GPU.ai
GPU compute marketplace, secure and community tiers
GPU.ai is a GPU compute aggregator that routes launches across 14+ upstream providers rather than operating its own fleet. Pricing below is independently tracked directly from GPU.ai's own pricing API.
View on GPU.ai →GPU.ai is a GPU compute aggregator/broker, not a dedicated GPU cloud with its own fleet. It routes each launch to whichever of its 14+ upstream providers can fulfill the request, which is why listings for the same GPU model vary by region and by tier rather than following one provider-wide rate. Secure tier means the host is a datacenter machine; community tier means the host is peer-owned hardware. Neither tier is consistently cheaper than the other — for some GPU models secure is cheaper than community, the reverse of the usual pattern.
| GPU Model | Region | On-Demand |
|---|---|---|
| RTX A4000 16GB (Community) | Asia-Pacific | $0.10 |
| RTX A4000 16GB (Secure) | EU-West | $0.12 |
| RTX 3080 10GB | US-East | $0.15 |
| RTX 3090 24GB (Community) | EU-East | $0.17 |
| RTX 4000 Ada 20GB (Community) | EU-West | $0.20 |
| RTX 4080 16GB | US-East | $0.27 |
| RTX A5000 24GB | Canada-Central | $0.27 |
| RTX 4000 Ada 20GB (Secure) | EU-West | $0.28 |
| RTX PRO 4500 32GB (Community) | EU-West | $0.33 |
| L4 24GB (Community) | EU-East | $0.34 |
| RTX 4090 24GB (Community) | Canada-Central | $0.34 |
| RTX 5080 16GB | Canada-Central | $0.34 |
| RTX A6000 48GB (Secure) | Canada-Central | $0.40 |
| L4 24GB (Secure) | India-South | $0.44 |
| RTX 5090 32GB (Community) | US-West | $0.48 |
| A40 48GB | Canada-Central | $0.49 |
| RTX 3090 24GB (Secure) | EU-East | $0.50 |
| RTX A6000 48GB (Community) | EU-East | $0.61 |
| RTX 6000 Ada 48GB (Secure) | US-Central | $0.65 |
| L40 48GB (Community) | US-Central | $0.69 |
| RTX PRO 4500 32GB (Secure) | EU-East | $0.72 |
| A100 40GB (Community) | EU-East | $0.74 |
| RTX 4090 24GB (Secure) | EU-East | $0.74 |
| RTX 6000 Ada 48GB (Community) | US-Central | $0.74 |
| L40 48GB (Secure) | US-Central | $0.78 |
| L40S 48GB (Community) | EU-West | $0.79 |
| L40S 48GB (Secure) | US-Central | $0.80 |
| A100 80GB (Community) | Asia-Pacific | $0.81 |
| A100 40GB (Secure) | India-South | $0.89 |
| RTX 5090 32GB (Secure) | EU-East | $0.99 |
| A100 80GB (Secure) | Canada-Central | $1.07 |
| RTX PRO 6000 96GB (Community) | EU-East | $1.69 |
| H100 PCIe 80GB | Canada-Central | $1.98 |
| RTX PRO 6000 96GB (Secure) | US-Central | $1.99 |
| H100 NVL 94GB (Community) | US-Central | $2.59 |
| H100 NVL 94GB (Secure) | US-Central | $3.19 |
| H100 SXM 80GB (Community) | EU-East | $3.36 |
| H100 SXM 80GB (Secure) | Asia-Pacific | $3.49 |
| H200 141GB (Community) | EU-West | $3.59 |
| H200 141GB (Secure) | India-South | $3.99 |
| B200 192GB (Community) | US-West | $6.26 |
| B200 192GB (Secure) | US-West | $6.79 |
| B300 288GB | EU-West | $7.89 |
Rates shown are the cheapest listed offering per GPU model and tier at fetch time — see the region note below. regularly tracked · last tracked 2026-09-23 · confirm current rates before procurement.
GPU rate covers compute and the bundled instance disk only. Additional disk beyond the baseline is billed separately per GB-hour where the listing marks disk as configurable.
The vCPU and RAM figures behind each row above describe the specific cheapest offering at fetch time, not a fixed provider-wide bundle shape, since this is a live marketplace, not a fixed instance catalog.
Region is a real, priced field on every GPU.ai listing, not a hidden routing detail — but coverage is sparse and asymmetric across GPU.ai's 9 tracked regions. A given GPU model and tier is typically only listed in a handful of those regions at any time, not all of them, so this isn't a standard IaaS region selector where every provider region carries every GPU. GPU.ai's own marketing describes automatic best-price routing across its upstream providers; that most likely governs which upstream host fulfills a request within a chosen region, not whether region itself is selectable.
GPU.ai is independently listed on GPUAdvisor. Pricing above is drawn directly from GPU.ai's own pricing API. Listings, prices and coverage are never influenced by commercial relationships. GPU.ai's public pricing page could not be cross-checked against the API figures above — it is fully client-rendered and returned no price data to a direct fetch; this is a blocked verification, not a skipped one.