, built for "try real quantum hardware," not "solve a business problem faster than a classical machine." The details of what's in the box matter a lot more than the word "quantum" in the headline.
What's actually for sale
The seller is Quantum Park — a joint center of Bauman Moscow State Technical University and VNIIA (a Rosatom research institute), led by the same team (Ilya Rodionov's group) that shipped Russia's first batch of photonic chips back in July. The product line is called SnowDrop and comes in four configurations of superconducting quantum coprocessors:
| Model | Qubits | Topology | Single-qubit fidelity | Two-qubit fidelity |
|---|---|---|---|---|
| SnowDrop 4QTCS | 4 | T-shaped | up to 99.96% | — |
| SnowDrop 8Q2TCS | 8 | T-shaped | 99.927% | 99.21% |
| SnowDrop 8QGCL | 8 | Grid | 99.9%+ | up to 99.55% |
| SnowDrop 10QGCL | 10 | Grid | 99.9%+ | up to 99.55% |
The platform is marketed as scalable to 48 qubits — that's a roadmap, not what ships today. Every model comes as a complete kit: the quantum chip itself, a cryogenic readout system, a Russian-made ultra-low-temperature cryostat called YARANGA with control signal switching, qubit control electronics, and a software stack for running algorithms. They're not selling a chip — they're selling a turnkey lab.
The price: 170 million rubles, and the questions that go unanswered
Only one number is public: the entry-level platform with the 4-qubit processor starts at 170 million rubles (roughly $1.9M) — comparable, as several outlets pointed out, to a four-bedroom apartment in central Moscow. Pricing for the 8- and 10-qubit versions hasn't been disclosed. Given that more qubits mean tighter cooling and connectivity requirements, a higher price is a safe bet, but I won't put a number on it — there isn't one in any public source.
This isn't really a complaint about secrecy: per-unit price lists are rare for this class of hardware globally. IBM and Google don't publish quantum system pricing either — they negotiate contracts individually. But it does mean there's currently no way to compare "cost per Russian qubit" to "cost per IBM qubit" in any meaningful way.
Fidelity: the numbers are real, the bar is different
The most concrete thing in this release is the fidelity data, and it's genuinely good for a superconducting architecture: single-qubit gate errors under 0.2%, two-qubit gate errors under 1% on the best qubit pairs. For context, the widely cited threshold below which error correction can actually suppress accumulated noise rather than add to it sits around 99% two-qubit fidelity. So the claimed numbers are formally above that threshold — that's a real engineering result, not marketing rounding.
The catch: "operation fidelity" and "useful quantum computation" are different metrics. SnowDrop tops out at 4-10 computational qubits, and any algorithm with real error correction needs orders of magnitude more physical qubits per logical qubit. So "fidelity on par with world-class systems" is an honest claim about individual operations — not a claim that this machine is solving problems classical computers can't.
Where this sits globally
Compared head-to-head on qubit count, it's not close: IBM already has chips with over a thousand physical qubits in its fleet, Google demonstrated the 105-qubit Willow chip below the error-correction threshold back in late 2024, and IonQ has dozens of algorithmic qubits on trapped-ion hardware with a completely different physics and different strengths. Formally, the Russian lineup is two orders of magnitude behind at launch.
But the qubit-count race isn't the only metric, and it's often not the most honest one — D-Wave has thousands of qubits, but that's an annealing architecture for a narrow class of optimization problems, not universal gate-based computing like SnowDrop. Realistically, the Russian team isn't competing with 2026-vintage IBM and Google flagships — it's roughly where those companies were five to seven years ago: "the first reliably working, non-one-off superconducting gate-model computer in the country." For the domestic market, that's a real milestone. For the global race, it's an entry, not a claim to leadership.
Who this is actually for
The buyer list isn't hidden: universities, research centers, and companies that need infrastructure to develop and test quantum algorithms — not consumer or even typical enterprise users. Before this announcement, the only way to access a Russian quantum coprocessor legally was through the Bauman Octillion cloud platform: since July 2025, it's run over 103,000 hybrid quantum-classical algorithms, with more than 9,000 cumulative hours of uptime. Selling hardware is the next step for anyone who's outgrown queue time on a shared cluster and needs their own installation for continuous experimentation.
Bauman's rector, Mikhail Gordin, framed the project's logic this way: an engineering university's job isn't just to create technology, but to bring it to practical use. After further equipping Quantum Park, the team claims a production capacity of 10-20 multi-qubit installations per year — not mass manufacturing in the usual sense, but hand-built units for specific customers, which is normal for hardware at this level.
The takeaway: an infrastructure story, not a consumer one
Strip away the "Russia vs. the world" headline effect, and what's left is a modest but genuinely interesting story: the country now has a reproducible product, tested across 103,000 runs, with an actual price tag on at least the base version — not a one-off prototype in a single lab. For a university or R&D team with 170+ million rubles and a real need to run quantum algorithms on dedicated hardware instead of queuing for cloud time, this is a reasonable purchase. For the market broadly, it's a signal that after years of cloud-only access, Russia's quantum program has moved into a phase where it's willing to sell a physical product, not just demo prototypes.
I wouldn't read this as "a quantum laptop in five years," though. The path from 10 research qubits to something that actually beats a classical computer on a real business problem is measured in decades worldwide, and Russia isn't an exception here.
Originally published at klukyanov.ru.
Shorter weekly write-ups (in Russian) — on Telegram.
Top comments (0)