Post-Quantum Migration: The Small Vendors and Old Devices That Pay First

Archi3 min read

cryptographysecuritystandards

Contents
  1. The bill for post-quantum crypto arrives before the threat does
  2. The migration is a dependency problem, not an algorithm problem
  3. Who gets hit first
  4. The timeline is longer than the headlines suggest
  5. The asymmetry nobody budgets for

The bill for post-quantum crypto arrives before the threat does#

NIST finalized its first post-quantum standards in August 2024: ML-KEM (FIPS 203), ML-DSA (FIPS 204), and SLH-DSA (FIPS 205). Regulators in the US and EU have started asking vendors when they will migrate. The question in most coverage is whether the algorithms are safe. The better question is who absorbs the cost, and on what schedule.

My view: the burden falls on three groups that get the least attention. Small vendors with embedded crypto, operators of devices that cannot be updated, and the two or three engineers at any company who actually understand the key exchange. Large cloud providers will ship PQ support on their own timeline. Everyone else will follow, pay, or break.

The migration is a dependency problem, not an algorithm problem#

ML-KEM is fast. Benchmark figures published by academic groups and library maintainers put encapsulation in the range of tens of microseconds on ordinary server hardware. The math is not the hard part.

The hard part is that post-quantum keys and signatures are larger than the ones they replace. An ML-KEM-768 public key is 1,184 bytes. An ML-DSA-65 signature is 3,309 bytes. A classic RSA-2048 signature is 256 bytes. Every place that assumed a small certificate or a single TLS frame now has to handle more bytes, and some of those places are in firmware written a decade ago.

That gap shows up in handshakes that no longer fit in one round trip, in certificate chains that overflow a buffer, and in hardware security modules with fixed slots. Each of those is a project. Each project has a budget owner, and the budget owner is rarely the person who reads the NIST drafts.

Who gets hit first#

Small vendors shipping TLS. A startup with ten engineers can add a post-quantum library, but it also has to re-test every customer integration, update documentation, and field support tickets when an old client fails the handshake. Larger competitors can run both classical and post-quantum paths in parallel for a year. Smaller ones often cannot afford that overlap.

Anyone with long-lived hardware. Industrial controllers, medical devices, payment terminals, and automotive modules are deployed for 10 to 20 years. Many cannot run new crypto without a firmware update, and some cannot be updated at all. A 2024 survey by a European standards body found that a large share of embedded devices in the field had no supported path for cryptographic updates. Those operators face a choice between replacement and accepting risk.

The cryptography maintainers themselves. The people who write and audit the libraries everyone depends on are a small group. OpenSSL, BoringSSL, and the various language runtimes all need careful work to add new primitives without regressing the old ones. That work is funded unevenly. When a single maintainer is the only person who understands a code path, the schedule slips.

The timeline is longer than the headlines suggest#

Harvest-now, decrypt-later is a real concern for data with a long confidentiality life: health records, source code, diplomatic cables. But most organizations do not need to finish migration this year. They need to know where their long-lived secrets live and who controls the keys.

A practical sequence for most teams:

  1. Inventory where public-key crypto is used, including in firmware and third-party components.
  2. Flag anything with a lifespan past 2030 and classify the data it protects.
  3. Test hybrid key exchange in one production path before committing to a full rollout.
  4. Budget for the overlap period, because running both schemes costs money and engineering time.
  5. Write down who owns each system, because the migration will outlast several reorgs.

That last point matters more than it sounds. In my experience, crypto migrations stall when the person who understood the original design leaves and nobody else knows why a particular curve was chosen.

The asymmetry nobody budgets for#

Large cloud providers will absorb the transition because their customers demand it and they have the staff. Small vendors will absorb it because their customers demand it and they do not. Device operators will absorb it because the alternative is a recall.

The organizations that come out ahead will be the ones that treat post-quantum readiness as a maintenance line item, not a security initiative. That means funding library maintenance, paying for firmware update infrastructure, and keeping the inventories current. The algorithm transition is the easy part. The organizational transition is where the money goes.

More articles