Good question — this is a common issue when you try to launch GPU instances (like g4dn.xlarge) in AWS but get an error such as: “Instance limit exceeded” “You do not have sufficient quota

This is indeed a common issue when trying to launch GPU instances in AWS. There are a few key reasons why you might encounter these errors:

1. Default vCPU Limits: When you create an AWS account, AWS sets default limits on the number of instances you can run on a per-Region basis. These limits are separate for different instance types and purchasing options (On-Demand vs Spot).

2. Separate Quotas for Instance Types: GPU instances like g4dn.xlarge have specific quotas that are different from regular instances. For example, there are separate quotas for "All G and VT On-Demand Instances" versus "All G and VT Spot Instance Requests."

3. Regional Availability: GPU capacity can vary significantly by region and Availability Zone.

To resolve these issues, you can:

1. Request a Quota Increase: Go to the Service Quotas page in the AWS console, select Amazon EC2, and request an increase for the specific GPU instance family you need (like "All G and VT On-Demand Instances").

2. Check Your Request Type: Ensure you're requesting the right type of instance. If you received a quota increase for Spot Instances but are trying to launch On-Demand Instances, you'll still get an error.

3. Try Different Availability Zones: GPU capacity can vary across Availability Zones within the same region.

4. Consider Alternative Instance Types: If you urgently need GPU capacity, consider using a different GPU instance family that might have more available capacity.

5. Try a Different Region: If your workload allows, you might find better GPU availability in other AWS regions.

The quota increase process typically takes some time to complete, so plan accordingly if you need GPU instances for time-sensitive workloads.
Sources
EC2 instance creation - vCPU limit issue | AWS re:Post
Troubleshoot Amazon EC2 instance launch issues - Amazon Elastic Compute Cloud
aws ap-northeast-1 gpu instance is not work!!! help | AWS re:Post

You need to verify and if necessary request service quota limit increase

Go to Service Quotas console of your desired Region to verify On-Demand Instance quota value for G instance types, i.e. Running On-Demand G and VT instances. Value is likely to be 0 for your account