Quantum computing has a dirty secret that the industry would prefer you not dwell on. Every qubit you have ever read about in a press release is a physical qubit. And physical qubits are unreliable. They lose their quantum state constantly. They produce errors at rates that make real computation impossible. The entire field has spent decades trying to solve a single problem: how do you build qubits that actually work long enough to be useful?
Quantinuum just posted the most convincing answer anyone has produced so far.
In a paper published on arXiv in late February, Quantinuum researchers demonstrated quantum computations using up to 94 error-protected logical qubits on the company’s 98-qubit Helios trapped-ion processor. The logical gate error rates hit roughly one error in 10,000 operations, significantly lower than the raw error rates of the physical hardware underneath. They also ran 48 fully error-corrected logical qubits using a more robust encoding scheme. Both results cleared “beyond break-even,” meaning the error protection actually improved computation accuracy instead of degrading it.
That last part is the important part. For years, adding error correction to quantum computers made them worse. The overhead of monitoring and fixing errors consumed more resources than it saved. Beyond break-even is the threshold where the protection starts paying for itself. Quantinuum crossed it with 94 qubits. Nobody else has done that at this scale.
How They Did It
The technique relies on something called iceberg codes, named because most of the structure sits below the surface. In the simplest version, just two extra monitoring qubits watch over the entire logical system. That is absurdly efficient. Traditional quantum error correction codes require massive overhead, sometimes 1,000 physical qubits per logical qubit. Iceberg codes achieved a physical-to-logical ratio near 1:1 for error detection and roughly 2:1 for full error correction.
The reason Quantinuum can pull this off is their trapped-ion architecture. Unlike superconducting qubits (Google, IBM), trapped ions can connect to any other ion in the processor. That all-to-all connectivity lets you implement error correction schemes that would be physically impossible on a chip where qubits can only talk to their neighbors. It is a genuine architectural advantage, not marketing.
To prove the system works on something real, the team ran a quantum simulation of the three-dimensional XY model of quantum magnetism using 64 error-detected logical qubits. They also generated a 94-logical-qubit GHZ entangled state with 94.9% fidelity. These are not toy demonstrations. They are the kind of computations that will eventually underpin drug discovery, materials science, and cryptographic analysis when the qubit counts get large enough.
Where This Sits in the Race
Quantinuum is not the only company making progress on error correction. Google’s Willow chip, announced in December 2024, demonstrated that errors could be reduced exponentially as qubit counts increased, a result the field calls going “below threshold.” That was a milestone. But Willow is a 105-qubit superconducting chip. Google showed the principle works. Quantinuum showed it works on computations with 94 logical qubits that outperform the raw hardware. Those are different claims.
IBM has published its own fault-tolerant roadmap centered on the Quantum Starling system, targeted for 2029, which would feature 200 logical qubits running 100 million error-corrected operations. That is ambitious. It is also three years away.
IonQ, Quantinuum’s closest competitor in trapped-ion technology, is working toward 256-qubit systems in 2026 and has posted strong qubit performance numbers from R&D prototypes. But IonQ has not published a comparable logical qubit demonstration at this scale. The company also carries baggage from a Wolfpack Research short report alleging that most of its revenue came from Pentagon contracts that were subsequently canceled. IonQ disputes the allegations. The uncertainty remains.
The competitive picture is messy, but one pattern is clear. The companies that can demonstrate fault-tolerant operations (not just promise them on a roadmap) are separating from the pack. Quantinuum just put up the strongest logical qubit numbers in the industry.
Follow the Money
None of this exists in a vacuum. Quantinuum filed a confidential S-1 with the SEC on January 14, beginning the formal process for what would be the first traditional quantum computing IPO. Every other quantum company that went public (IonQ, D-Wave, Rigetti) used SPAC mergers, which let them project future revenue without the scrutiny of a full SEC registration. Quantinuum is betting its financials can withstand that scrutiny.
The last private round valued the company at $10 billion, with a $600 million raise drawing in NVIDIA NVentures, JPMorgan, and Amgen. The IPO is expected to target $20 billion and raise roughly $1 billion. Honeywell retains a 54% majority stake and provides the manufacturing backbone that no pure-play startup can replicate.
The timing of this research paper is not accidental. You publish your best results when you need the market to pay attention. An IPO roadshow goes better when you can say “we just ran 94 logical qubits beyond break-even” than when your last headline is six months old.
Governments are paying attention too. The UK just committed £2 billion to quantum computing procurement and scale-up, targeting operational infrastructure by the early 2030s. Quantinuum, formed from the merger of Honeywell Quantum Solutions and Cambridge Quantum, has deep UK roots. That £2 billion is not earmarked for Quantinuum specifically, but the company is positioned to capture a significant share of it.
What It Actually Means
Ninety-four protected logical qubits is a genuine engineering milestone. It is also not commercially useful scale. Real applications in cryptography, pharmaceutical modeling, and financial simulation need error-corrected systems orders of magnitude larger. The qubit count needs to grow. The error rates need to drop further. The cost per operation needs to come down.
But the math is starting to work. For the first time, encoding qubits in error-protecting schemes makes computation better instead of worse, at a scale that is not trivial. That is the inflection point the entire field has been waiting for.
Whether $20 billion is the right price for that inflection point depends on how fast the next steps come. Quantinuum says it will deliver universal, fully fault-tolerant quantum computing by 2029. If they are right, this valuation looks cheap. If the timeline slips, investor patience will be tested against Honeywell’s willingness to keep writing checks.
Ninety-four logical qubits. One error in 10,000 operations. The physics is cooperating. Now comes the hard part: turning physics into a product someone will pay for.
—
For inquiries and analysis contact laterstack@proton.me