GPUs for training

As of November 5, 2026 — Cards with 80 GB or more: from $0.65 per GPU-hour on the NVIDIA RTX PRO 6000 Blackwell (Hostinger), 20 cards qualify. Rates are each card’s cheapest live on-demand rate card across the providers we track, read hourly, every one linked to the page it was read from.

Training a model from scratch spreads its weights, gradients and optimizer states - 16 bytes per parameter in mixed precision with Adam - over many GPUs, so the card is chosen for its memory and for the links between cards in a node, not for fitting the model alone; interconnect is not ranked here. Below, every card with 80 GB or more; each card page lists the providers that price a whole 8-GPU node.

Cards with 80 GB or more

A 7B model alone takes 7 billion × 16 bytes = 112 GB of training state, more than one 80 GB card holds: training is split across cards from the start.

Cards with 80 GB or more: cards with 80 GB or more, cheapest first, as of November 5, 2026.
Card, memory	Cheapest on-demand	Provider, read	Median rate card	$ per GB-hour	Deploy
NVIDIA RTX PRO 6000 Blackwell · 96 GB	$0.65 from	Hostinger captured Nov 5, 00:40 UTC	$2.20 42 providers	$0.007	Deploy ↗
NVIDIA GB10 (DGX Spark) · 128 GB	$0.79	GPUwerk captured Nov 5, 00:40 UTC	$0.82 2 providers	$0.006	Deploy ↗
NVIDIA A100 80GB · 80 GB	$0.92	Yotta Labs captured Nov 5, 00:40 UTC	$1.99 63 providers	$0.011	Deploy ↗
Intel Gaudi 2 · 96 GB	$1.10	Denvr Dataworks captured Nov 5, 00:40 UTC	$1.16 2 providers	$0.011	Deploy ↗
AMD Instinct MI300X · 192 GB	$1.71 from	TensorWave captured Nov 5, 00:40 UTC	$2.65 7 providers	$0.009	Deploy ↗
NVIDIA H100 80GB (form factor not stated) · 80 GB	$1.72	Gcore captured Nov 5, 00:40 UTC	$3.75 28 providers	$0.022	Deploy ↗
NVIDIA H100 SXM · 80 GB	$1.79 GPU only	Cudo Compute captured Nov 5, 00:40 UTC	$3.29 41 providers	$0.022	Deploy ↗
NVIDIA RTX PRO 6000 Blackwell Max-Q · 96 GB	$1.80	EmpirioLabs captured Nov 4, 01:40 UTC	$1.86 2 providers	$0.019	Deploy ↗
NVIDIA H100 PCIe · 80 GB	$1.98 from	GPU.ai captured Nov 5, 00:40 UTC	$2.82 10 providers	$0.025	Deploy ↗
AMD Instinct MI325X · 256 GB	$2.25	Bentaus captured Nov 5, 00:40 UTC	$2.25 3 providers	$0.009	Deploy ↗

10 more cards qualify, dearer: NVIDIA GH200 $2.29, NVIDIA H200 $2.29, NVIDIA H100 NVL $2.92, AMD Instinct MI355X $2.95, NVIDIA H200 NVL $3.29, AMD Instinct MI350X $4.00, NVIDIA B200 $4.50, NVIDIA B300 $5.89, NVIDIA GB200 NVL72 $6.94, NVIDIA GB300 NVL72 $13.30.

How these figures are read

A card qualifies by its memory alone; speed, interconnect and software support are not ranked here. The rate is the card’s cheapest live on-demand rate card - each provider once, at its cheapest on-demand rate, community capacity and marketplace floors out, read within 48 hours - and the median is taken over the same rate cards, as on every card page. Memory figures are floors from the arithmetic above, not measurements: context length, batch size and the serving stack change them. Methodology.

Sources

Rajbhandari et al., ZeRO: Memory Optimizations Toward Training Trillion Parameter Models (2019)

Other workloads: GPUs for LLM inference · GPUs for fine-tuning · GPUs for image generation · Cheapest GPU memory per hour.

Data updated: November 5, 2026