The quantum computing press release cycle has a consistent structure: company announces qubit count milestone, stock moves, academics note the number is physically meaningless without error rates, everyone moves on. We've been watching this loop for a while. The interesting question — the one worth spending analytical energy on — isn't how many physical qubits a system has. It's what the logical qubit overhead looks like and whether anyone has a credible path to fault-tolerant operation below 1,000 physical qubits per logical qubit.
That number matters because it determines when quantum computing escapes the regime where it can only run toy benchmarks and enters the regime where it solves commercially valuable problems. Current best-in-class hardware requires between 1,000 and 10,000 physical qubits to encode a single logical qubit with enough protection to run non-trivial circuits. At 10,000:1 overhead, you need 10 million physical qubits to solve a 1,000-qubit problem. Nobody has 10 million physical qubits. Nobody will have them for a long time.
Surface Codes: The Standard and Its Constraints
Surface codes are the dominant error correction framework for superconducting qubit systems because their error threshold — roughly 1% per gate operation — is achievable with current hardware. The math is beautiful. Qubits are arranged on a 2D lattice, and you measure stabilizer operators that detect errors without collapsing the logical state. When an error occurs, the decoder infers what happened and applies a correction.
The catch is locality. Surface codes work because you only need to measure interactions between neighboring qubits on the lattice. That maps well onto superconducting planar chip geometries where coupling is physically limited to adjacent qubits. The problem is that local decoding under a realistic error model with correlated noise is harder than the idealized version, and every additional microsecond of decoder latency shows up as an effective increase in the logical error rate.
The classical processing layer underneath a fault-tolerant quantum computer may need to make error correction decisions in under 1 microsecond. That's a real-time computing problem that most quantum hardware companies haven't fully reckoned with yet.
Decoder hardware is one of the cleanest investment themes in quantum computing that almost nobody is looking at. A fast decoder based on minimum-weight perfect matching (MWPM) running on FPGAs or custom ASICs is a prerequisite for any fault-tolerant system — and it's almost entirely decoupled from the qubit modality. Whether the underlying hardware uses superconducting transmons, trapped ions, or photonic qubits, you need classical control electronics that can keep up. The companies building that infrastructure will matter regardless of which qubit technology wins.
What the Physical Error Rate Has to Be
A surface code with code distance d (meaning a d×d lattice of physical qubits per logical qubit) suppresses errors exponentially in d — but only if the physical error rate is below the threshold. Below threshold, every increase in code distance gives you exponentially better logical error rates. Above threshold, scaling up makes things worse.
The threshold for depolarizing noise in a surface code is around 1%. Real hardware has correlated errors, leakage, and crosstalk that the idealized model doesn't include. In practice, you want physical gate fidelities above 99.5% — ideally 99.9% — before you start worrying about logical qubit overhead ratios. Several superconducting systems are now demonstrating two-qubit gate fidelities in the 99.5-99.9% range on small chips with 10-100 qubits.
The problem is that those fidelity numbers typically degrade as you add more qubits to the chip. Crosstalk between qubits, frequency crowding, and substrate losses all get worse with scale. The companies we find interesting are those with systematic solutions to cross-talk — whether through qubit frequency allocation algorithms, improved substrate materials, or 3D packaging geometries that reduce parasitic coupling. Some of these are EDA problems, some are materials problems, some are packaging problems. None of them are simple.
Trapped Ions and the Connectivity Advantage
Superconducting qubits dominate the hardware conversation because of IBM and Google's scale. But trapped ion systems have a connectivity advantage that changes the error correction arithmetic significantly. Ion trap qubits can be made to interact with any other qubit in the trap, not just nearest neighbors. That connectivity enables more efficient codes — specifically, low-density parity check (LDPC) codes — that in principle achieve similar logical error rates with lower physical qubit overhead than surface codes.
The practical tradeoff is speed. Ion-ion gate operations take microseconds; superconducting gates take nanoseconds. For error correction to work, you need to complete stabilizer measurement cycles faster than the decoherence time of your physical qubits. Ion traps have decoherence times measured in seconds, which gives them a large T2/gate time ratio. The physics is different but the systems engineering challenge — building chips and control hardware that scale — is comparably hard.
We track a half-dozen companies building trapped ion systems, two doing photonic quantum computing, and the superconducting field closely. Our investment thesis here is less about picking the winning qubit modality and more about finding companies that solve the shared infrastructure problems: decoder hardware, cryogenic RF control electronics, and calibration software that keeps multi-qubit systems operating near their theoretical fidelity limits over long runs.
The Commercial Timeline Is Not What You've Been Told
Our data suggests fault-tolerant quantum advantage on commercially relevant optimization problems is 7-12 years away, not 2-3. The press releases are optimistic because there's fundraising happening. That's fine — it's how the industry works. But as investors, we're designing our portfolio for the realistic timeline, not the aspirational one. The companies we back now need businesses that survive the interim period: quantum-classical hybrid algorithms, simulation tools, and calibration software that has value even on noisy intermediate-scale quantum (NISQ) hardware.
QuantVault, our Boston-based portfolio company, exemplifies this approach. Their error-corrected qubit arrays are designed for financial modeling use cases where approximate answers computed faster than classical methods have economic value even before full fault-tolerance is achieved. The target isn't perfect — it's better than the alternative on specific problem classes. That's a much cleaner commercial proposition than "we'll have a universal quantum computer in three years."
Building quantum hardware, control systems, or quantum-classical software? We want to understand what you're working on. Get in touch.
