← Back to blog
Cloud Security

Quantum deadlines: which date should you actually plan for?

Allen Clark

Allen Clark

A row of desk calendars standing along a glowing timeline that recedes into the distance, set against a faint lattice of connected nodes suggesting a quantum computer, in navy and teal.

Last week I spent a day at a quantum and AI security summit in London and heard more dates than I could write down. Q-Day was put at anywhere from next year to 2035. One speaker said the estimate had dropped from 2035 to 2030 to 2027 in twelve months. Another said some organisations have to comply by 1 January 2027, which is effectively now.

If you're a CISO or a board member trying to work out when to start, that's not much help. These dates measure different things, so they need pulling apart.

Date one: when a quantum computer could break today's encryption

This is Q-Day and nobody knows when it will be. What we do have are published estimates of how big a quantum computer would need to be to break RSA-2048. Those estimates have come down quickly.

  • In 2019, Craig Gidney and Martin Ekerå estimated about 20 million noisy qubits running for about eight hours.
  • In 2025, Gidney put it at under a million noisy qubits running for under a week, using the same hardware assumptions.
  • In February 2026, Iceberg Quantum published a design needing fewer than 100,000 physical qubits running for about a month. It relies on longer-range qubit connectivity that hasn't yet been shown working at that scale and was still a preprint when it came out.
Bar chart on a logarithmic scale titled "The estimates are falling, but the gap is still large". It shows the qubits estimated to be needed to break RSA-2048, falling from about 20 million noisy qubits (Gidney and Ekerå, 2019) to under 1 million (Gidney, 2025) to under 100,000 physical qubits (Iceberg Quantum preprint, 2026). A shorter amber bar shows today's largest machines at a few thousand noisy qubits at most. A footnote says these are estimates of what would be needed, not machines that exist.
Each estimate has cut the size of machine needed, but nothing built today comes close.

Today's machines have at best a few hundred to a few thousand noisy qubits. So the gap is closing on paper, but these are estimates of what would be needed rather than machines that exist. The Global Risk Institute's latest expert survey, published in March, put the chance of a cryptographically relevant quantum computer within ten years at between 28% and 49%. That is still a probability rather than a date.

One speaker went further and said quantum computing is already available as a service from certain countries. You send them your encrypted data, wait a while and get it back unencrypted. A service like that would be illegal and nobody would advertise it, so the absence of public evidence doesn't prove much either way. What I can say is that it would need a machine far beyond anything publicly known. A 2025 study by armasuisse and ETH Zurich ran Shor's algorithm on cloud quantum computers and found a substantial gap between what they can do and what breaking real encryption needs. If such a service does exist, it means someone has quietly built far more than the published research suggests is possible today. Either way the conclusion doesn't change, because harvest now, decrypt later is already a reason to act on long-lived data.

Date two: other countries' and standards bodies' deadlines

The 1 January 2027 date is real, but it doesn't apply to most of us. It comes from the NSA's CNSA 2.0 guidance and covers new acquisitions for US National Security Systems. It's a procurement rule for one part of the US government rather than an estimate of when quantum computers will arrive.

Singapore was also mentioned as having told large enterprises to submit migration plans. The Cyber Security Agency of Singapore requires owners of Critical Information Infrastructure to submit plans by 31 March 2027 and to finish migrating by the end of 2031. Other organisations are encouraged to finish by 2035. There is also a grant covering up to 70% of the cost of pilot projects, open until October 2027.

The US standards body has now added its own dates. NIST finalised its transition report, IR 8547, on 27 October. It deprecates quantum-vulnerable public-key algorithms such as RSA and elliptic curve after 2030 and disallows them after 2035. Deprecated means they can still be used if the owner accepts the risk. Separately, White House orders require US federal agencies to move their high-value systems to post-quantum key establishment by the end of 2030 and to post-quantum signatures by the end of 2031.

The EU has gone for a similar shape. Its coordinated roadmap, published in June 2025, asks member states to have done the groundwork by the end of 2026, including cryptographic inventories and supply chain dependency maps. High-risk use cases should be migrated by the end of 2030, when quantum-vulnerable public-key mechanisms shouldn't be used on their own. Medium-risk use cases follow by 2035.

None of these applies to a UK business directly. They do show how seriously other governments are taking it, with 2030 emerging as the date for high-risk systems almost everywhere. Anyone selling into the US or EU may also meet the requirements through their customers' contracts long before any UK rule applies.

Timeline from 2026 to 2035 titled "Which quantum date applies to you?". 1 January 2027, US national security systems, new purchases under CNSA 2.0. 31 March 2027, Singapore critical infrastructure plans due. 2028, highlighted, UK NCSC discovery done and migration plan in place. 2030, EU high-risk systems migrated and NIST deprecates RSA and elliptic curve. 2031, UK highest-priority migration work done and Singapore critical infrastructure migrated. 2035, UK migration complete and NIST disallows RSA and elliptic curve.
Most of these dates belong to other governments. For UK organisations the one to work to is 2028, when you're expected to know where your cryptography is and have a plan.

Date three: the UK milestones

The NCSC has set three dates for UK organisations.

  • By 2028, organisations should have found where they use cryptography and have a migration plan in place.
  • By 2031, the highest-priority migration work should be done and the plan turned into a full roadmap.
  • By 2035, migration should be complete.

Most of the coverage I see quotes 2035. For most organisations the date that matters is 2028, because that's when you're expected to know what you've got. That is a discovery exercise and it takes longer than people expect. Cryptography sits in applications, network kit, certificates, code signing, backups and almost every supplier's product.

Date four: how long your data needs to stay secret

This is the one only you can set and it's the one that makes the others urgent. Data stolen today can be stored and decrypted later once the capability exists. Health records, intellectual property, customer data and anything else that still needs to be confidential in ten or fifteen years is already exposed if it travels over encryption that a future quantum computer could break.

Michele Mosca put this as a simple test. If the time your data needs to stay secret plus the time it takes you to migrate is longer than the time until a capable quantum computer exists, you're already late. That holds whichever Q-Day estimate you believe.

Diagram titled "Mosca's test: are you already late?". The top bar adds how long your data must stay secret to how long it takes you to migrate. The bottom bar, which is shorter, shows the time until a capable quantum computer exists. The part of the top bar that runs past the bottom one is marked as exposed. A line underneath says that if the top bar is longer than the bottom one, data you protect today is already at risk. A footnote says the bar lengths are illustrative.
Add how long your data must stay secret to how long your migration will take. If that's longer than the time until a capable quantum computer exists, you're already late.

Some of this is already happening without anyone asking. Cloudflare says about 70% of the browser traffic reaching its network now uses hybrid post-quantum key exchange. That's the easy part of the job. The harder part is everything that doesn't upgrade itself, such as internal applications, older network kit, VPNs, embedded devices and the systems suppliers run on your behalf.

Signatures work differently from encryption. Forging a signature needs a quantum computer at the moment of the forgery, so harvesting doesn't apply in the same way. The risk there is long-lived records and code that have to be trusted for decades. That's a good reason to treat signatures and encryption as separate parts of the plan.

What I'd do

  • Start the discovery work now and aim to finish it well before 2028. Include suppliers, because most of the cryptography you depend on sits inside someone else's product.
  • Sort what you find by how long the data needs to stay secret, so the longest-lived confidential data goes first.
  • Ask every critical supplier what their quantum-safe plan is and when they'll ship it.
  • Build in crypto-agility so changing an algorithm or certificate becomes routine. Public TLS certificate lifetimes are already being cut to 47 days from 15 March 2029, so that automation is needed anyway.
  • Be wary of anything that relies on a single dramatic date to justify the spend. The risk to long-lived data stands up on its own.

The dates that came up at the summit were all real in some sense, but only one of them is a UK deadline and only one of them is about your data. Which date is your organisation working to?

Sources

    Quantum deadlines: which date should you actually plan for? | djinn six