atlas
trn1…
Try

3 of 1326 instances · cheapest is trn1.2xlarge at $1.344/hr

Instance Workload vCPUs Memory Local disk Network On-demand/hr ↑ Per vCPU
trn1.2xlargeGPU & accelerators 8 32 GiB 475 GB NVMe 12.5 Gbps $1.344 $0.1680
trn1.32xlargeGPU & accelerators 128 512 GiB 7.6 TB NVMe 800 Gbps $21.50 $0.1680
trn1n.32xlargeGPU & accelerators 128 512 GiB 7.6 TB NVMe 1600 Gbps $24.78 $0.1936