Which technology to choose for quantum safe migration? The choice matters now because the deadlines are no longer abstract.
CNSA 2.0 expects new US national security systems to be quantum-safe from January 2027, and the EU's coordinated roadmap asks critical infrastructure to finish its high-risk transition by 2030. Meanwhile, "harvest now, decrypt later" means traffic recorded today can be decrypted once a capable quantum computer exists. The links that carry your longest-lived data are already exposed.
Hybrid post-quantum cryptography (PQC) is the only one of the three routes to quantum-safe data in transit that is cheap to deploy, light to run, scales across a real network and has the backing of the security agencies. Quantum key distribution (QKD) and pre-shared keys (PSK) each solve part of the problem, but they move the cost somewhere else: into new physical infrastructure for QKD, and into your operations team for PSK.
Post-quantum cryptography. New public-key algorithms, such as ML-KEM (FIPS 203) and FrodoKEM, built on maths problems that quantum computers are not known to solve efficiently. In hybrid mode a PQC algorithm runs alongside a proven classical one, so the connection stays safe even if either is later broken. It is software: it runs on standard hardware, over any network, at any distance.
Quantum key distribution. Two endpoints exchange key material encoded in single photons over a dedicated optical fibre or free-space link. Any eavesdropping on that quantum channel disturbs the photons and can, in principle, be detected. QKD only produces keys. It still needs a classical or PQC mechanism to authenticate the endpoints, and a conventional cipher to encrypt the data.
Pre-shared keys. A symmetric secret is generated once, carried to both ends out-of-band, and mixed into the key exchange. Symmetric ciphers such as AES-256 resist known quantum attacks, so a well-kept PSK blunts harvest-now-decrypt-later. RFC 8784 standardises this for IPsec as an interim measure. The protection is only as good as the process that creates, moves, rotates and revokes every key.
The major Western security agencies agreed: migrate to PQC, treat QKD as a niche add-on at most, and use pre-shared keys only as a bridge.
| Body | On QKD | On PQC and PSK |
| NSA (US) | Does not recommend QKD for national security systems unless its limitations are overcome: partial solution, special hardware, trusted-relay insider risk, hard to validate, DoS-prone | CNSA 2.0 mandates PQC (ML-KEM-1024, ML-DSA-87) with AES-256 |
| NCSC (UK) | Does not endorse QKD for government or military use; notes it needs dedicated hardware and cannot authenticate on its own | PQC is the recommended mitigation |
| ANSSI, BSI, NLNCSA, Swedish Armed Forces (joint paper, 2024) | QKD is not yet mature enough for most use cases; at best a defence-in-depth complement in niche settings | PQC migration should be the priority |
| IETF RFC 8784 | – | Mixing PSKs into IKEv2 is a stop-gap until PQC key exchange is available |
The EU is investing in quantum communication infrastructure through EuroQCI, so QKD will exist in some government and research networks. That is a strategic, long-horizon programme. It does not change the near-term advice for enterprises and critical-infrastructure operators, which points to PQC.
PQC has the lowest total cost of ownership because it adds no new physical layer and no new manual process. QKD is the most expensive on both counts; PSK looks cheap until you count staff time.
PQC rolls out on the network you already have; QKD is a multi-year infrastructure project; PSK is quick for a handful of links and slows down with every site you add.
PQC effort grows with the number of sites; QKD and PSK effort grows with the number of site pairs, which is roughly the square of the sites.
For a full mesh of N sites, QKD needs N(N−1)/2 point-to-point quantum links and PSK needs N(N−1)/2 separate secrets. PQC needs N encryptors, each joining the network once, with keys negotiated automatically between any pair.
| Sites | PQC encryptors | QKD links or PSK secrets (full mesh) | PSK manual rotations a year (quarterly) |
| 5 | 5 | 10 | 40 |
| 10 | 10 | 45 | 180 |
| 20 | 20 | 190 | 760 |
| 50 | 50 | 1,225 | 4,900 |
Real networks are rarely full meshes, and hub-and-spoke designs cut the numbers. But they do so by routing through central points, which for QKD means more trusted relays and for PSK means a higher-value key store. Distance and bandwidth matter too: PQC works across any routed path and NQX encrypts at up to 100 Gbps, while QKD reach and key rates fall as fibre length grows.
Pre-shared keys do not work in large environments. We know this from experience. As the company behind the Secure Shell (SSH) protocol, SSH Communications Security has spent decades helping enterprises regain control of static keys that spread faster than anyone could track, rotate or revoke. Manually managed secrets break down well before a network reaches hundreds of links, and pre-shared keys for network encryption follow the same pattern.
The future-proof choice is the one that can change: a crypto-agile PQC platform absorbs new algorithms and new threats by software update, while QKD and PSK lock you into fixed physical or procedural setups.
| Criterion | Hybrid PQC (NQX) | QKD | Pre-shared keys |
| Agency recommendation | Primary control (NSA, NCSC, BSI, ANSSI) | Not recommended as a primary control | Interim stop-gap |
| Cost to deploy | Low: standard hardware, existing links | Very high: photonics, dedicated fibre, relays | Low |
| Cost to maintain | Low to moderate: algorithm lifecycle | Very high: physical upkeep plus authentication layer | High: manual rotation and audit |
| Complexity and effort | Low: transparent bump-in-the-wire | Very high: multi-year infrastructure project | Low at first, rising with each site |
| Scalability | Linear in sites, any distance, up to 100 Gbps | Per pair, distance-limited | Per pair, manual |
| Future-proofness | High: crypto-agile, hybrid hedge | Limited: implementation risk, still needs PQC | Limited: no forward secrecy |
Make hybrid PQC the default for every link that carries data with a long confidentiality lifetime, starting with the most exposed ones. Pre-shared keys don't work at scale, so keep them only where a legacy system can't yet do PQC, with a dated plan to retire them. Consider QKD only if a regulator or a specific national program requires it, and then as an extra layer on top of PQC, not instead of it.
SSH's NQX is built for exactly this: hybrid key agreement with FrodoKEM and ML-KEM alongside classical algorithms, Layer 2 and Layer 3 encryption at wire speed up to 100 Gbps, automated key generation and rotation, and a software-defined engine that takes new algorithms without new hardware. Explore NQX quantum-safe network encryption or try the interactive PQC vs QKD vs PSK comparison.
What is the difference between PQC and QKD? Post-quantum cryptography (PQC) is software. It uses new mathematical algorithms that quantum computers are not known to break, and it runs over any network. Quantum key distribution (QKD) is hardware. It uses quantum physics to share keys over dedicated fiber or satellite links. PQC protects data wherever it travels. QKD protects keys only across its own optical links.
Is QKD more secure than PQC? Not in practice today. QKD's theoretical guarantees depend on specialized hardware, trusted relay nodes and classical authentication underneath. In a January 2024 joint position paper, France's ANSSI, Germany's BSI and the Dutch and Swedish national security agencies concluded that QKD isn't mature enough and recommended prioritizing PQC. In October 2026 the NSA restated that it does not recommend QKD for national security systems.
Are pre-shared keys quantum-safe? Symmetric 256-bit pre-shared keys resist known quantum attacks, so yes, in principle. The problem is scale. A full mesh of N sites needs N(N–1)/2 keys, each distributed out of band, rotated and revoked by hand. That is manageable for a few fixed links and fragile across a growing network.
Can PQC, QKD and pre-shared keys be combined? Yes. Standards already allow mixing extra key material into a post-quantum handshake, for example IKEv2 post-quantum pre-shared keys (RFC 8784) alongside ML-KEM. Organizations with QKD infrastructure often add PQC as an independent second mechanism instead of relying on QKD alone.
Which approach should we choose? For most organizations, hybrid PQC is the default. It works over existing networks, scales to many sites and updates in software. Pre-shared keys fit small, fixed, high-assurance links. QKD makes sense only where dedicated optical infrastructure already exists, and then as an extra layer, not a replacement.
How does SSH NQX deliver quantum resistance? NQX uses hybrid post-quantum key exchange, ML-KEM and FrodoKEM alongside classical algorithms, in a software-based crypto engine. It runs over any network path without dedicated fiber or quantum hardware. New algorithms arrive as software updates.