Build it from superconducting circuits or from semiconductor spins, it makes little difference. These units are delicate. The outside world perturbs them, and they suffer bit-flips and phase-flips that collapseThe instant a measurement forces a qubit from a blend of possibilities into one definite classical value. their computational state long before a single useful algorithm finishes. Here is the honest read: no physical system anyone has built is good enough to serve, on its own, as a reliable building block.1
So the thinking has to move away from the single particle and toward the logical qubitA reliable 'qubit' built by bundling many error-prone physical qubits together with error correction. These are the units that actually matter for useful computing.. Instead of trusting one fragile unit, we build a collective thing out of a large array of physical qubits. Through an architecture called the surface codeThe leading error-correction scheme for superconducting qubits. It needs many physical qubits to protect one logical qubit., those physical bits work together to detect and manage their own errors.2 What comes out is a single logical unit that performs better than the parts it is made of. Notice what does the work here. It is not better physics. It is the surprising math of topological tricks and an architecture built, as I will show, out of nothingness.
Figure 1: The overhead pyramid. Thousands of physical qubits hold up a single logical one, roughly 14,500 to 1 for the 2,000-bit factoring problem.
The 14,500-to-1 Ratio: Strength in Massive Numbers
In the surface codeThe leading error-correction scheme for superconducting qubits. It needs many physical qubits to protect one logical qubit., distance d is the minimum number of physical operations it takes to define a logical operator. It is the size of our redundancy. Push d up, and the logical error rate falls exponentially. That is the good news. The cost is steep. For the machine to be reasonably fault-tolerant, the path stops being about better qubitsThe basic unit of a quantum computer. Like a 'bit' in a normal computer, but instead of being only 0 or 1 it can be 0, 1, or a blend of both at once. and becomes about massive arrays of them.1
A reasonably fault-tolerant logical qubit that can be used effectively in a surface code takes of order 103 to 104 physical qubits.
For the 2,000-bit factoring problem in particular, the numbers point to an overhead of something like 14,500 physical qubits for every one logical qubit.2 The hard part of this problem has moved. It used to live on the laboratory bench. It now lives on the architectural blueprint.
Figure 2: The distance ladder. Pushing the code distance d up drives the logical error rate down exponentially, as long as the physical error rate stays below the threshold.
The 1% Revolution: Why “Good Enough” Is the New Perfect
The surface codeThe leading error-correction scheme for superconducting qubits. It needs many physical qubits to protect one logical qubit. has won the architectural race for one reason: it forgives. Most other error-correction schemes, the Steane code or the Bacon-Shor code among them, fall apart the moment the error rate climbs past 2 × 10−5, which is 0.002 percent.5 The surface code holds together up to an error threshold of roughly 1 percent.2, 6
That is three orders of magnitude, and it is the whole reason the surface code is the only realistic route to a large machine. It works by holding the array in a quiescent state, a simultaneous eigenstate of all the stabilizers across a two-dimensional grid. It does not chase every flicker with a physical correction. It detects instead. It finds the bit-flips and phase-flips and tracks them in classical software. We do not have to repair the hardware in real time. We only have to know exactly how it broke, so we can account for the break when we read out the final measurementReading a quantum system, which forces it out of its blend of possibilities into a single definite 0 or 1 and ends its quantum behaviour..3
Figure 3: Computing with holes. A logical CNOT is performed by braiding one kind of hole in a loop around the other, so the logic is fixed by the path rather than the hardware.
Computing with “Holes”: The Architecture of Nothingness
There are two kinds of hole, and the difference is simply which stabilizer you switch off:
Hole type
How it is made
Z-cut (smooth)
Turn off the measure-Z stabilizers.
X-cut (rough)
Turn off the measure-X stabilizers.
Inside these holes we define the logical operators. The XL operator is not one pulse. It is a chain of physical X operators stretched across the lattice. The ZL operator is the same idea, a chain of physical Z operators. So the information does not live in the presence of a signal. It lives in the structured absence of the lattice itself.2
Classical control software runs the dance and keeps the books on what each move means mathematically. Because the logic is fixed by the topology of the path, by the plain fact that one hole went around another, the system shrugs off small local bumps. A local error can flicker all it likes. It cannot undo the fact that the braid happened. That is what topological protection means in practice: the burden of precision moves off the physical hardware and onto the software tracking the paths.
The Billion-Qubit Reality Check: The Scale of Shor’s Algorithm
Here is the resource breakdown for that one calculation:
Component
Requirement
Execution time
26.7 hours
Computational logical qubits
~4,000
Physical qubits for logic
~58 million
Total physical qubits (footprint)
~1 billion
Where does the gap come from? Almost none of those billion qubitsThe basic unit of a quantum computer. Like a 'bit' in a normal computer, but instead of being only 0 or 1 it can be 0, 1, or a blend of both at once. are thinking. They are working in a distillation factory. To perform a Toffoli gateA three-qubit gate that flips a target qubit only when two control qubits are both 1. Costly to build from simpler gates, so it is a common stress test of a machine., the machine consumes particular quantum statesThe full description of a quantum system's condition at a moment in time, such as whether a qubit is 0, 1, or a mix., the ancilla states written |AL⟩. These states are born dirty and noisy. We have to spend large stretches of the lattice purifying, or distilling, them until they are clean enough to use in logic. The billion-qubit footprint is the bill for that purification, run at industrial scale.2, 8
Conclusion: The Future Is a Managed Lattice
The logical qubit is not an object you can hold. The thing it is, is a statistical consensus among thousands of physical bits, an equilibrium managed inside a noisy room. As the first useful quantum computers come into view, I think we have to accept what is happening underneath: the era of pure physics is giving way to the era of systems engineering.4
A logical qubit is a statistical consensus among thousands of physical bits.
The breakthrough will not arrive as one perfect material. It will arrive as the ability to manage enormous lattices of good-enough components. The question is no longer how to stop the noise. It is how to build an architecture clever enough to compute straight through it. So the real question for the coming decade is a plain one. Can we build a machine with a billion parts that behaves as a single, silent whole?
Devitt, S. J., Munro, W. J. & Nemoto, K. “Quantum Error Correction for Beginners.” Reports on Progress in Physics76, 076001 (2013). DOI: 10.1088/0034-4885/76/7/076001. Preprint: arXiv:0905.2794.
Fowler, A. G., Mariantoni, M., Martinis, J. M. & Cleland, A. N. “Surface Codes: Towards Practical Large-Scale Quantum Computation.” Physical Review A86, 032324 (2012). DOI: 10.1103/PhysRevA.86.032324. Preprint: arXiv:1208.0928.
Terhal, B. M. “Quantum Error Correction for Quantum Memories.” Reviews of Modern Physics87, 307–346 (2015). DOI: 10.1103/RevModPhys.87.307.
Google Quantum AI and Collaborators. “Quantum Error Correction Below the Surface Code Threshold.” Nature638, 920–926 (2025). DOI: 10.1038/s41586-024-08449-y. Preprint: arXiv:2408.13687.
Cross, A. W., DiVincenzo, D. P. & Terhal, B. M. “A Comparative Code Study for Quantum Fault Tolerance.” Quantum Information & Computation9, 541–572 (2009). Preprint: arXiv:0711.1556.
Raussendorf, R. & Harrington, J. “Fault-Tolerant Quantum Computation with High Threshold in Two Dimensions.” Physical Review Letters98, 190504 (2007). DOI: 10.1103/PhysRevLett.98.190504. Preprint: arXiv:quant-ph/0610082.
Shor, P. W. “Polynomial-Time Algorithms for Prime Factorization and Discrete Logarithms on a Quantum Computer.” SIAM Journal on Computing26, 1484–1509 (1997). DOI: 10.1137/S0097539795293172. Preprint: arXiv:quant-ph/9508027.
Bravyi, S. & Kitaev, A. “Universal Quantum Computation with Ideal Clifford Gates and Noisy Ancillas.” Physical Review A71, 022316 (2005). DOI: 10.1103/PhysRevA.71.022316. Preprint: arXiv:quant-ph/0403025.
