Unable to create On-Demand Capacity Reservation for Instance Type p4de.24xlarge

The issue you're experiencing with creating an On-Demand Capacity Reservation for the p4de.24xlarge instance type is likely due to insufficient capacity in the selected Availability Zone. This is a common problem, especially for specialized and high-demand instance types like the p4de.24xlarge, which uses powerful GPUs.

Here are some steps you can take to address this issue:

1. Try different Availability Zones: Capacity can vary between Availability Zones within a region. Attempt to create the Capacity Reservation in other Availability Zones within your chosen region.

2. Check other regions: If your workload allows, consider creating the Capacity Reservation in a different AWS region where capacity might be more readily available.

3. Wait and retry: Capacity constraints can be temporary. You may want to wait for some time (possibly a few hours or even days) and then try again, as AWS continuously works on provisioning additional capacity.

4. Contact AWS Support: If you urgently need this specific instance type and capacity, reaching out to AWS Support might be helpful. They may be able to provide more detailed information about capacity availability or potentially assist with your specific needs.

5. Consider alternative instance types: If your workload allows some flexibility, you might want to explore other GPU-enabled instance types that could meet your requirements.

6. Look into Reserved Instances or Savings Plans: While this doesn't solve your immediate need, for long-term use, these options can provide a higher likelihood of capacity being available when you need it.

Remember that for high-demand, specialized instance types like the p4de.24xlarge, capacity can be limited and may fluctuate based on overall demand. It's always a good idea to plan ahead and consider backup options when working with such instance types.
Sources
Cannot use p4 instances | AWS re:Post
Insufficient capacity error while trying to create on-demand p4d instance | AWS re:Post

This error typically occurs due to insufficient capacity for that specific instance type (p4de.24xlarge) in the selected Availability Zone (us-east-1a). These are highly specialized GPU-backed instances, and availability can be limited due to high demand.

Here are some things you can try:

Workarounds and Recommendations: Try a different Availability Zone (AZ): In the same region (us-east-1), attempt to create the capacity reservation in other AZs like us-east-1b, us-east-1c, etc.

Try a different AWS region: Consider testing in GPU-heavy regions like us-west-2, us-east-2, or eu-west-1, which might have better capacity for ML/AI workloads.

Retry after a delay: Capacity issues can be temporary. AWS might free up capacity if other customers release their instances.

Open a support case with AWS: AWS Support can sometimes help you request capacity for high-demand instance types or provide visibility into planned availability.

Consider other instance types: If you're flexible, look into p4d.24xlarge, g5.48xlarge, or inf2 instances depending on your workload’s requirements.

Additional Notes: Keep in mind that Capacity Reservations are immediately billed at the On-Demand rate once provisioned—even if you don’t use the instance.

You can also explore EC2 Capacity Blocks or EC2 Spot blocks if your workload allows for flexibility.

Let me know if you have a strict workload requirement or timeline—I’m happy to help explore fallback options.