
Amazon EC2 Capacity Blocks now offer instant start times, block extensions, and up to six-month reservations for GPU and ML chip instances, providing more flexible machine learning infrastructure provisioning.


<div>
<p>
Amazon Web Services has announced three new features for EC2 Capacity Blocks for ML, designed to improve GPU and ML chip instance provisioning:
</p>
<ul>
<li>Near-instantaneous capacity access for ML workloads</li>
<li>Ability to extend Capacity Block durations during ongoing jobs</li>
<li>Option to reserve GPU/ML chip capacity for up to six months</li>
</ul>
<p>
Key capabilities include:
</p>
<ul>
<li>Reserve 1-64 instances (up to 512 GPUs or 1,024 Trainium chips)</li>
<li>Support for P5e, P5, P4d, and Trn1 instances</li>
<li>Available in multiple US and Asia Pacific regions</li>
</ul>
<p>
These enhancements provide more flexibility and responsiveness for machine learning workload capacity management.
</p>
</div>


Amazon EC2 Capacity Blocks now supports instant start times and extensions

Related articles

Related articles

Nov 22
2024
Amazon EC2 introduces provisioning control to launch instances on On-Demand Capacity

Mar 19
2026
Amazon EC2 Fleet now supports interruptible Capacity Reservations

Nov 24
2025
Amazon EC2 announces interruptible Capacity Reservations

Oct 17
2025
Announcing Amazon EC2 Capacity Manager