Quantum computers are crossing the fault-tolerance threshold — and the implications are larger than most realize
For most of its history, quantum computing has been a field defined by gap between promise and practice. Processors with a hundred or a thousand qubits made headlines, while researchers acknowledged in the fine print that those qubits were too error-prone to do useful computation. The era of NISQ — Noisy Intermediate-Scale Quantum — devices produced remarkable physics but limited practical results.
That picture is changing. In late 2024, Google published results from its Willow processor demonstrating something researchers had been chasing for decades: quantum error correction that improves exponentially as the system scales. It was the clearest evidence yet that the engineering path to fault-tolerant quantum computing is real, not theoretical.
What fault tolerance actually means
A quantum bit, or qubit, is fragile. Interactions with the environment — vibrations, electromagnetic fields, thermal noise — cause decoherence, collapsing quantum states before a computation can complete. Current physical qubits have error rates that make them useless for algorithms requiring millions of gate operations.
Quantum error correction addresses this by encoding a single logical qubit across hundreds or thousands of physical qubits. The redundancy lets the system detect and correct errors in real time without measuring the logical qubit directly (which would destroy its quantum state). The catch is the overhead: a fault-tolerant quantum computer capable of breaking RSA-2048 encryption is estimated to require roughly 4,000 logical qubits — and each logical qubit may need 1,000 physical qubits to maintain. That means millions of physical qubits at high quality.
The critical metric is whether error correction scales well. In previous systems, adding more physical qubits to protect a logical qubit sometimes made things worse as the additional components introduced new error pathways. Google's Willow results showed that error rates dropped exponentially as they scaled up the error-correction code size — a "below threshold" result that demonstrates the fundamental viability of the approach.
The competitive landscape
Google's superconducting qubit approach is one of several competing architectures. IBM has committed to a roadmap that reaches 100,000+ qubit systems within this decade, focusing on quantum volume and error rates as the key metrics rather than raw qubit count. IBM's systems are accessible through the cloud and have become the primary platform for academic quantum computing research.
Microsoft has taken a different physical bet. Rather than building qubits from superconducting circuits, Microsoft has been pursuing topological qubits based on exotic quasiparticles called Majorana fermions. The theoretical advantage is that topological qubits are inherently more resistant to certain types of decoherence, potentially requiring fewer physical qubits per logical qubit. In 2025, Microsoft announced results consistent with creating and measuring Majorana-based qubits — though the field is watching carefully as the approach remains unproven at scale.
IonQ, Quantinuum, and others are working with trapped-ion architectures, which achieve lower error rates per gate operation than superconducting systems but run slower and face different scaling challenges. The diversity of approaches reflects genuine uncertainty about which physical platform will win the race to fault tolerance at scale.
What quantum computers will actually be used for
The "break encryption" framing dominates public discourse, but it's the least interesting near-term application and the furthest from practical realization. The applications that will arrive first are in quantum chemistry and materials science.
Simulating molecular behavior is classically intractable above a certain size — the computational cost grows exponentially with the number of electrons being modeled. Quantum computers are naturally suited to this problem because they can represent quantum states efficiently. Applications include designing new catalysts for industrial chemistry, discovering battery materials with higher energy density, and modeling protein-drug interactions for pharmaceutical development.
Optimization problems — logistics routing, portfolio optimization, scheduling — are another candidate, though the quantum advantage for these applications is less clear-cut than for quantum chemistry. The field is still working out where quantum provides genuine speedups versus classical heuristics.
The cryptography urgency
While fault-tolerant quantum computers capable of breaking current encryption are still years away, the threat is real enough that governments are acting now. NIST finalized its first post-quantum cryptography standards in 2024, and US agencies have been given timelines to migrate cryptographic infrastructure. The concern is "harvest now, decrypt later" — adversaries collecting encrypted data today with the intent to decrypt it once quantum capabilities mature.
Organizations managing sensitive data with long classification periods — government secrets, medical records, financial data — face the most urgent migration timelines. Standard web traffic encrypted with TLS is less immediately threatened, but the migration to quantum-resistant algorithms will eventually affect every piece of internet infrastructure.
A realistic timeline
Useful fault-tolerant quantum computers — systems that can solve problems beyond classical reach in commercially valuable domains — are most likely 7-15 years away. The recent milestones are genuine and significant, but the engineering gap between today's best systems and the millions of high-quality qubits needed for large-scale applications remains enormous.
What has changed is that the path is now clearer. The physics works. The error correction approaches are scaling as theory predicted. The remaining challenges are engineering: fabricating millions of qubits with consistent quality, operating them at millikelvin temperatures at scale, building the classical control systems fast enough to handle real-time error correction. These are hard problems, but they're engineering problems rather than fundamental physics obstacles. That distinction matters.