Deploy AI models on H100, A100, H200, and AMD MI300X GPUs with up to 40% cost savings. Lightning-fast machine learning inference on enterprise GPU infrastructure.
Estimate training time and cost from model size, token count, and GPU choice — priced against Aquanode's live marketplace, per GPU / hr.
The math
Total FLOPs ≈ 6 × parameters × tokens (2 FLOPs/param/token forward + 4 backward — the standard dense transformer approximation; it overstates true FLOPs for mixture-of-experts models). Training time = total FLOPs ÷ (GPU's peak BF16 FLOP/s × assumed utilization). Cost = training time × the GPU's live $/hr rate.
Worked example: 7B params, 300B tokens, at assumed 40% utilization
GPU
Provider
Rate (per GPU / hr)
Estimated time
Estimated cost
A16
Vultr
$0.059/hr
3645.8 days
$5,163
V100
SimplePod
$0.070/hr
2916.7 days
$4,900
RTX 3060
SimplePod
$0.070/hr
3645.8 days
$6,125
A40
Vultr
$0.075/hr
3645.8 days
$6,563
RTX 4070
SimplePod
$0.080/hr
3645.8 days
$7,000
All figures assume a single GPU — a real multi-GPU cluster trains faster but is not linearly modeled here (interconnect overhead varies by topology).
Try your own numbers
1.26e+22
Total training FLOPs
40 TFLOP/s
Effective throughput
3645.8 days
Estimated training time
A16 (Vultr)
GPU used
$0.059/hr
Per-GPU rate
$5,163
Estimated cost (1 GPU)
Formula: FLOPs = 6 × parameters × tokens. Time = FLOPs ÷ (peak GPU FLOP/s × utilization). Cost = time × live per-GPU $/hr. Assumes a single logical accelerator — a real multi-GPU cluster scales throughput roughly linearly minus interconnect overhead, which is not modeled here.
