Quantum Readiness Has a Deadline. Nobody Knows the Date.

Somewhere, on a disk you will never see, there may be a copy of what your API said this morning.

It is encrypted, so whoever holds it can't read it. They don't need to yet. Storage is cheap, and they are betting that a machine able to undo the encryption arrives while your data still matters.

The industry calls this harvest now, decrypt later. It is why quantum computing belongs in your planning now.

Harvest now, decrypt later Two lanes of the same encrypted traffic between a browser and an API. In both, an attacker copies the traffic into storage today. Years later a large quantum computer exists. Traffic protected by RSA or ECDH key exchange is decrypted and the patient records are readable. Traffic protected by hybrid ML-KEM key exchange stays locked. TODAY: THE TRAFFIC IS RECORDED LATER: A QUANTUM COMPUTER RSA OR ECDH KEY EXCHANGE Browseruser session APIpatient-portal ATTACKERRecorded traffic ENCRYPTEDUnreadable for nowDECRYPTEDPatient records readable HYBRID ML-KEM KEY EXCHANGE Browseruser session APIpatient-portal ATTACKERRecorded traffic ENCRYPTEDUnreadable for nowSTILL LOCKEDML-KEM holds
Figure 1. Harvest now, decrypt later. Traffic recorded today is readable later unless its key exchange already uses a post-quantum algorithm.

The number keeps shrinking

In 2019, the best public estimate said that breaking a 2048-bit RSA key would take a quantum computer with around 20 million qubits.[1] It sounded like science fiction, and most of us filed it there.

In 2025, one of the same researchers at Google redid the math and came back with fewer than one million.[2] Nobody had to build anything to get there. It was better algorithms, worked out on paper.

In February 2026, the startup Iceberg Quantum published a design that needs fewer than 100,000.[3] In October, Alice & Bob put its own figure at 22,593.[4]

These are designs for machines nobody has built yet, and each rests on its own assumptions. But the estimates keep falling while the machines keep growing, and nobody can tell you the year they meet.

Estimated qubits to break RSA-2048 Published estimates of the noisy physical qubits needed to break a 2048-bit RSA key, on a log scale: about 20 million in 2019, under one million in 2025, under 100,000 in February 2026, and 22,593 in October 2026. PHYSICAL QUBITS TO BREAK RSA-2048 (LOG SCALE) 10k100k1M10M100M~20 million2019Gidney and EkerÄ< 1 million2025Gidney, Google< 100,000Feb 2026Iceberg Quantum22,593Oct 2026Alice & Bob ABOUT 900x FEWER IN SEVEN YEARS
Figure 2. Published estimates of the noisy physical qubits needed to break RSA-2048. Each is a design for a machine nobody has built yet, under its own assumptions.

Y2K at least printed its date on the calendar. The first people to own a machine that breaks RSA will have every reason to keep quiet about it.

The people who set the rules have stopped waiting

Ask ten experts when that day comes and you will get ten answers. The people who write the standards stopped arguing about the date and picked their own.

NIST published the replacement algorithms in 2024. In October 2026 it finalized the transition plan: today's public-key cryptography is deprecated after 2030 and disallowed after 2035.[5] That covers RSA and elliptic curves, which is nearly every key, certificate and signed token your systems rely on.

The browsers did not wait. Chrome and Firefox already negotiate a post-quantum key exchange with any server that offers one.

If 2030 sounds far away, remember SHA-1. Researchers showed it was weak in 2005,[6] and browsers stopped accepting it in 2017. That was twelve years to retire a single hash function. This time it is every public key you own.

The question nobody can answer

Every migration like this starts with the same question: where, exactly, do we use cryptography?

Some of the answer is in code written by people who left years ago. Some is buried in config files and in libraries you never chose on purpose. And some is in no repository at all, because it is whatever your load balancer agrees to when a client says hello.

Most teams answer it with a spreadsheet and a round of interviews. Six months later the spreadsheet is wrong.

So we built the inventory

NightVision's quantum readiness answers that question from both places.

It starts in your source. You already run nightvision openapi extract to map your API. That same run now reads the cryptography your code chooses, across seven languages, down to the line where the RSA key is generated or the token is signed. It runs on your machine, and your code never leaves it.

It finishes on the wire. Every NightVision scan greets your hosts the way a current browser does and records whether they can negotiate a post-quantum key exchange. It notes which algorithms sign the tokens in your traffic and what your identity provider publishes. It only connects and reads, and it works on the services behind your firewall.

Everything lands on one screen, sorted by what to do about it: Fix now, Migrate by 2030, Plan by 2035, or Ready. When a customer or an auditor asks for proof, the inventory exports as a Cryptography Bill of Materials.

One inventory from source and the wire Findings from the source, such as an RSA-2048 key generated in code, and from a scan, such as a host that does not offer a post-quantum key exchange, land on one Quantum readiness screen sorted into Fix now, Migrate by 2030, Plan by 2035 and Ready. The inventory exports as a CycloneDX Cryptography Bill of Materials. IN YOUR SOURCEnightvision openapi extractauth/keys.py:42 · RSA-2048 ON THE WIRENightVision scanapi.example.com · no ML-KEM QUANTUM READINESSapi.example.com, 48 findings Fix now3Migrate by 203014Plan by 20359Ready22 EXPORTCBOMCycloneDX 1.7 ✓ ready for the auditor
Figure 3. What your code chooses and what your hosts negotiate land on one screen, sorted by when each has to change, and export as a Cryptography Bill of Materials. Counts are illustrative.

Start counting

Replacing your cryptography will take years. Finding it takes one command, and you can run it this afternoon:

nightvision openapi extract . --no-target

Someone may already be saving your traffic for later. The least you can do is know what they will find.

References

  1. Craig Gidney and Martin Ekerå, How to factor 2048 bit RSA integers in 8 hours using 20 million noisy qubits, 2019.
  2. Craig Gidney, How to factor 2048 bit RSA integers with less than a million noisy qubits, 2025.
  3. The Quantum Insider, Q-Day Just Got Closer: Three Papers in Three Months Are Rewriting the Quantum Threat Timeline, March 2026.
  4. The Quantum Insider, Alice & Bob Estimates 22,593 Qubits to Break RSA-2048, October 2026.
  5. NIST, IR 8547, Transition to Post-Quantum Cryptography Standards, October 2026.
  6. Xiaoyun Wang, Yiqun Lisa Yin and Hongbo Yu, Finding Collisions in the Full SHA-1, CRYPTO 2005.

Start your cryptography inventory

Quantum readiness is included with NightVision. It reads the cryptography your code chooses and what your running hosts negotiate, and rates each finding by when it has to change. Read how it works.