Quantum computing has made a meaningful step toward breaking today’s public-key cryptography, but this is not a quantum computer breaking RSA. It is a better estimate of how one future machine might do it, and that distinction matters. Alice & Bob’s new resource analysis puts RSA-2048 at 22,593 physical qubits and 3.14 days of runtime on a fault-tolerant superconducting cat-qubit architecture, roughly 40 times fewer physical qubits than an earlier Google estimate.

That is genuine progress, because quantum security planning depends on resource estimates, not raw qubit-count theater. A machine with 100,000 noisy qubits is not automatically closer to cryptographically relevant computation than a machine with 20,000 better-protected qubits. The useful unit is the complete stack: physical error rates, connectivity, correction overhead, decoding speed, logical-gate fidelity, and the time required to run a deep version of Shor’s algorithm.

But the commercial takeaway is not that RSA-2048 is about to fall. The estimate assumes a particular cat-qubit architecture, a fault-tolerant system that does not yet exist at this scale, and engineering performance that must hold across millions or billions of operations. The important change is more uncomfortable: the safety margin around public-key cryptography is narrowing on paper, while migration programs take years in the real world. Waiting for a demonstration is not prudence. It is a plan to discover the problem after the infrastructure has become impossible to replace quickly.

Alice & Bob’s result is real science, not quantum-washing. It attacks a serious bottleneck in quantum computing, the number of physical qubits required to create reliable logical qubits. Conventional error correction spreads one logical qubit across many physical qubits because every component is noisy. Cat qubits encode information in oscillator states designed to suppress one class of errors natively, potentially reducing the correction burden.

The number that matters is not 22,593 by itself. It is the product of qubits and runtime, under stated assumptions. Alice & Bob estimates RSA-2048 factoring at 22,593 physical qubits over 3.14 days, while elliptic-curve attacks at the 256-bit level require 29,588 physical qubits and about 61.8 minutes. That second figure is strategically more important for many modern systems. Elliptic-curve cryptography protects digital signatures, key exchange, blockchain infrastructure, and a large fraction of internet authentication. It is not enough to ask whether a future quantum computer can factor RSA. The attack surface includes every public-key primitive that quantum algorithms weaken.

The estimate also exposes how misleading vendor qubit counts can be. A physical qubit is a component, not a unit of useful computation. It can lose its state, suffer leakage, be misread, or interact imperfectly with its neighbors. A logical qubit is the error-corrected object that algorithms actually need. The conversion between the two depends on the physical error model and the code architecture. Two companies can advertise similar physical-qubit counts while being separated by orders of magnitude in useful computational depth.

That is why this announcement should not be compared casually with IBM’s processor roadmap or Google’s error-correction demonstrations. Those programs may be stronger in other dimensions, including fabrication, control, logical-qubit experiments, or system integration. Alice & Bob’s advantage is architectural specialization. Cat qubits can lower overhead for a particular noise bias, but they do not make arbitrary errors disappear. The burden shifts toward suppression of the remaining error channel, fast syndrome extraction, reliable coupling, and a decoder that can keep up with the machine.

The most aggressive reading of the announcement is therefore wrong. This is not evidence that Alice & Bob has a cryptographically useful machine. It is evidence that the resource estimate has moved into a range where national-security planners and large enterprises should stop treating quantum attacks as an abstract century-scale scenario.

There is also a subtle caveat in the comparison with Google’s estimate. A “40 times fewer qubits” claim is meaningful only if the two studies use comparable assumptions about error rates, architecture, gate scheduling, error correction, parallelism, and acceptable failure probability. Resource estimates are models, not race results. The model can improve while the hardware remains far away.

The timeline for quantum risk is now clearly ahead of the timeline for general-purpose quantum advantage. That sounds contradictory, but it is not. Breaking a cryptographic key is a highly structured task with an algorithm designed for it. Running a commercially valuable chemistry, optimization, or machine-learning workload may require different connectivity, deeper circuits, better data loading, and fault tolerance that is harder to achieve. Quantum computers can become dangerous to encryption before they become broadly useful to enterprise computing.

For executives, the correct response is migration, not panic. RSA and elliptic-curve systems embedded in certificates, firmware signing, identity systems, industrial control equipment, payment infrastructure, and archived communications cannot be replaced on a weekend. The difficult part is inventory and dependency mapping, especially where vendors have hidden public-key cryptography inside appliances and long-lived devices. Post-quantum cryptography is less glamorous than a processor launch, but it is the actual near-term quantum business.

The industry winners will not necessarily be the companies with the most qubits. They will be the companies that can demonstrate a credible path from physical hardware to logical operations with transparent error budgets. That favors architectures with a strong noise model, scalable fabrication, efficient decoding, and a clear answer to how the system runs continuously rather than merely producing an impressive isolated experiment.

Watch three things next. First, whether Alice & Bob publishes measured hardware parameters that close the gap between its model and its machine. Second, whether the estimate survives independent benchmarking under comparable assumptions. Third, whether competitors respond with lower resource estimates that include realistic failure probabilities rather than headline qubit counts.

The broader signal is decisive: quantum progress is becoming less about adding qubits and more about reducing the cost of reliable computation. Alice & Bob has not broken RSA. It has made the engineering target harder to dismiss, and that is precisely why cryptographic migration belongs on executive agendas now.