Quantum roadmaps are usually compared on two numbers: gate fidelity and coherence time. Both are real, and neither tells you how many physical qubits a useful machine needs. The number that decides that is connectivity, because error correction is a neighbourhood operation, and a qubit that cannot reach its neighbours needs a chain of swaps to pretend that it can.
A transmon array is lithographed onto a chip, so coupling is fixed by the layout at fabrication time. Nearest-neighbour gates are fast, measured in tens of nanoseconds, but a patch of surface code needs each qubit to touch four others in a regular grid. That grid is achievable, yet any long-range interaction inside the processor becomes a sequence of SWAP gates, and every SWAP injects the very error the code is trying to remove.
The resource that decides time-to-value is physical qubits per logical qubit, and that number is set by connectivity at least as much as by gate fidelity.
Part One: What is actually being compared
To compare superconducting transmon arrays with neutral-atom arrays honestly you first have to name the quantity being optimised. In fault-tolerant qubit layout it is tempting to use the headline specification, but the binding quantity is the one that appears in the bill after a full year of operation: the sum of the committed cost, the per-unit cost and the cost of the occasions when the system behaves badly. Once that is written down, most of the specification-sheet argument simply disappears.
The cost structure behind superconducting transmon arrays is dominated by a fixed commitment that must be paid before the first unit of value appears, while neutral-atom arrays converts that commitment into a variable charge on every unit. That difference is invisible in a per-unit comparison and decisive in a budget, because a fixed cost rewards scale and a variable cost punishes it. Whichever side your volume forecast lands on determines the answer long before the specification sheet does.
Part Two: The arithmetic that decides it
Neutral atoms take the opposite trade. Individual atoms are held in optical tweezers and can be physically rearranged between operations, which makes connectivity almost arbitrary and lets a code patch be reconfigured in software. The price is clock speed: moving and re-cooling atoms takes microseconds, three orders of magnitude slower than a transmon gate. A logical operation that needs a thousand correction rounds therefore finishes in milliseconds on one platform and seconds on the other.
The second-order consequence is organisational. Choosing superconducting transmon arrays creates a dependency on a capability that must be staffed, monitored and kept current; choosing neutral-atom arrays keeps the capability outside the team and turns it into a vendor relationship. Neither is free, and the one that looks cheaper in a spreadsheet is frequently the one that adds a standing meeting to the calendar.
Part Three: Where the simple model breaks
That ratio is not a curiosity; it changes the business case for any algorithm whose value depends on wall-clock time. A chemistry simulation that tolerates a slow logical clock can exploit flexible connectivity and use far fewer physical qubits. A real-time control or trading workload cannot, no matter how good the fidelity looks. The two platforms are optimising different objective functions, and quoting one number from each produces a comparison that does not mean anything.
At 99.9 percent two-qubit fidelity a logical qubit costs roughly 1000 physical qubits, and a 6-fold increase in swap depth can cancel years of fidelity improvement.
Part Four: The cost that arrives later
Scaling also arrives from different directions. Transmon chips hit a wiring and cryostat bottleneck: every extra qubit needs a coaxial line, a microwave chain and a slice of dilution-fridge cooling power. Neutral-atom systems hit a laser and vacuum bottleneck: more tweezers need more optical power and better beam steering, but the control hardware stays outside the chamber. One scaling curve is set by cryogenics, the other by optics and throughput.
Expect a split by scale rather than by industry. Large operators can amortise the fixed side of superconducting transmon arrays and will move first; smaller ones keep neutral-atom arrays because their utilisation is spiky and their tolerance for operational complexity is low. The interesting competition happens in the middle, where a managed service can make either option look like the other.
Part Five: What this changes outside the engineering team
Error correction amplifies both curves. A logical qubit with a target failure rate near one in a trillion needs roughly a thousand physical qubits at 99.9 percent two-qubit fidelity, and that multiplier falls sharply as fidelity improves but rises sharply as connectivity worsens. A six-fold increase in swap depth can erase a fidelity gain that took years of materials work to achieve, which is why layout papers and materials papers have to be read together.
The everyday analogue is a warehouse with fixed shelving versus one with mobile racks. Fixed shelves make each individual pick fast; mobile racks make any two items neighbours at the cost of a slow rearrangement. Which warehouse wins depends entirely on whether the order list is predictable, and a quantum error-correction cycle is just such an order list.
A market sizing sanity check
Public and private spending on quantum hardware is on the order of 8 to 12 billion dollars a year, and control electronics, cryogenics and laser systems already take a larger share than the qubit chips themselves. If fault-tolerant prototypes reach the few-hundred-logical-qubit range by 2031, the enabling control layer plausibly becomes a 6 to 11 billion dollar annual market, split between microwave infrastructure and programmable optics depending on which connectivity model survives.
All figures here are an author's synthesis of published ranges and a stated model, not a measurement of a shipped system. The point of the arithmetic is to locate the crossover and expose which assumption drives it; a vendor quote will differ, and the direction of that difference is more informative than the final number.
The comparison that matters
Transmons buy gate speed and pay for it with fixed topology and a cryogenic wiring wall. Neutral atoms buy reconfigurable connectivity and pay for it with a slow logical clock. Choose by the number of correction rounds your workload needs per second, not by the headline coherence time.
The honest caveat is that this comparison is sensitive to one assumption: that the workload keeps its current shape. Shift the mix toward the case where neutral-atom arrays is strong and the result changes without any technology changing at all. That is why the durable output of an evaluation is not a verdict but the threshold at which the verdict flips.
A machine with a slower logical clock is not a worse machine; it is a machine priced for a different set of algorithms.
SOL: HCqPmPJM3NVtS6YrmVaTZYnuP8XZ2PFgQZKT94VK1LkS
Creating this takes real effort — thank you for the support.
Top comments (0)