Billing & payments
GPU server SLA credits: how they work
Every server is covered by a 99.9% monthly availability commitment for power, network and hardware. If we fall short, you receive a credit of up to 100% of that server's monthly fee.
On this page
GPU server SLA credits are what you receive when one of your FullGPU servers falls below 99.9% monthly availability. The commitment covers power, network connectivity and hardware on every server you rent. Credits range from 10% to 100% of that server's monthly fee, and you request them through support within 30 days after the end of the affected month.
How much credit you receive
| Monthly availability | Credit (share of the server's monthly fee) | Downtime in a 30-day month |
|---|---|---|
| Below 99.9% | 10% | more than about 43 minutes |
| Below 99.5% | 25% | more than 3.6 hours |
| Below 99.0% | 50% | more than 7.2 hours |
| Below 95.0% | 100% | more than 36 hours |
For example, a single NVIDIA H100 at $2,190 per month that reaches 99.3% availability, about five hours of downtime in a 30-day month, is below 99.5% but not below 99.0%. Its credit is 25%, or $547.50.
How downtime is counted
Downtime is a period during which your server is unavailable because of a failure of power, network or hardware on our side. It is counted from when we detect the incident or you report it, until it is resolved.
Monthly availability is the total minutes in the month minus downtime, divided by the total minutes in the month. A 30-day month has 43,200 minutes, so 99.9% availability allows about 43 minutes of downtime.
Report an incident as soon as you notice it, through the contact form or the channel provided with your server. Downtime is counted from detection or from your report, so an early report keeps the record accurate. Our support guide lists what to include.
Failed GPUs and components are replaced as a priority. When a repair cannot be completed quickly, we offer an equivalent server where available.
What does not count as downtime
- Scheduled maintenance announced at least 48 hours in advance.
- Issues caused by your software, configuration or use, such as a driver you installed or a firewall rule that blocks SSH.
- Suspension under our terms of service or our acceptable use policy.
- Attacks exceeding the DDoS protection included with the service.
- Events outside our reasonable control.
Our engineers still help with problems in your own software stack. Those problems just do not count toward the SLA.
How to request a credit
Send a request through support within 30 days after the end of the affected month. Include the server's IP address and the dates and times of the incident, with your time zone. Credits are applied to your next invoice, or refunded if your service ends.
Under our terms of service, service credits are your primary remedy for unavailability. The full commitment, including bandwidth and support, is in our service level agreement.
Frequently asked questions
Are SLA credits paid out in cash?
Credits are applied to your next invoice. If your service ends, they are refunded instead.
Does planned maintenance count as downtime?
Not when it is scheduled maintenance announced at least 48 hours in advance. Downtime covers unavailability caused by a failure of power, network or hardware on our side.
Is a crash caused by my own software covered?
No. Issues caused by your software, configuration or use are excluded from the SLA, but our engineers can still help you troubleshoot through support.
Still stuck?
Send us the details — server IP, what you ran and any error output. An engineer replies 24/7.