The Canyon ReviewAuth providers, priced at the MAU cliff
Auth0, Clerk, Supabase Auth, Cognito, FusionAuth and Keycloak priced at 1k, 10k and 100k monthly active users. The cliffs are the story.
At 1,000 users everything is free and the choice barely matters. At 100,000 users the same choice spans two orders of magnitude on the invoice. Auth pricing is cliff pricing; this teardown maps the cliffs. Short verdict: Clerk for velocity, Supabase Auth or Cognito for price, FusionAuth or Keycloak when you'd rather own it.
Criteria: cost at 1k/10k/100k MAU 45%, integration speed 25%, migration-out story 20%, enterprise-feature ransom 10%. List prices as of June 2026.
The table
| Rank | Provider | 10k MAU | 100k MAU, roughly |
|---|---|---|---|
| 1 | Supabase Auth | in the $25 plan | low hundreds |
| 2 | Cognito | ~$0 (50k free) | low hundreds |
| 3 | Clerk | ~$25+ | four figures |
| 4 | FusionAuth | self-host free | flat, your infra |
| 5 | Keycloak | $0 | your platform team's time |
| 6 | Auth0 | free tier ends | call, and bring a chair |
Reading the cliffs
Cognito's 50,000 free MAU remains the quiet giant of the category, per the pricing page. The developer experience is the tax: hosted UI styling, quirky APIs, and enough sharp edges that teams buy Clerk anyway. That purchase is rational; just know you're paying for ergonomics, not capability.
Clerk earns its rank by making the first week effortless: drop-in components, sane session handling, organizations that work. The per-MAU meter past the free tier is where you should model growth with real numbers, because B2C apps with 100k casual users will feel it.
The enterprise-feature ransom deserves its own paragraph. SAML SSO and SCIM live in top tiers across the industry; that's the pattern sso.tax has cataloged for years. If your roadmap says "enterprise customers, eventually," check what tier the letters S-A-M-L sit in before you integrate anything.
The skip
Auth0, for new small products. Once the category's default answer, its post-acquisition pricing became a cliff face: modest MAU counts route to sales conversations, and features migrate tierward. Existing integrations work fine; there's no urgency to leave. But starting fresh in 2026 with this table in hand, you can get the same OIDC flows cheaper, or the same polish from Clerk with public numbers.
Migration honesty, since auth is the stickiest dependency you'll pick: export your password hashes' feasibility varies wildly by provider. Ask that question before user one, not user 100,000.