Public conversation about quantum computing tends to focus on speed, the idea that a quantum computer will simply outrun a classical one at certain problems. That framing isn’t entirely wrong, but it skips over the actual bottleneck that has occupied most quantum computing researchers for the past decade: errors. Qubits, the basic unit of quantum information, are extraordinarily fragile, and the difficulty of keeping them stable long enough to do useful, reliable computation is the central obstacle standing between today’s quantum machines and genuinely practical applications.
Why Qubits Are So Error-Prone
Classical computer bits are stable by comparison, a transistor reliably holds a 0 or a 1 state, and error rates in classical computing are low enough that they rarely factor into everyday use. Qubits, by contrast, rely on quantum states that are inherently delicate, sensitive to tiny fluctuations in temperature, electromagnetic interference, and even the physical vibrations of the equipment housing them. This sensitivity, ironically, is also what gives qubits their computational power, but it comes at the cost of a phenomenon called decoherence, where a qubit loses its quantum state and the information it was holding becomes unreliable, often within fractions of a second.
Every operation performed on a qubit also carries a chance of introducing an error, and because quantum algorithms typically require many operations chained together, these small per-operation error rates compound quickly. A qubit with a seemingly small individual error rate can still make a long computation useless if the errors accumulate faster than they can be identified and corrected.
Why This Is Harder to Fix Than It Sounds
The standard approach to this problem is quantum error correction, which involves encoding one logical, reliable qubit using many physical qubits working together, so that errors in individual physical qubits can be detected and corrected without disturbing the overall computation. This works in principle and has been demonstrated in research settings, but the practical cost is steep. Current error correction schemes can require somewhere in the range of dozens to over a thousand physical qubits to create a single reliable logical qubit, depending on the error correction code and error rates involved.
- Physical qubits currently need extensive redundancy to produce a single reliable logical qubit
- Error rates compound across the many operations required for meaningful quantum algorithms
- Environmental sensitivity means qubits require extreme isolation, often near absolute zero temperatures
This means that the widely cited qubit counts companies use to market their quantum processors, hundreds or a few thousand qubits, are largely physical qubits, not the smaller number of stable logical qubits that would actually be usable for a complex, error-corrected computation. The gap between physical and logical qubit counts is one of the more consistently misunderstood aspects of quantum computing coverage in the media.
What Progress Actually Looks Like Right Now
Researchers have made real, incremental progress on error correction and qubit stability in recent years, including demonstrations of error correction schemes that reduce error rates as more physical qubits are added, a milestone that wasn’t reliably achievable a few years earlier. Companies working on different qubit technologies, superconducting circuits, trapped ions, photonic qubits, are each pursuing somewhat different trade-offs between qubit stability, operation speed, and scalability, and it’s not yet clear which underlying approach will end up being the most practical for building large, error-corrected quantum computers.
Why This Matters More Than Raw Speed Claims
Quantum computers have already demonstrated an ability to outperform classical computers on very narrow, specifically designed benchmark tasks, but these demonstrations don’t translate into practical advantages for the kinds of real-world problems, drug discovery, materials science, cryptography, that quantum computing is ultimately expected to help with. Those applications require long, complex, error-corrected computations that current hardware simply can’t sustain reliably yet. Progress toward that point depends far more on reducing error rates and improving qubit stability than on any raw increase in qubit count or processing speed, which is why researchers in the field tend to talk about error correction milestones far more than the speed comparisons that dominate popular coverage.
The timeline for when error rates will be low enough to support broadly useful quantum applications remains genuinely uncertain, with estimates from serious researchers ranging from several years to a couple of decades depending on the application and the pace of continued error correction breakthroughs. What’s clear is that the qubit count on a marketing slide says very little about how close a given machine actually is to doing something practically useful, and the more meaningful number, if a comparison is really needed, is the error rate behind it.