Equal1’s RacQ: A Fab-Compatible Quantum Sandbox, Not a Quantum Advantage Machine

Quantum finally did something normal. It showed up on a loading dock.

For ten years, if you wanted to touch a qubit you needed a dilution fridge, a PhD, and a room nobody else was allowed in. Now two startups have put it in a 19-inch rack and shipped it like a server.

Last week I tore down SAXON Q’s SXQ128. This week it’s Equal1’s RacQ.

Same form factor. Completely different bet.

Above the fold — if you’re briefing your boss

Equal1’s RacQ is a 6-qubit silicon spin machine that lives in a standard rack, plugs into a 1.6kW outlet, weighs 400kg, and hides a closed-cycle cryocooler holding 0.3 Kelvin inside the box so you don’t have to build a lab around it. It’s built on a normal commercial CMOS process — GlobalFoundries 22FDX — with 35 monolithic quantum cells on a single die. Dell is demoing it integrated with a PowerEdge R770, PowerSwitch networking, and their Quantum Intelligent Orchestrator as a peer accelerator.

It’s not a supercomputer. It’s not going to break RSA. It is the first time you can roll a chip that could actually be made in a real fab into a data center you already own and start learning the hybrid quantum-classical toolchain in days.

Line to steal for the meeting: SAXON got quantum out of the cryostat by changing physics. Equal1 got quantum into the rack by hiding the cryostat.

1. The engineering win is real (and totally different from SAXON)

SAXON solved placement. Equal1 solved integration.

SAXON took NV centers in diamond — notoriously random, like throwing darts blindfolded — and got placement yield from roughly 10% to 85%. That’s a materials win.

Equal1 did something more boring and harder. They took spin qubits, which live as little quantum dots trapping a single electron spin, and made them work in standard silicon.

From their technology page: the element is silicon spin qubits, the state is electron spin up/down, the substrate is silicon, the operating temperature is 0.3K. Quantum dots demonstrated in commercial CMOS, with tunable coupling and charge sensing demonstrated.

Then the packaging trick. Analog control — RF generation and readout — sits microns from the qubits, on-die, in cryogenic CMOS. No cable loss, no latency. Digital control logic handles gate and measurement timing, cycle-accurate, adjacent to the analog. Error correction isn’t bolted on: it’s an embedded ARM core doing on-die ML inference with a detect → decode → correct loop at sub-cycle latency.

Their unit is what they call a quantum tile — qubits, control, and correction in one silicon unit, repeated across the UnityQ die. One tile. Many tiles. UnityQ.

That is the opposite of exotic. That is manufacturable. That is the whole bet.

2. The compute claim is not (and the tax is different)

Here is what they actually list under qubit performance:

• 1Q gate fidelity (average): 99.9% @ 140ns — randomized Clifford
• 2Q gate fidelity (average): 99.3% @ 200ns — interleaved
• Readout fidelity: 99% @ 10µs — parity readout

SAXON’s problem was disconnection. Equal1’s problem is taxation.

SAXON had 128 people in 16 soundproof rooms of 8, with no connection between rooms. You can’t host a 128-person meeting.

Equal1 has 6 people in a single hallway, but only neighbors can talk directly. To entangle qubit 1 and qubit 6, you shuffle the state down the line using SWAPs. Each SWAP is 3 CNOTs. So 4 SWAPs equals 12 CNOTs of tax before your actual gate fires.

So that 99.3%?

• Nearest-neighbor pair: 99.3%
• Long-range pair needing 4 SWAPs: 0.993¹³ ≈ 91.2%
• And your 200ns gate becomes 2.6 microseconds of noise accumulation.

“Full connectivity via SWAP” is true the way “my apartment is beachfront via a 4-hour drive” is true. It’s not free.

Three things that aren’t on the slide:

Threshold vs. affordable

Credit to Equal1 for actually publishing 2Q fidelity — SAXON omitted it entirely. 99.3% is technically above the ~99% surface code threshold everyone quotes. But barely above threshold means your physical-to-logical overhead is brutal. You don’t get error correction, you get a very expensive physics lesson.

Speed is the actual story

140ns / 200ns with 10µs readout is fast. Superconducting is ~30ns, ions are ~50,000ns. Sitting in the middle means you can run 10k–100k variational shots per second. For the hybrid workloads RacQ is designed for, shots-per-second matters more than qubit count.

It’s 6, not 128

You can simulate 6 fully-entangled qubits on your phone. 2⁶ = 64 amplitudes. That’s not a criticism, it’s context. This is not where you run Shor’s.

3. What this is / what this isn’t

What this is: a fab-compatible sandbox. The entire quantum stack — qubits, analog RF, digital timing, ARM cores for error correction — integrated on a CMOS chip you could actually tape out at GlobalFoundries, with the cryogenics hidden in a rack your facilities team can roll in. A way to learn the toolchain that will matter if silicon spin rides the $1T semiconductor learning curve.

What this isn’t: a quantum advantage machine. A PQC threat. An Nvidia replacement. A 128-qubit monolithic processor. It doesn’t give you more qubits — it gives you fewer qubits that could actually be built at scale.

4. SAXON vs. Equal1: for your roadmap

Don’t put them in the same bucket.

5. Does this shift your PQC timeline?

No. Same answer as SAXON.

Breaking RSA-2048 needs thousands of logical qubits. Even with Gidney’s 2025 results bringing estimates from millions down to under a million physical qubits at high fidelity, you are at 6 physical qubits. You are not factoring RSA tomorrow.

Does it shift your infrastructure timeline? Yes.

SAXON proved you can get hands-on time without cryogenics. Equal1 proves you can get hands-on time without a custom supply chain, inside the same rack and orchestrator Dell already sells you.

Who this is for / who it isn’t for

Who it’s for: data center managers who want to test quantum orchestration as a peer accelerator in a standard rack. Developers who live in variational loops and care about shots-per-second more than qubit count. University programs and innovation teams who want silicon without a cryogenics budget or a custom materials lab.

Who it isn’t for: anyone who read “commercially available” and heard “production-ready.” Anyone hunting immediate quantum advantage. And anyone without reinforced flooring and a dedicated 20A circuit — that 400kg and 1.6kW spec is real.

The verdict

SAXON built a deployable sandbox. Equal1 built a fab-compatible sandbox.

It won’t replace Nvidia. It won’t break RSA.

But if you believe quantum advantage will come not from the most exotic physics but from the technology that can ride the $1T semiconductor learning curve, RacQ is the first box you can actually roll into your data center to start learning that toolchain.

For SAXON I said: stop counting qubits, start counting connections.

For Equal1: stop counting qubits, start counting wafers.

At Firebringer Quantum, we stress-test deep tech claims so leadership teams can make smart quantum investments without the hype hangover.