How much Internet traffic already negotiates post-quantum encryption?

As of October 31, 2026, the measured answer is: most of the client side, a small fraction of the server side. About 70% of browser-generated HTTPS traffic reaching Cloudflare’s network already negotiates hybrid post-quantum key agreement, per the figure Cloudflare published on September 29, 2026 — up from 29% of human-generated web traffic at the start of 2025 and 52% in early December 2025, per the Cloudflare Radar 2025 Year in Review of December 15, 2025. Meanwhile only about 15% of the origin servers Cloudflare connects to support it, per the same September 29 post. The public record still shows 0 RSA-2048 keys broken by a quantum computer and 0 breaches attributed to a quantum attack, re-verified by Rosetta Quantum today, October 31, 2026. This page puts the whole measured series next to its sources, and next to the regulatory clocks it will collide with.

What does “negotiates post-quantum encryption” actually measure?

It measures one specific step: key agreement in TLS — whether a browser and a server derive their shared session secret with the hybrid X25519+ML-KEM exchange instead of a purely pre-quantum one. ML-KEM is the lattice-based mechanism NIST standardized as FIPS 203 on August 13, 2024. Nothing in the percentages on this page is about signatures or certificates: the certificate your server presents is still, in virtually every deployment, RSA or ECDSA, and the migration of signatures runs on its own, slower track — US Executive Order 14412 with OMB M-26-15, as summarized by Cloudflare on June 23, 2026, will require post-quantum signatures in federal high-value assets only by December 31, 2031, a year after its key-exchange deadline.

Key agreement went first for a reason the house has covered in detail: recorded ciphertext can be decrypted later, so the confidentiality of today’s traffic against a future machine depends on the key exchange used today. That is the harvest-now-decrypt-later arithmetic, and it is the stated rationale of the CISA/NSA/NIST quantum-readiness factsheet of August 21, 2023. The numbers below are therefore confidentiality numbers, not authentication numbers — a distinction the marketing around “quantum-safe” products rarely makes.

How fast did the client side move?

From 29% to about 70% in roughly 21 months, with every reading sourced and dated. The series, as registered on Rosetta Quantum’s public PQC board: 29% of human-generated web traffic post-quantum encrypted at the start of 2025 and 52% in early December 2025, both from the Cloudflare Radar 2025 Year in Review published December 15, 2025; approximately 67% of human HTTPS requests on April 15, 2026, from Cloudflare Radar’s published reading of that date; and about 70% of browser-generated traffic at the publication of Cloudflare’s September 29, 2026 post. The three phrasings are close cousins rather than one continuous instrument — the what-we-don’t-know block below prices that caveat.

The striking part is why it moved: almost nobody migrated anything. Browsers update themselves, and the default did the work. Chrome negotiates hybrid ML-KEM by default from version 131, Edge from 131, Firefox from 132 on desktop and 145 on Android, Safari from version 26, per Cloudflare’s protocol support matrix consulted today, October 31, 2026. A CISO who did nothing in 2025 still saw their users’ browsers go post-quantum. That free ride ends at the server room door.

Why is the server side stuck near 15%?

Because servers do not update themselves. Cloudflare scanned the origins it connects to and reported 12.8% supporting post-quantum key agreement on September 8, 2026 — alongside roughly 45 billion post-quantum origin connections per day — and about 15% in its September 29, 2026 post. Set 15% of origins against 70% of browser traffic and the asymmetry is the finding: the visitor half of the handshake is largely ready, and the operator half is largely not. When a PQC-capable browser meets a classical-only origin, the handshake silently falls back to pre-quantum key agreement — the traffic is encrypted, but it is the recordable kind.

The software for the lagging half already ships with post-quantum defaults: OpenSSL negotiates X25519MLKEM768 by default from 3.5.0, Go from 1.24, Node.js from 22.20.0 and 24.5.0, Caddy from 2.10.0, NGINX when compiled against OpenSSL 3.5+, per the same support matrix of October 31, 2026. The gap is not missing software; it will close at the speed of fleet upgrades, load-balancer configs and appliance refreshes — which is exactly why regulators set deadlines a decade before anyone claims a cryptographically relevant quantum computer will exist.

Which stacks negotiate post-quantum key agreement by default today?

The matrix below is the deployment reality a security team can test against, one row per stack, all from Cloudflare’s protocol support page consulted October 31, 2026:

A one-line test exists for any endpoint you operate: a TLS client built on OpenSSL 3.5+ will negotiate X25519MLKEM768 against your edge if your edge can; if the handshake comes back classical, that endpoint is on the recordable side of the ledger.

What should a security manager do with these numbers?

Read them as a deadline schedule, not as a threat report. First: the 70% figure of September 29, 2026 means your users are already asking for post-quantum confidentiality on nearly every connection — whether they get it is decided by your side of the handshake. Second: the binding dates are regulatory, not speculative. NIST IR 8547, final since October 27, 2026, will deprecate RSA and ECC after 2030 and disallow them after 2035; EO 14412 will require post-quantum key exchange in US federal high-value assets by December 31, 2030; the UK NCSC timeline of March 20, 2025 will expect discovery done by 2028, high-priority migration by 2031 and completion by 2035 — the house mapped all of them when RSA’s expiry dates went final. Third: prioritize by data lifetime, not by fear — the harvest-now-decrypt-later arithmetic in the previous PQC WATCH piece turns your retention periods into your migration deadline. And for calibration: Rosetta’s Q-Day call — floor 2031, window 2033–2039, registered as prediction RQ-PRED-QDAY-001 in the house forecast — is a sealed, refutable judgment, not a measurement, and nothing in today’s board moves it.

PQC Board — status as of 2026-10-31

The latest reading of every indicator Rosetta Quantum tracks for this channel, with its source and reading date — four rows refreshed today:

The two zeros are first-class data of this channel: they are re-verified against the public record on every edition of this board, and the day either stops being zero will be a bigger story than any percentage above.

What we don’t know

We do not know today’s live share: Cloudflare Radar’s dashboard renders its current value in JavaScript, so this page cites only published, dated readings — the true share on October 31, 2026 may sit above or below the ≈70% of September 29. We do not know how representative Cloudflare’s vantage is: it is one network’s sample of the web, and we found no comparable published series from another large edge provider to triangulate against. The three client-side metrics — human-generated web traffic in 2025, human HTTPS requests in April 2026, browser-generated traffic in September 2026 — are close but not identical instruments, so the series should be read as a trend, not as one curve. These numbers say nothing about VPNs, SSH, email, messaging or data at rest. And nobody measures when a cryptographically relevant quantum computer will arrive: the regulatory dates above are declared policy, and Rosetta’s 2033–2039 window is a sealed judgment that will be graded, not a fact.

Rosetta Quantum publishes verdicts with raw, reproducible data. This is educational content, not a product claim, and it recommends no vendor.