PQC Cost Calculator
Post-quantum TLS is slower and heavier than what you run today. This tells you what that costs on your traffic, in dollars, before you commit to it.
Network and infrastructure leads, and CISOs sizing PQC across a TLS-heavy estate.
Vendors quote percentages on a benchmark you can't inspect. Nobody tells you what the percentage means on your bill.
A dollar figure from handshakes we measured daily on real hardware, plus the arithmetic that produced it.
What are you comparing?
Pick the key exchange you run today and the one you are considering. Every timing here is measured on real hardware — not modelled, not vendor-supplied.
How much traffic do you handle?
Only handshakes that run a full key exchange cost anything — resumed sessions skip it entirely. This is the biggest lever on the figure in step 4.
What do you pay for compute and bandwidth?
Defaults are AWS list prices. If you have negotiated rates, put them in — the answer moves with them. Press Enter or Calculate to apply.
What it costs
Your traffic, your rates, our measurements.
more than X25519, at 705 full handshakes/sec — 1.85B handshakes over 1 month.
$11.94 today against $193 after — a 16.16× change.
This is an ongoing operating cost for as long as the traffic runs, not a migration project fee.
The unit figure — multiply by your own volume if you would rather not trust the archetype.
What the same traffic costs you today, versus after.
What drives the difference
Every suite you selected
| Suite | Handshake | Bytes out | vs classical | CPU | Egress | Total / 1 month | × |
|---|---|---|---|---|---|---|---|
| X25519Classical | 286.1 µs | 32 B | baseline | $6.60 | $5.34 | $11.94 | 1.00× |
| X25519 + ML-KEM-768Hybrid | 266.1 µs | 1.09 KB | −7.0% | $6.14 | $187 | $193 | 16.16× |
How that number is worked out
No model, no hidden coefficient. Four multiplications, and you can check every one.
- 1Count the handshakes that actually happen1,500/sec minus the 53% that resume an existing session leaves 705/sec doing a real key exchange.
- 2Multiply by what one handshake costs in CPUWe measured 266.1 µs median against 286.1 µs for classical. The difference × your handshakes × your $/vCPU-hour.
- 3Multiply by what one handshake costs on the wire1.09 KB leaves your server per handshake instead of 32 B. That extra × your handshakes × your $/GB egress. Outbound only — inbound is free.
- 4Add the two, over your chosen periodAcross 1 month. That is the figure in step 4, and the split between the two terms is the bar above it.
The handshake timings are measured. Everything else is either a public figure we cite or a number you typed. Nothing is interpolated — there is no value between two of our runs.
Sources
Every figure this page depends on. The first entry changes with your selection.
Bank core-banking gateway · 53% session reuse · one month
The figures the calculator starts on, rendered statically. Everything above moves them.
| Suite | Handshake | Bytes out | CPU $/mo | Egress $/mo | Total $/mo | × |
|---|---|---|---|---|---|---|
| X25519 + ML-KEM-768 | 266.1 µs | 1.09 KB | $6.14 | $187 | $193 | 16.16× |
| P-256 + ML-KEM-768 | 215.0 µs | 1.13 KB | $4.96 | $192 | $197 | 16.53× |
| ML-KEM-768 | 61.4 µs | 1.06 KB | $1.42 | $182 | $183 | 15.32× |
| X25519 | 286.1 µs | 32 B | $6.60 | $5.34 | $11.94 | 1.00× |
It covers the TLS key exchange only — no certificate-chain signing, no packet-level effects, AWS list pricing. A real estate has negotiated rates, its own session-reuse behaviour, and a certificate chain that matters. This gets you the shape and the order of magnitude. It does not replace measuring your own.
Now do it against your actual infrastructure.
This page prices one key exchange against one set of assumptions. A migration decision needs the rest of it: which of your endpoints are actually negotiating what, how much of your traffic resumes rather than handshakes, what your certificate chain does to the picture, and where the cost lands across a whole estate rather than a single gateway.
We measure what your estate actually negotiates today, rather than assuming one uniform baseline across every endpoint.
Session-reuse rates and handshake volumes taken from your own telemetry, not an industry archetype we had to cite because we did not know you.
Certificate reissuance, HSM headroom, and the internal service-mesh multiplier that usually dwarfs the edge — priced together.
We do not sell migration tooling and we do not resell anyone’s PQC product. We measure, and we tell you what the measurements mean for your bill — which is the only reason to trust a number that came from a vendor’s competitor or from us.