Quantum computers are crossing the fault-tolerance threshold — and the implications are larger than most realize

Quantum computing has lived in a peculiar state for the past decade: simultaneously overhyped in press releases and underappreciated in its genuine technical difficulty. The milestones of 2024 and 2025 — Google's Willow chip, Microsoft's Majorana-based results, IBM's expanding roadmap — mark something real. But to understand why they matter, you first need to understand what quantum computers are actually trying to overcome.

NISQ Devices vs. Fault-Tolerant Quantum Computers: A Crucial Distinction

Most quantum computers that exist today fall into a category researchers call NISQ devices — Noisy Intermediate-Scale Quantum systems. "Noisy" is the operative word. Quantum bits (qubits) are extraordinarily fragile. They lose their quantum state through a process called decoherence, caused by heat, electromagnetic interference, vibration, even cosmic rays. Every quantum operation introduces errors. In a NISQ device, these errors accumulate faster than they can be corrected, which means computations are limited in both depth (the number of sequential operations) and reliability.

NISQ devices are real, programmable quantum computers — but their practical usefulness is constrained. They can demonstrate interesting quantum phenomena and may outperform classical computers on narrow, specially constructed benchmarks. What they cannot do is run the sustained, deep quantum circuits required for the applications that make quantum computing genuinely transformative: breaking modern encryption, simulating complex molecules, or solving industrial-scale optimization problems.

Fault-tolerant quantum computers are something different in kind. A fault-tolerant system uses quantum error correction to detect and fix errors continuously as the computation runs, allowing arbitrarily long and complex quantum circuits to execute reliably. The challenge is that error correction itself requires overhead — significant overhead.

What "Fault Tolerance" Actually Requires: Logical vs. Physical Qubits

The central challenge of fault-tolerant quantum computing is the distinction between physical qubits and logical qubits. A physical qubit is the actual hardware component — a superconducting circuit, a trapped ion, a photon. A logical qubit is an error-protected quantum unit built by encoding one logical qubit across many physical qubits, using redundancy to detect and correct errors without disturbing the underlying quantum state (which you can't simply copy or measure without destroying it).

How many physical qubits does one logical qubit require? It depends on the quality of the physical qubits and the error correction code being used. Current estimates for practical fault tolerance range from hundreds to thousands of physical qubits per logical qubit. Google's surface code error correction scheme, for example, requires roughly a thousand physical qubits to produce a single logical qubit reliable enough for deep computation — and that number drops as physical qubit quality improves.

This is why quantum computing progress cannot be measured simply by qubit count. A system with 1,000 noisy physical qubits and a system with 1,000 high-quality physical qubits enabling ~1 reliable logical qubit are almost entirely different things.

Google's Willow Chip: Below-Threshold Error Correction

In late 2024, Google announced results from its Willow quantum processor that represent a genuine milestone. The key achievement was demonstrating "below-threshold" error correction — meaning that as Google added more physical qubits to its error correction scheme, the error rate of the logical qubit went down exponentially, rather than up.

This matters enormously because it had never been cleanly demonstrated before at this scale. Prior quantum error correction experiments showed that adding more physical qubits helped, but not consistently enough and not exponentially. Willow's results showed that the surface code error correction approach actually works as theoretically predicted — each additional layer of physical qubit redundancy multiplies the reliability improvement.

Google also reported that Willow solved a specific benchmark problem in five minutes that would take today's fastest classical supercomputers an estimated 10 septillion years. That headline number is real but requires context: the benchmark was designed specifically to be hard for classical computers and easy for quantum computers, not to solve any practical problem. The more significant result is the error correction scaling, which is the foundational requirement for everything else.

Microsoft's Topological Approach: Majorana Particles

Microsoft has pursued a fundamentally different strategy for building stable qubits, one that has been controversial and difficult for years: topological qubits based on Majorana zero modes. In 2025, Microsoft published experimental results claiming to have created and measured Majorana particles in a semiconductor device — an achievement that had eluded researchers for over a decade.

The appeal of Majorana-based qubits is that they are theoretically far more stable than conventional qubits. Their quantum information is stored non-locally, meaning local disturbances cannot easily corrupt it. If the approach proves scalable, it could dramatically reduce the number of physical qubits needed per logical qubit — potentially by orders of magnitude compared to surface code approaches.

Microsoft's results remain under peer review and independent verification, and the path from a demonstrated Majorana particle to a functioning logical qubit built from them involves many additional engineering challenges. But if the approach validates, it could represent a fundamentally different trajectory for fault-tolerant quantum computing.

IBM's Roadmap: 100,000+ Qubits by the End of the Decade

IBM has taken the most public and detailed roadmap approach to quantum development. The company has consistently hit its annual milestones: Eagle (127 qubits, 2021), Osprey (433 qubits, 2022), Condor (1,121 qubits, 2023), Heron (improved qubit quality focus, 2023). IBM's stated goal is to reach systems with 100,000+ physical qubits before 2030, alongside improvements in qubit quality and connectivity that would make error correction practical at scale.

IBM's strategy differs from Google's in emphasis: rather than pursuing a single breakthrough chip, IBM has focused on building quantum computing infrastructure — cloud access, tooling, developer ecosystems — while steadily improving hardware. The Heron processor in particular represented a quality-over-quantity shift, prioritizing the error rate improvements needed for error correction over raw qubit counts.

The CRQC Threshold: What It Takes to Break RSA-2048

One application drives urgency around fault-tolerant quantum computing more than any other: cryptography. Specifically, the concept of a cryptographically relevant quantum computer, or CRQC — a system capable of running Shor's algorithm at sufficient scale to break RSA-2048 encryption in a practical timeframe.

Estimates for how many logical qubits this requires have been steadily revised. Recent analyses suggest that breaking RSA-2048 would require somewhere between 4,000 and 10,000 logical qubits, running for hours to days. Given the physical-to-logical qubit ratios current architectures require, that translates to millions of physical qubits of sufficient quality. We are still many years away from a CRQC.

But "many years" is not "never," and the harvest-now-decrypt-later threat means the problem is already present even if the CRQC is not. Nation-state adversaries are plausibly archiving encrypted traffic today with the intention of decrypting it once quantum computers are capable enough. Data that must remain confidential for a decade or more is already at risk under this threat model.

Near-Term Practical Applications: Where Quantum Earns Its Keep

The cryptographic threat is the stick; here is the carrot. Fault-tolerant quantum computers will likely first prove their practical value not by breaking encryption but by simulating quantum systems — chemistry, materials science, and biology at the molecular level.

Classical computers cannot efficiently simulate quantum mechanical systems because the computational complexity scales exponentially with system size. A quantum computer does not have this problem: it is itself a quantum mechanical system and can simulate others directly. This means quantum computers could simulate protein folding and molecular binding with a precision impossible for classical systems, potentially accelerating drug discovery by orders of magnitude. New catalyst design for industrial chemistry — including carbon capture and nitrogen fixation — could become tractable. Materials with exotic quantum properties could be designed computationally before any atom is physically arranged.

Optimization problems — logistics routing, financial portfolio optimization, supply chain scheduling — are also expected to benefit from quantum speedups, though the extent and timing of those benefits are more contested in the research community.

Why "Quantum Supremacy" Doesn't Mean Useful Quantum Computing

Google first claimed "quantum supremacy" in 2019, when its Sycamore processor completed a specific sampling task faster than any classical computer could. IBM subsequently disputed the claim, and the benchmark task itself had no practical application. Similar dynamics have repeated with every subsequent "supremacy" or "advantage" demonstration, including the Willow results.

These demonstrations are scientifically meaningful — they confirm that quantum hardware can outperform classical hardware on at least some tasks, which was not obvious a decade ago. But they do not demonstrate useful quantum advantage on problems the world actually needs to solve. That requires fault tolerance, and fault tolerance requires logical qubit overhead that current systems cannot yet provide at the scale needed for real applications.

The distinction matters for evaluating vendor claims. A company announcing a "quantum advantage" on a benchmark is not necessarily claiming their system is useful for your problems. Read carefully.

Timeline Reality Check: Significant Milestones, Measured Expectations

The 2024-2026 period represents a genuine inflection point in quantum computing development. The Willow chip's below-threshold error correction result, Microsoft's Majorana particle work, and IBM's continued hardware progress all demonstrate that the theoretical foundations of fault-tolerant quantum computing are yielding to engineering. These are not incremental NISQ improvements — they are steps toward a qualitatively different kind of quantum computer.

But the gap between where the field is and where fault-tolerant quantum computers need to be for practical use remains large. Building a system with thousands of high-quality logical qubits — which requires millions of physical qubits — involves engineering challenges in cryogenic cooling, qubit connectivity, control electronics, and fabrication that will take years to resolve. Conservative estimates from researchers who have examined the hardware requirements closely put practically useful fault-tolerant quantum computers for general computation in the 2030s. Aggressive estimates push some specialized applications earlier. No credible estimate puts a CRQC in the next two or three years.

What Organizations Should Actually Do Now

Given this landscape — genuine progress, but practical fault-tolerant quantum computing still years away — what should organizations be doing?

Start post-quantum cryptography migration now. NIST finalized its post-quantum cryptography standards in 2024 (ML-KEM, ML-DSA, SLH-DSA). The cryptographic migration needed to protect against a future CRQC is a multi-year infrastructure project. Organizations with long-lived sensitive data, critical infrastructure, or national security obligations cannot wait for quantum computers to actually arrive. The harvest-now-decrypt-later threat makes this a present problem.

Inventory your cryptographic exposure. Know where RSA, elliptic-curve cryptography, and Diffie-Hellman key exchange appear in your infrastructure. TLS certificates, SSH keys, code signing, VPN configurations, encrypted databases — all need to be mapped before they can be migrated.

Engage with vendors on PQC roadmaps. Enterprise software vendors vary widely in their post-quantum readiness. If a vendor has no credible PQC migration roadmap, that is a procurement risk worth raising now rather than in 2029.

Monitor quantum hardware developments selectively. Not every quantum computing announcement warrants a strategic response, but the key technical milestones — sustained logical qubit demonstrations, below-threshold error correction at scale, CRQC timeline revisions from credible research — should be tracked by technology leadership.

The inflection point the industry has been waiting for is arriving — just not on the timeline the headlines tend to suggest. Fault-tolerant quantum computing is coming. The window to prepare is now, and it is still open.