For CFO · CIO · COO · Procurement

AI Vendor Lock-In Cost.

"How locked-in are we, really?" Most exec teams find out at renewal — when the leverage conversation is already over. This calculator estimates the all-in cost of leaving one AI vendor today, plus a lock-in multiplier (switching cost ÷ annual fee) that sorts vendors into four bands: portable · committed · locked · captured. Pair with the SLA Negotiator for the renewal walkthrough and the Vendor Renewal Decision for the keep/replace verdict.

Inputs

All numbers are yours. Math runs in your browser; nothing leaves.

The base fee — what you'd save if you walked away tomorrow. Excludes implementation costs you already paid (those are sunk).
Your raw data + derived artifacts (embeddings, fine-tunes). If you can't estimate, ask the vendor's customer success team — they should know. If they don't, that's its own finding.
Cloud-standard is $0.05–$0.12/GB. Some AI vendors waive egress in their contract — set to 0 if your contract confirms it. Some charge egress + per-API-call retrieval fees — that's the high end.
240 hr
Eng/ML hours to retrain a fine-tune, rebuild a RAG corpus, port prompts, recompute embeddings on the new vendor. 80–500 hr typical for one model dependency; 1000+ hr if you have multiple fine-tunes or a custom pipeline.
3 mo
Engineering FTE-months to re-integrate the new vendor: SDK swap, auth flow, monitoring, eval harness, runbook updates. 1–6 FTE-months typical; longer if the vendor is woven through many products.
2 mo
Overlap where you pay both vendors during migration. 1-2 mo for a clean cutover; 3-6 mo for a phased migration with parallel traffic. Set to 0 only if you have a hard switchover (rare and risky).
Salary + benefits + overhead for the engineers doing the migration. US senior IC ≈ $200–260K. Use your real loaded rate, not a market-rate guess.
Verdict
You are

Move the sliders to see the math update.

All-in switching cost
$—
egress + retrain + retooling + parallel
Lock-in multiplier
—×
switching cost ÷ annual fee

Cost breakdown

Egress
$—
Retrain / rebuild
$—
Retooling
$—
Parallel-run
$—

The math, on the page

Four components, all transparent. Argue with the framework by running them yourself.

ComponentFormula
EgressegressGb × egressPricePerGb
Retrain / rebuildretrainHours × (fteCost / 2080)
RetoolingretoolingMonths × (fteCost / 12)
Parallel-rungraceMonths × (annualSpend / 12)
Total switching costegress + retrain + retooling + parallel
Lock-in multipliertotal / annualSpend

Verdict bands: < 0.5× portable · 0.5–1.5× committed · 1.5–3.0× locked · > 3.0× captured.

What this calculator deliberately doesn't model: opportunity cost of staying (the value a better vendor would deliver), brand/customer-trust cost of switching during a public renewal, regulatory cost of a re-attestation cycle on the new vendor. Use the number as a floor on the leverage conversation, not as a final verdict.

Note on egress assumption: Many AI vendors don't expose raw data egress as a line item — you may be migrating prompts + configurations + outputs but rebuilding embeddings from scratch on the new side. If that's your case, set egress GB to 0 and put the equivalent eng-hours into the retrain bar.