Quantum computers have been called "the technology of the future" for decades, but in the last couple of years strong signs have emerged that this future is now bumping up against a concrete threshold. That threshold has a name: error correction. The physical fragility that keeps a quantum bit (qubit) from behaving like a reliable classical bit has long been the field's deepest problem — and results announced back-to-back by Google and IBM now show that this problem has become mathematically manageable.
Why Is Qubit Fragility Such a Big Problem?
A classical computer bit is either 0 or 1, and keeping it that way is relatively easy. A qubit, by contrast, exists in a state called superposition — carrying both a 0 and a 1 possibility at once — and this fragile state can be disturbed by the smallest vibration, temperature fluctuation, or electromagnetic noise in its environment. This is called decoherence. Keeping a single qubit stable long enough to actually compute something has been quantum computing's central engineering problem.
Physically driving a single qubit's error rate down to zero isn't possible. So since the mid-1990s, researchers have pursued a different strategy: combine many "physical" qubits so that their joint state forms a single "logical" qubit that is far more resistant to error. This is known as quantum error correction (QEC).
The "Threshold": Why Does a Magic Boundary Matter?
QEC theory rests on a simple idea: if the physical qubits' error rate is below a certain threshold, adding more qubits to the system reduces the overall error exponentially. But if the error rate sits above that threshold, the opposite happens — adding more qubits makes things worse, because every new qubit brings its own share of error into the system.
That's why "getting below threshold" became something close to a holy grail in the quantum computing community. It had been known theoretically since 1995, but no hardware had ever demonstrated it conclusively in an experiment — until Google's Willow chip.
Google's Willow Chip: A Turning Point
Google's researchers grew the qubit array on Willow step by step: first a 3×3 grid, then 5×5, then 7×7. At each step, the expectation was clear: if the system was genuinely below threshold, the logical error rate should roughly halve with each growth step. According to the results published in Nature, that's exactly what happened — each time the code distance increased by two, the logical error rate dropped by a factor of about 2.14.
In the largest configuration (a 101-qubit, distance-7 surface code), the logical error rate per cycle fell to about 0.143%. Even more significantly, this logical qubit's "memory lifetime" exceeded that of the single best physical qubit in the system by a factor of roughly 2.4 — meaning error correction produced a result that outlived the very components it was protecting. This is what the field calls "breakeven," a threshold that had eluded researchers for years.
What Does "Surface Code" Actually Mean?
The method used in the Willow experiment is a scheme called the surface code. The idea: physical qubits are arranged in a checkerboard-like grid; some carry the actual data, while others interspersed among them continuously "measure" their neighbors to check for errors. Because these measurements can be made without disturbing the underlying information, the system can correct errors without collapsing the computation. As the grid grows (from 3×3 to 7×7), each physical qubit gains more neighbors and therefore more error-detection power — and what Google demonstrated is that this scaling behaves exactly as theory predicted.
IBM's Roadmap: From Error Correction to Quantum Advantage
IBM is tackling the problem from a slightly different angle: not just reducing errors, but building a hardware-software pipeline that can detect and correct them in real time. According to the company's published roadmap, it is working on a 120-qubit Nighthawk processor paired with a real-time error-correction decoder; the speedup targeted for that decoder was reportedly achieved a full year ahead of schedule.
The Loon Architecture and the Connectivity Problem
For error-correcting codes to work efficiently, qubits need to be connected to each other not just strongly but flexibly. IBM's Loon chip architecture links qubits across the chip using connectors called "c-couplers," allowing each qubit to interact with up to six different neighbors. This is regarded as a critical engineering step toward making scalable error-correcting codes practically implementable. IBM states that, building on these steps, it aims to demonstrate a verified "quantum advantage" by the end of the year.
Why Does This Matter?
This progress in error correction means the biggest obstacle standing between quantum computers and practical tools — rather than laboratory curiosities — is starting to be cleared. The quantum advantage anticipated in fields ranging from molecular simulation for drug discovery, to designing new materials, to cryptography and optimization problems, all require a large number of reliably operating logical qubits. Willow's demonstrated below-threshold behavior and IBM's real-time correction targets are seen as the first concrete evidence that this logical-qubit count will keep climbing in the years ahead.
There are, of course, still major engineering hurdles to clear — chief among them holding thousands of physical qubits together coherently and keeping manufacturing costs reasonable at that scale. Even today's most advanced systems turn hundreds of physical qubits into only a few dozen reliable logical ones, while practical applications are discussed in terms of thousands or even millions of logical qubits. Still, this marks a shift from a question that was debated a decade ago as "is this even possible" to one now framed as "how fast can this scale."
Competition across the industry appears to be accelerating this progress as well: alongside Google and IBM, other companies are pursuing similar thresholds with their own qubit architectures, from superconducting circuits to trapped ions. Which architecture will ultimately win out isn't clear yet, but the fact that error correction has moved from a theoretical debate to a measurable engineering metric marks a critical turning point in the field's maturation.