The first time a launch fails with a quota error, most people assume they did something wrong. They did not. Every major provider caps what a new account can run, and the cap is deliberately low. This guide explains what a quota is, why it exists, how vCPU-based limits are counted and what marketplace options labelled with a vCPU figure, such as “AWS vCPU32” and “OVH vCPU34 Account”, are describing.
What a service quota is
A quota, which AWS used to call a limit, is a ceiling on how much of a particular resource an account can hold at once. Quotas exist per service and often per region, and they cover far more than compute: numbers of IP addresses, storage volumes, load balancers and API request rates all have them. The quota that bites first is almost always compute, because virtual machines are the first thing anyone creates.
Quotas are not billing limits. You can have a valid payment method and a large credit balance and still be refused a launch because the regional vCPU ceiling is reached.
Why new accounts start small
Providers cap new accounts for three reasons. Capacity planning: unlimited launches from unknown accounts would make regional capacity unpredictable. Fraud and abuse control: stolen payment details are typically used to spin up as much compute as possible quickly, so low defaults limit the damage. Customer protection: a misconfigured script cannot run up an enormous bill in an hour if the account can only run a handful of machines.
The AWS fact page notes that default new-account limits can be as low as a handful of vCPUs, and that the number of concurrent EC2 instances or the availability of certain GPU instance families is often restricted until a quota increase is approved. Quotas grow as an account demonstrates legitimate, paid use.
How vCPU quotas are counted
Modern compute quotas are expressed in vCPUs rather than instance counts. A vCPU is one virtual processor thread as presented to the operating system. The provider sums the vCPUs of every running instance in a family and region and compares the total to the ceiling.
- A quota of 32 vCPUs could be one 32-vCPU machine, four 8-vCPU machines or sixteen 2-vCPU machines in that region and family.
- Stopped instances usually do not count; terminated ones never do.
- Specialized families such as GPU instances often have their own separate, and much lower, quota.
On AWS, quotas are viewed and requested through the Service Quotas console, per region and per instance family. On OVHcloud, public cloud projects carry their own quotas on instances, vCPUs, memory and storage, and increases are requested through the control panel or support. The listings’ vCPU labels refer to these ceilings.
What the listed vCPU options mean
The Buy AWS Account page lists “AWS vCPU32” and “AWS vCPU32 With CloudFront Enable”, along with “AWS Old vCPU32/64/128/256/512 Account” and Bedrock-oriented options such as “AWS vCPU512 10K RPM Bedrock Account”. The Buy OVH Account page lists an “OVH vCPU34 Account”. In each case the number describes an approved compute quota rather than a piece of hardware you are buying.
| Option label | What the figure describes | What it does not describe |
|---|---|---|
| AWS vCPU32 | An approved regional vCPU ceiling for standard instance families | Any running servers, credit balance or GPU quota |
| AWS vCPU512 10K RPM Bedrock Account | A larger vCPU ceiling paired with a Bedrock request-rate profile | Model access outside the listed profile |
| OVH vCPU34 Account | A public cloud quota expressed in vCPUs | Dedicated servers or VPS, which are ordered separately |
The AWS fact page adds an important caveat: confirm the actual quotas that matter to your workload, especially EC2 instance families, GPU types and regional availability, and treat console exports from the Service Quotas console as the most reliable evidence. Quotas are regional, so a ceiling in one region says nothing about another.
Requesting an increase the right way
Increase requests go through the provider and are reviewed against the account’s history. A request is more likely to be approved when it is specific and proportionate:
- State the exact quota, region and instance family.
- Explain the workload in plain terms and why it needs the requested amount.
- Ask for what you need in the next few months, not a theoretical maximum.
- Have a valid payment method attached and a clean billing history.
Requests take time and are not always granted immediately, which is the practical reason some teams look for accounts that already carry an approved ceiling. That is legitimate only when the account is held and operated within the provider’s terms; quotas belong to the account, and providers can review any account that appears to have changed hands improperly.
Designing around quotas instead of fighting them
Quotas are easier to live with when they are part of the plan. Spread non-critical workloads across regions so no single ceiling blocks everything. Request increases early, before a launch date. Use smaller instance sizes for development so the production family keeps its headroom. Track quota use alongside cost in your monthly review.
For a side-by-side look at the listed vCPU-labelled options, see vCPU quota accounts compared. The wider set of AWS options is decoded in AWS account types explained, and OVHcloud’s product families are covered in OVHcloud accounts: VPS, dedicated and public cloud. Further quota questions are answered in the AWS section of the cloud FAQ.
