Learn · The quantum threat
The quantum threat, explained
What the risk is, how it arises, what it would take, when it could arrive, what it affects, and how organisations should respond.
In short: a large, error-corrected quantum computer could break the public-key cryptography (RSA and elliptic-curve) that secures almost every digital system. Nobody knows exactly when such a machine will exist, but data copied today can be decrypted later, signatures could be forged, and replacing cryptography takes years. That is why regulators already expect organisations to plan the move to quantum-safe cryptography.
1. What is the quantum threat?
The quantum threat is the risk that a cryptographically relevant quantum computer (CRQC) will make today’s public-key cryptography insecure. Public-key cryptography lets strangers on the internet agree secret keys and prove who they are. It underpins HTTPS websites, online banking, VPNs, email security, software updates, payment cards and digital identity.
The threat does not come from today’s quantum computers, which are far too small and error-prone. It comes from the fault-tolerant machines that several companies are now building, and from attackers who collect encrypted data in the meantime.
2. How today’s cryptography works
| Job | Typical algorithms today | Effect of a large quantum computer |
|---|---|---|
| Agreeing a secret key (key exchange) | ECDH, X25519, RSA key transport, Diffie-Hellman | Broken by Shor’s algorithm |
| Proving identity (digital signatures, certificates) | RSA, ECDSA, EdDSA | Broken by Shor’s algorithm |
| Encrypting the data itself | AES, ChaCha20 | Weakened by Grover’s algorithm; 256-bit keys remain strong |
| Hashing and integrity | SHA-2, SHA-3 | Modestly weakened; longer outputs remain strong |
3. Who and what is affected
Almost every organisation is affected, because public-key cryptography is built into the systems everyone uses. The table shows where it sits and what a quantum computer would put at risk.
| System | Typical cryptography | Main quantum risk |
|---|---|---|
| Websites and APIs (HTTPS/TLS) | ECDHE key exchange; RSA or ECDSA certificates | Harvest now, decrypt later; impersonation |
| VPNs and remote access | IKE/IPsec or TLS with ECDH and RSA | Harvest now, decrypt later |
| Certificates and PKI | RSA and ECDSA certificate chains | Forged certificates; long-lived root keys |
| Software and firmware updates | RSA or ECDSA code signing | Malicious updates signed with forged keys |
| Email security | S/MIME and PGP with RSA or ECC | Stored email decrypted later |
| Remote administration | SSH host and user keys | Server impersonation |
| Payments and banking messaging | RSA and ECC in cards, HSMs and interbank links | Fraud; long-lived financial data exposed |
| Digital identity and smart cards | RSA and ECC keys on cards and tokens | Credential forgery |
| Blockchains and digital assets | ECDSA and EdDSA signatures | Theft by forging transaction signatures |
| IoT and operational technology | Often fixed RSA or ECC in firmware | Devices that cannot be upgraded in time |
4. What is not broken
Symmetric encryption such as AES-256 and hash functions such as SHA-384 and SHA-512 are expected to remain secure. Quantum computers also do not break the new post-quantum algorithms that NIST standardised in August 2024. See Standards.
5. How a quantum computer breaks it
Why RSA and elliptic curves are safe today
RSA is secure because multiplying two large prime numbers is easy, but working backwards from the product to the primes is extraordinarily hard. Elliptic-curve cryptography relies on a similar one-way problem called the discrete logarithm. The best classical methods grow so slowly that factoring a 2048-bit RSA key is far beyond any existing computer. For comparison, the largest RSA number factored publicly with classical computers, RSA-250 (829 bits) in 2020, took about 2,700 core-years of computing.
Shor’s algorithm, step by step
In 1994 the mathematician Peter Shor showed that a quantum computer could solve both problems efficiently. The trick is to turn factoring into a question about a repeating pattern, which quantum interference is very good at finding.
- Step 1 (classical): pick a random number and set up a function that repeats with a hidden period. If you know the period, simple arithmetic gives you the factors.
- Step 2 (quantum): place a register of qubits in superposition, so it represents a huge range of inputs at once.
- Step 3 (quantum): compute the repeating function on all of those inputs together. This “modular exponentiation” is the most expensive part of the whole algorithm.
- Step 4 (quantum): apply the quantum Fourier transform. Interference cancels out most outcomes and makes measurements cluster around values that reveal the period.
- Step 5 (classical): use the period to compute the factors with a greatest-common-divisor calculation. The run is repeated a few times if needed.
A version of the same idea solves the elliptic-curve discrete logarithm problem, which breaks ECDH key exchange and ECDSA or EdDSA signatures.
Grover’s algorithm
Lov Grover’s 1996 algorithm speeds up brute-force search, roughly halving the effective strength of a symmetric key. AES-128 would offer about 64 bits of security against such an attack, while AES-256 would still offer about 128 bits, which is why long-term guidance recommends 256-bit keys. Grover’s algorithm is also hard to run in parallel, which limits its practical impact further.
6. What it would take
Shor’s algorithm needs a long, error-free calculation: billions of quantum operations in a row. No physical qubit is reliable enough for that, so the machine must use logical qubits, each built from many physical qubits through error correction. Three numbers describe the size of the machine needed:
- Logical qubits: how much error-corrected working memory the algorithm needs.
- Physical qubits: the actual hardware, including the qubits used for error correction and for producing special “magic states”.
- Run time: how long the calculation takes, which depends on how fast error-correction cycles run.
| Estimate | Target | Logical qubits | Physical qubits | Run time |
|---|---|---|---|---|
| Gidney and Ekerå (2019) | RSA-2048 | — | About 20 million | About 8 hours |
| Gidney, Google (May 2025) | RSA-2048 | About 1,400 | Fewer than 1 million | Less than a week |
| Roetteler et al., Microsoft (2017) | 256-bit elliptic curve | About 2,330 | — | — |
| Google Quantum AI (2026) | 256-bit elliptic curve | Fewer than 1,200 to 1,450 | — | — |
| Chevignard, Fouque, Schrottenloher (Eurocrypt 2026) | 256-bit elliptic curve | About 1,100 to 1,200 (more operations) | — | — |
The physical-qubit estimates assume qubits with about a 0.1% error rate and error-correction cycles of around a microsecond, which is better than most of today’s hardware but within the range engineers are targeting. In 2026, the company Iceberg Quantum claimed RSA-2048 could be broken with fewer than 100,000 physical qubits using a different family of error-correcting codes; this claim needs hardware with more flexible qubit connections and is still being examined.
How far away is that from today?
| Today (2026) | Needed to break RSA-2048 (2025 estimate) | |
|---|---|---|
| Physical qubits in the largest processors | Hundreds to a few thousand | Fewer than 1 million |
| Error-corrected logical qubits | Small-scale demonstrations | About 1,400 running for days |
| Error correction “below threshold” | First shown by Google in December 2024 | Needed at very large scale |
The gap is still large. But each step in the table is an engineering problem rather than an open question of physics, and companies such as IBM have published roadmaps to fault-tolerant machines with hundreds of logical qubits by the end of the decade.
7. When could it happen?
No one can give a date. The most widely cited expert survey, the Global Risk Institute’s Quantum Threat Timeline Report (published March 2026), found that experts put the chance of a CRQC within 10 years at 28–49% and within 15 years at 51–70%, the highest estimates in the survey’s history.
Regulators have translated this uncertainty into deadlines: most expect high-risk systems to move to quantum-safe cryptography by around 2030 and everything else by 2035. See the Regulation Tracker.
8. How the threat plays out
Harvest now, decrypt later

Encrypted traffic and files can be copied now and stored. Anything that must stay confidential for many years, such as health records, financial data, trade secrets and government information, is already exposed. National security agencies have warned that well-resourced adversaries may already be collecting data this way.
Forged signatures, identities and updates
Once a CRQC exists, attackers could forge digital signatures: impersonating websites, issuing fake certificates, signing malicious software updates or approving fraudulent transactions. Systems with long-lived signing keys, such as certificate authorities and firmware update channels, need to change early.
Wider and systemic effects
- Digital assets and payments: systems that rely on elliptic-curve signatures could see funds moved with forged transactions.
- Signed records: contracts and audit trails protected by today’s signatures may become harder to rely on as evidence.
- Financial stability: the G7 Cyber Expert Group warns that quantum computing could expose financial institutions’ data and that coordinating upgrades will take many years.
- Migration risk: new implementations can contain bugs, such as the 2024 “KyberSlash” timing weaknesses reported in several Kyber implementations, and larger keys can break older equipment.
- Long-lived devices: industrial systems, vehicles, medical devices and smart cards bought today may still be in service when attacks become possible.
9. How to respond
- Govern. Make quantum risk a board-level topic with a named owner.
- Discover. Build an inventory of where public-key cryptography is used, including in suppliers’ products.
- Prioritise. Start with systems that protect long-lived data or act as roots of trust.
- Migrate. Adopt the NIST standards (ML-KEM, ML-DSA, SLH-DSA), often in hybrid mode first.
- Stay agile. Design systems so algorithms can be changed by configuration in future.
10. Common myths
| Myth | Reality |
|---|---|
| “Quantum computers can break all encryption.” | They threaten public-key cryptography. Symmetric encryption with 256-bit keys and modern hash functions remain strong, and post-quantum algorithms are designed to resist them. |
| “It is decades away, so we can wait.” | Experts give a substantial chance within 10 years, data stolen today is already at risk, and migration takes years. |
| “We need a quantum computer to protect ourselves.” | No. Post-quantum cryptography runs on ordinary computers and is already available in mainstream software. |
| “Quantum key distribution is the answer.” | For most organisations, security agencies recommend post-quantum cryptography rather than QKD, which needs special hardware and links. |
- Global Risk Institute, Quantum Threat Timeline Report 2025 (published Mar 2026)
- Gidney, How to factor 2048 bit RSA integers with less than a million noisy qubits (arXiv, May 2025)
- Microsoft Research: Roetteler et al., quantum resource estimates for elliptic curve discrete logarithms (Asiacrypt 2017)
- Google Research: Gidney et al., resource estimates for 256-bit elliptic curves
- Chevignard, Fouque and Schrottenloher, Reducing the number of qubits in quantum discrete logarithms on elliptic curves (Eurocrypt 2026)
- The Quantum Insider, Q-Day just got closer (Mar 2026), on the Iceberg Quantum Pinnacle claim
- Google, Meet Willow (Dec 2024)
- IBM roadmap to Starling, as reported June 2025
- US Treasury: G7 Cyber Expert Group statement on quantum computing (dated 25 September 2024)
- NIST FIPS 203 (ML-KEM), with FIPS 204 and 205 (Aug 2024)
- Boudot et al., Comparing the difficulty of factorization and discrete logarithm: a 240-digit experiment (CRYPTO 2020), and the RSA-250 factorisation announcement (2020)
