Does Vercel have an SLA? The 99.99% Enterprise uptime commitment, its credit tiers, its two gaps, and how to claim
Vercel's Enterprise SLA targets 99.99% monthly uptime for the parts of the platform that serve your content. Misses pay 10%, 25% or 50% of the month's fee, capped at 50%, claimed in writing with log files within 30 days of becoming eligible. Hobby and Pro plans get no uptime SLA — and two narrow bands of downtime pay nothing at all.
When Vercel has a bad hour, it is rarely one team's problem — the marketing site, the app front end and the preview deployments can all go at once. The question that lands on the platform or FinOps lead afterwards is short: *does Vercel owe us anything?* The answer depends on your plan, on which part of Vercel failed, and — unusually — on exactly how many minutes it was down.
Does Vercel have an SLA with service credits?
Yes — but only on Enterprise. Vercel publishes an Enterprise Service Level Agreement (last updated 18 July 2024) under which it targets 99.99% uptime each calendar month, excluding Excused Downtime. If a month falls short, the credit is 10%, 25% or 50% of that month's fee depending on how far uptime fell, capped at 50% of the amount due for the month. You must give written notice within 30 days of becoming eligible, with log files showing the date and time of the outage — without them the credit is forfeited.
The Hobby and Pro plans have no published uptime SLA, so no credit is owed on them however long an outage runs.
What does the Vercel Enterprise SLA actually promise?
| Term | Vercel Enterprise SLA |
|---|---|
| Who it covers | Enterprise customers (the document is Vercel's *Enterprise* SLA) |
| Target | 99.99% uptime, excluding Excused Downtime |
| Wording | "commercially reasonable efforts" to make the platform available 99.99% of the time |
| Period | Each calendar month |
| Formula | Uptime ÷ (total minutes in the month − Excused Downtime) × 100 |
| Scope | Platform components that serve customer content to your websites and applications |
| Not in scope | Components that do not serve customer content — the SLA's own examples are the API and the CLI |
| Credit base | The month's fee for the Services |
| Credit form | Monetary credit applied to future use, issued within 60 days |
| How to claim | Written notice, with log files documenting the outage date and time |
| Deadline | Within 30 days of becoming eligible for a credit |
| Cap | 50% of the amount due for the Services for the month |
| Remedy | Credits are the sole and exclusive remedy |
Source: Vercel's published Enterprise Service Level Agreement, read on 8 October 2026; Ontracko's Vercel SLA profile carries the same tariff.
How much downtime does Vercel's 99.99% allow?
Very little. In a 30-day month (43,200 minutes, assuming no Excused Downtime), 99.99% leaves about 4.3 minutes. Vercel's credit table then prices the shortfall like this:
| Monthly uptime | Downtime in a 30-day month | Credit |
|---|---|---|
| 99.99% or above | Up to ~4.3 min | None — target met |
| Between 99.98% and 99.99% | ~4.3 to ~8.6 min | None — no row in the table |
| 99.1% to 99.98% | ~8.6 min to ~6.5 h | 10% |
| Between 99.0% and 99.1% | ~6.5 h to ~7.2 h | None — no row in the table |
| 95% to 99% | ~7.2 h to 36 h | 25% |
| Below 95% | More than 36 h | 50% |
The bold rows are not our invention. Vercel's table reads "99.1% to 99.98%", "95% to 99%" and "< 95%" — so a month landing strictly between 99.98% and 99.99%, or strictly between 99.0% and 99.1%, is below the 99.99% target and yet sits in no band. Ontracko reproduces the table as printed rather than smoothing it over; if your Order Form closes those gaps, it controls. For the general arithmetic, see what a 99.9% SLA actually allows.
What counts as Vercel downtime — and what doesn't?
Two filters decide whether an incident counts.
Scope: does the failing component serve your content? The SLA covers the platform serving customer content to your websites and applications, and expressly leaves out components that do not — naming the API and the CLI as examples. That matters, because many Vercel incidents are not content-serving incidents. A dashboard, analytics or observability outage, or an API-only problem, can be a genuinely bad day for your team while your production sites keep serving — and those minutes do not obviously count. A builds-only incident, where already-deployed sites stay up, is similarly hard to fit inside "serving customer content". Check the Vercel status page for which components were affected, not just that an incident happened.
Excused Downtime. Vercel excludes outages caused by the failure of any third-party vendor, the Internet in general, factors outside its reasonable control, your own (or third-party) software or hardware, and force majeure. An upstream provider's outage that takes Vercel with it is, on the document's wording, excused — and it comes out of the denominator as well as the downtime.
How to claim a Vercel SLA credit
- Confirm you are covered: you need an Enterprise agreement with Vercel. Hobby and Pro have no uptime SLA.
- Confirm the failing component served customer content to your sites or apps — not the API, the CLI, or another non-serving component.
- Rule out Excused Downtime: a third-party vendor failure, a general Internet problem, your own code or infrastructure, or force majeure.
- Add up the month's qualifying downtime and compute uptime against the month's minutes less Excused Downtime. Check which band it lands in — including whether it lands in one of the two unpriced gaps.
- Pull log files that show the date and time of the downtime from your own monitoring or request logs. The SLA makes them a condition of the claim.
- Send Vercel written notice within 30 days of becoming eligible for the credit, attaching the logs. The SLA does not define the eligibility date precisely, so the safe practice is to count the 30 days from the incident itself rather than from month-end.
- Check the credit lands: Vercel applies credits to future use within 60 days. They are not cash refunds.
A worked example
Say your Enterprise agreement costs $10,000 a month, and in a 30-day month your production site was not served for a cumulative 2 hours (120 minutes) of qualifying, non-excused downtime. Uptime is (43,200 − 120) ÷ 43,200 ≈ 99.72% — inside "99.1% to 99.98%" — so the tier is 10%, and the credit is 10% × $10,000 = $1,000, well under the 50% monthly cap.
Now say the same month had only 6 minutes of downtime. Uptime is (43,200 − 6) ÷ 43,200 ≈ 99.986% — below the 99.99% target, but above 99.98% — so it falls in the first gap and the table prices it at nothing. Short incidents are the common case, which is why the gap matters more than it looks.
The credit calculator on the Vercel SLA page runs the same table — gaps included — against your own spend.
Frequently asked questions
Does Vercel have an SLA on the Pro plan?
No. Vercel's published uptime SLA is its Enterprise Service Level Agreement. The Hobby and Pro plans carry no uptime commitment and no service credit, so an outage on those plans is not compensated under a published SLA.
What is Vercel's uptime guarantee?
On Enterprise, a 99.99% monthly uptime target for the platform components that serve your content, excluding Excused Downtime. Vercel's wording is "commercially reasonable efforts", and the commitment is backed by service credits rather than refunds.
How much is a Vercel SLA credit?
10% of the month's fee for uptime from 99.1% to 99.98%, 25% for 95% to 99%, and 50% below 95%. Credits for a month are capped at 50% of the amount due for that month.
Am I owed a credit if Vercel was down for two hours?
On Enterprise, two hours of qualifying, content-serving downtime in a 30-day month is about 99.72% uptime, which pays the 10% tier — provided the outage was not Excused Downtime and you file in time with log files. On Hobby or Pro, no credit is owed.
How long do I have to claim a Vercel credit?
Thirty days from the time you become eligible for the credit, by written notice with log files documenting the outage date and time. Missing the window or the logs forfeits the credit.
Does a Vercel dashboard or API outage count?
Usually not. The SLA covers components that serve customer content to your websites and applications, and names the API and CLI as examples of components it does not cover. If your production sites kept serving, the incident is unlikely to count.
Does Vercel refund money for outages?
No. Credits are monetary credits applied to future use, issued within 60 days, and they are the sole and exclusive remedy for a missed SLA.
Methodology & caveats
Every term above — the Enterprise scope, the 99.99% target and its "commercially reasonable efforts" wording, the uptime formula, the content-serving scope and its API/CLI examples, the Excused Downtime list, the 10/25/50% table as printed, the written-notice and log-file requirements, the 30-day window, the 60-day application, the 50% monthly cap and the sole-remedy clause — is transcribed from Vercel's published Enterprise Service Level Agreement (last updated 18 July 2024), read on 8 October 2026. Downtime minutes are our arithmetic on a 30-day month with no Excused Downtime; the two unpriced gaps follow directly from the table's printed boundaries. Your Enterprise agreement may vary these terms; where it does, it controls. Ontracko scores status-page incident duration, which is evidence a qualifying event occurred, not the uptime figure Vercel computes — which is why the claim needs your own logs. For context across vendors, see which SaaS vendors actually pay SLA credits, SLA claim deadlines by vendor, why SLA credit claims get denied, the reliability rankings and the SLA glossary.
*At 99.99%, a Vercel claim can turn on a few minutes — and on whether those minutes fell in a priced band at all. Ontracko watches the Vercel status feed for free, times every content-serving incident, and flags the month the moment it crosses a band, so the 30-day window doesn't close on you. Monitor Vercel free — 8% only on recovered credits. See Vercel live status or open the Vercel SLA page.*
Related reading
How to calculate your SLA credit: the formula, worked examples, and the mistakes that zero it out
An SLA credit is a percentage of the monthly fee for the affected service — not your whole bill — set by how far uptime fell below the commitment. Here's the three-step formula, a downtime cheat sheet, two worked examples (tiered and flat), and why a real breach sometimes still calculates to zero.
Does Supabase have an SLA? The 99.9% Enterprise uptime commitment, its credit tiers, and how to claim
Supabase's SLA commits to 99.9% availability per calendar month — but only for Enterprise customers on an Order Form. Below it, credits run 10% to 30% of the affected service's monthly fees, capped at 20% of a year's fees, and must be emailed within 30 days of month-end with 5-minute-interval evidence.
Does OpenAI or Anthropic owe you an SLA credit when the API goes down?
OpenAI prints a 99.9% uptime SLA on its Scale Tier and Fast mode pages — for Enterprise customers only — but publishes no credit schedule, no uptime definition and no claim window. Anthropic publishes nothing at all and disclaims uninterrupted service by name. Here is what each one actually owes you.
Why do SLA credit claims get denied? The six exclusions that void a claim
A real outage is not the same thing as claimable downtime. Six exclusion families — plan tier, scheduled maintenance, your architecture, minimum-duration floors, third-party carve-outs and your own configuration — decide whether a vendor pays. Here is how each one works, with the vendors that use it.
Or browse the SLA glossary and the reliability rankings.
Stop leaving SLA credits on the table
Ontracko monitors your vendors, catches every SLA breach, and drafts the claim. Free — 8% only on recovered credits.
Start monitoring free