The single milestone that decides when quantum computing becomes practical has a dry name and enormous stakes: fault tolerance. Here is what it means and why the race to reach it is heating up.
Today's quantum computers are real, impressive, and error prone. Their qubits are fragile, and small mistakes creep into every calculation. For now, that limits them to short tasks and careful demonstrations. The whole field is organized around fixing this, and the finish line has a name. Fault tolerance is the point at which a quantum computer can correct its own errors faster than they appear, so it can run long, useful programs reliably.
- Quantum errors are the main barrier, not the number of qubits alone.
- Error correction bundles many physical qubits into one stable logical qubit.
- The leading roadmaps now target fault-tolerant machines before the end of the decade.
Why errors are the real problem
It is tempting to track progress by counting qubits, the way we once counted megahertz. That misses the point. A machine with thousands of noisy qubits can still be useless if errors pile up before it finishes. Quality matters more than raw quantity. The goal is not just more qubits, but qubits stable enough, and connected well enough, to trust. This is why a smaller, cleaner machine can outperform a larger, noisier one.
The idea of a logical qubit
The fix is clever. Instead of trusting any single fragile qubit, you spread one unit of information across many physical qubits and use the group to constantly check and repair itself. The result is called a logical qubit, and it is far more stable than any of its parts. The cost is steep, since each logical qubit can take many physical ones, but the principle works. It turns an unreliable component into a dependable one, which is exactly the trade computing has made before in other forms.
The signal that changed the mood
For years, error correction was mostly theory. That has shifted. Google's Willow processor, unveiled at the end of 2024, showed something the field had been waiting for: as the team made the logical qubit larger, the error rate went down rather than up. Crossing that threshold, where bigger means better instead of noisier, is the proof of principle that error correction scales. IBM, meanwhile, has published a detailed roadmap aimed at a fault-tolerant machine around the end of the decade. When serious companies put dates on paper, it usually means the remaining problems look like engineering, not mystery.
Who is in the race
The contest is broad and well funded, which is healthy. Different groups are betting on different kinds of qubits, from superconducting circuits to trapped ions to neutral atoms and photons, and each approach has real strengths. No one knows yet which will win, or whether several will coexist for different jobs. What matters for everyone watching is that the competition is pushing quality, stability, and error correction forward on every front at once.
What to do while you wait
Fault tolerance is not here yet, and that is fine. The right posture is neither hype nor dismissal. Watch the error rates and the roadmaps, not the qubit counts. Learn which of your problems are the kind quantum will help with, so you recognize the moment when it arrives. And build relationships with the platforms doing the work now, because access and know-how compound. The companies that treat the run-up to fault tolerance as preparation, rather than spectator sport, will be ready the day the machines are.
The overhead problem
The reason fault tolerance is hard comes down to arithmetic. Because individual physical qubits are so error-prone, error correction works by spreading the information of one reliable logical qubit across many physical ones, which constantly check and repair each other. The catch is the ratio. Depending on the quality of the underlying hardware and the cleverness of the error-correcting code, it can take dozens, hundreds, or even more physical qubits to produce a single dependable logical one.
This is why the race to fault tolerance is really two races at once. One is to build more physical qubits. The other, just as important, is to reduce how many physical qubits each logical qubit requires, by improving hardware quality and inventing more efficient codes. A breakthrough that lowers the overhead can be worth as much as a breakthrough that adds qubits, because it brings the same goal closer with less hardware.
It also explains why the companies that lead on fidelity have an advantage in this race. Better physical qubits mean lower overhead, which means fewer of them are needed to reach a useful number of logical qubits. Quality and quantity work together, and progress on either front moves the finish line closer.
Different roads to the same goal
The contest to reach fault tolerance is being run on several tracks at once, with different companies betting on different kinds of qubits, and that diversity is a strength for the field. Superconducting approaches bring speed and manufacturing maturity. Trapped ions and neutral atoms bring exceptional quality and flexible connectivity that suits error correction well. Photonic approaches bring room-temperature operation and natural networkability.
Each of these paths has produced real milestones, and no one yet knows which will reach large-scale fault tolerance first, or whether several will arrive together for different uses. What matters for everyone watching is that the competition is pushing quality, error-correction techniques, and scale forward on every front simultaneously. A breakthrough on any track tends to lift the whole field, because the underlying science is shared.
This is why the recent demonstrations that error correction actually improves as a machine scales were so significant across the board. They did not just validate one company's approach. They confirmed that the central premise of the entire endeavor is sound, which is the kind of result that energizes every team chasing the goal.
What changes when we get there
It is worth picturing what fault tolerance enables, because that is the reason the effort commands such investment. A fault-tolerant quantum computer could simulate molecules and chemical reactions directly, accelerating the discovery of new medicines, better batteries, more efficient fertilizers, and novel materials, all designed from first principles rather than slow trial and error. It could solve optimization problems of a scale that defeats today's best methods, with applications across logistics, energy, and finance.
These are not incremental improvements. They are the kind of capabilities that could reshape entire industries, which is why governments and companies are willing to pour resources into a goal that remains a few years out. The promise of fault tolerance is the promise of quantum computing itself, finally made reliable enough to deliver.
The arrival will not be a single dramatic moment but a gradual crossing, as machines accumulate enough logical qubits to tackle steadily more valuable problems. The first useful fault-tolerant applications will likely be narrow and specialized, expanding outward as the machines grow. But each step past the threshold opens doors that were firmly closed before.
Reading the signposts
For anyone tracking the field, the right things to watch are not the splashiest qubit-count announcements but the markers of genuine progress toward reliability. Watch for demonstrations that error rates fall as logical qubits grow larger. Watch for the published roadmaps from serious players, and whether they are met on schedule. Watch for the steady improvement in fidelity and the shrinking overhead of error correction.
These signposts tell you whether the field is on track, and increasingly they point in an encouraging direction. The companies setting firm dates for fault-tolerant machines, and hitting their interim milestones along the way, are giving the world a credible timeline to plan against. The shift from vague someday to concrete schedule is itself one of the clearest signs of how far the field has come.
The practical takeaway is the same as ever. Fault tolerance is not here yet, but it is no longer a distant dream, and the prudent move is to understand the milestones, follow the right metrics, and be ready to act when the machines cross the threshold in your field. The race is real, it is being run by serious people, and the finish line is finally coming into view.
The economics of getting there
Fault tolerance is not only a scientific milestone. It is an economic turning point, because it is the moment quantum computing crosses from interesting to indispensable for the problems it suits. Before fault tolerance, a quantum computer can demonstrate advantages on narrow, carefully chosen tasks. After it, the machine can run long, valuable programs reliably, and the calculus for businesses changes entirely.
That is why so much capital is flowing into the field now, well ahead of the payoff. Companies and governments understand that the organizations positioned at the moment of arrival, with the relationships, the expertise, and the use cases already in place, will capture the early value, while latecomers spend years catching up. The investment is a bet on being ready, not on the technology being finished.
It also explains the intensity of the competition. The first players to deliver fault-tolerant machines will set standards, attract customers, and build the ecosystems that tend to entrench early leaders. The prize is not merely scientific credit but a position at the foundation of a new kind of computing, which is worth a great deal.
What to do before it arrives
For business leaders, the right response to the race for fault tolerance is neither to wait passively nor to overcommit prematurely. It is to prepare. That means building a basic literacy in what quantum can and cannot do, identifying which of your hardest problems, in chemistry, optimization, or simulation, are the kind quantum will eventually help with, and keeping a light watch on the providers serving your industry.
It can also mean starting small now, experimenting on the machines available today to build the institutional know-how that will matter when the technology matures. The companies that begin learning early, even on limited hardware, will recognize the moment of arrival and move quickly, while those starting from zero will be at a disadvantage.
The deeper point is that fault tolerance is coming on a timeline that serious people are now willing to put on paper, and that changes the question from whether to prepare to how. The race to fault tolerance is one of the most consequential contests in technology, and understanding where it stands is the foundation for being ready when its winners cross the line.
Jason Kumpf tracks the milestones that will make quantum computing practical. He is Head of US Revenue at Razorpay, a board advisor, angel investor, and speaker. More about Jason.