NVIDIA H100 80GB (form factor not stated) — what a GPU-hour costs
Listed rates, not availability. Read . How to read these figures ↓
Not enough monthly history for the H100 80GB (form factor not stated) to state a trend yet.
Size your workload
Who is cheapest today, and what a month costs.
GPUsHours per monthRegionH100 80GB (form factor not stated) by provider
Every listed rate on every market, cheapest first — the top card may be capacity that can be reclaimed mid-run.
| Provider · market and risk · instance, region | $ / GPU-hour | Read | Vs. median | 7 days | Listed since | A month at the selected workload | Source |
|---|---|---|---|---|---|---|---|
| CloudRiftspecialistreserved · low riskH100 · global | $1.70 | captured | −39% vs median | 7d — | since | $9,928 for the selected workload | Deploy ↗ |
| Gcorespecialistserverless · medium risk1×H100 · Everywhere Inference · €1.53 listed · ECB 1.1355 on September 29, 2026 | $1.74 | captured | −38% vs median | 7d — | since | $10,146 for the selected workload | Deploy ↗ |
| Sestercespecialiston-demand · low riskfrom | $1.83 | captured | −34% vs median | 7d no change | since | $10,687 for the selected workload | Deploy ↗ |
| GMI Cloudspecialiston-demand · low riskfrom | $2.00 | captured | −28% vs median | 7d no change | since | $11,680 for the selected workload | Deploy ↗ |
| Deep Infraspecialiston-demand · low riskH100 80GB · global | $2.20 | captured | −21% vs median | 7d — | since | $12,848 for the selected workload | Deploy ↗ |
| Daytonaspecialistspot · high riskNvidia H100 · GPU only | $2.27 | captured | −19% vs median | 7d — | since | $13,257 for the selected workload | Deploy ↗ |
| Vultrspecialistspot · high riskvbm-112c-2048gb-8-h100-gpu · global | $2.30 | captured | −18% vs median | 7d — | since | $13,432 for the selected workload | Deploy ↗ |
| Koyebspecialistserverless · medium riskH100 · global | $2.50 | captured | −10% vs median | 7d no change | since | $14,600 for the selected workload | Deploy ↗ |
| EmpirioLabsspecialiston-demand · low riskH100 80GB · global | $2.60 | captured | −7% vs median | 7d — | since | $15,184 for the selected workload | Deploy ↗ |
| Gcorespecialiston-demand · low risk1×H100 · Virtual GPU · €2.36 listed · ECB 1.1355 on September 29, 2026 | $2.68 | captured | −4% vs median | 7d — | since | $15,650 for the selected workload | Deploy ↗ |
| E2E Networksspecialiston-demand · low riskNVIDIA H100 80 GB · global | $2.69 | captured | −4% vs median | 7d no change | since | $15,710 for the selected workload | Deploy ↗ |
| Massed Computespecialiston-demand · low riskH100 (80GB) · global | $2.73 | captured | −2% vs median | 7d no change | since | $15,943 for the selected workload | Deploy ↗ |
| Northflankspecialiston-demand · low riskNVIDIA H100 80GB · GPU only | $2.74 | captured | −2% vs median | 7d no change | since | $16,002 for the selected workload | Deploy ↗ |
| Lyceumspecialiston-demand · low riskH100 VM · global | $2.79 | captured | 0% vs median | 7d — | since | $16,294 for the selected workload | Deploy ↗ |
| Cerebriumspecialistserverless · medium riskH100 · GPU only | $3.40 | captured | +22% vs median | 7d no change | since | $19,847 for the selected workload | Deploy ↗ |
| Leafcloudspecialiston-demand · low riskams-1 · €3.15 listed · ECB 1.1355 on September 29, 2026 | $3.58 | captured | +28% vs median | 7d — | since | $20,889 for the selected workload | Deploy ↗ |
| immers.cloudspecialiston-demand · low riskh100-4.16.128.120 · global · 328.16 RUB listed · Bank of Russia 0.011844 on September 30, 2026 | $3.89 | captured | +39% vs median | 7d — | since | $22,699 for the selected workload | Deploy ↗ |
| Daytonaspecialiston-demand · low riskNvidia H100 · GPU only | $3.95 | captured | +42% vs median | 7d — | since | $23,068 for the selected workload | Deploy ↗ |
| falspecialistserverless · medium riskH100, 80GB · global | $4.50 | captured | +61% vs median | 7d — | since | $26,280 for the selected workload | Deploy ↗ |
| Hugging Face Inference Endpointsspecialiston-demand · low riskNVIDIA H100 x1 (aws) · aws | $4.50 | captured | +61% vs median | 7d — | since | $26,280 for the selected workload | Deploy ↗ |
| Replicatespecialiston-demand · low riskgpu-h100 · global | $5.49 | captured | +97% vs median | 7d — | since | $32,062 for the selected workload | Deploy ↗ |
| Basetenspecialiston-demand · low riskH100, 80 GiB VRAM · global | $6.50 | captured | +133% vs median | 7d no change | since | $37,959 for the selected workload | Deploy ↗ |
| Cloud.ru Evolutionspecialiston-demand · low riskВиртуальная машина с GPU 52 vCPU/472 Gb RAM/2 GPU H100 NVLink · global · 854.00 RUB listed · Bank of Russia 0.011844 on September 30, 2026 | $10.12 | captured | +263% vs median | 7d — | since | $59,072 for the selected workload | Deploy ↗ |
Add your GPU cloudPrices come from your public pricing page, never from this form. Listing is free and can’t be bought. How providers are listed, and who is not, and why | |||||||
Per single GPU-hour; node prices divided by GPU count. Tap a provider to open its source page. Vs. median compares every market against the median rate card above — a spot rate is cheaper because it can be taken back, not because the provider is cheaper. Risk is read off the market, not off the provider: spot and community capacity can be reclaimed mid-run, serverless is shared capacity, billed by the second, a shared GPU is a card other tenants use too, on-demand and committed capacity stays yours while it runs. Community capacity and shared GPUs never count as on-demand: they are out of every median and every “cheapest”. Since is the day that figure was first listed unchanged — for a quote we have never seen change, the day of our first read of it. 7d is the listed figure against the same quote seven days earlier; it appears once a quote has been read for a week.
How to read these figures
We read published price lists, not availability: a listed rate can be sold out, waitlisted or absent from the region you need, and whether a card is free to rent is not something we measure.
Oldest read behind these figures: , from this build’s own hourly run. A quote not read again within 48 hours leaves every figure.
Hopper · 80 GB VRAM. Publicly listed rental prices at 63 providers, normalised to a single GPU-hour, plus what the card costs to buy. Every figure links to the page it was read from.
Today’s GPU-hour
Median rate card per card: each provider once, at its cheapest on-demand rate, community capacity and marketplace floors out — the card median every page on this site prints. The H100 SXM is headed by the Price Index instead, which has rules of its own (USD listings only). $/GB divides it by the card’s memory; spread is the dearest rate over the cheapest, with a dot per provider between them. Tap a card to open its page.
Memory| Card, $ per GB, spread and providers | Where each provider sits between cheapest and dearest | Median rate card, cheapest and dearest |
|---|---|---|
| GB300 NVL72288GB$0.048/GB · 1.8× spread · 2 providers | $13.91$9.82–$18.007d +3.1% | |
| GB200 NVL72186GB$0.071/GB · 3.4× spread · 4 providers | $13.25$8.00–$27.047d no change | |
| B300288GB$0.028/GB · 3.6× spread · 18 providers | $7.94$4.89–$17.807d −1.4% | |
| B200192GB$0.035/GB · 4.4× spread · 27 providers | $6.79$3.69–$16.117d −0.7% | |
| MI355X288GB$0.021/GB · 2.9× spread · 3 providers | $5.99$2.95–$8.607d no change | |
| MI350X288GB$0.018/GB · 1.5× spread · 2 providers | $5.08$4.00–$6.167d +4.7% | |
| H200141GB$0.031/GB · 5.1× spread · 38 providers | $4.38$2.09–$10.607d −0.6% | |
| H100 NVL94GB$0.046/GB · 2.6× spread · 7 providers | $4.32$2.92–$7.667d no change | |
| GH20096GB$0.040/GB · 2.8× spread · 3 providers | $3.87$2.29–$6.507d no change | |
| H200 NVL141GB$0.027/GB · 1.5× spread · 3 providers | $3.79$3.62–$5.60 | |
| H100 SXM80GB$0.044/GB · 6.2× spread · 29 providers | $3.49$1.79–$11.067d −0.1% | |
| H100 PCIe80GB$0.038/GB · 3.6× spread · 8 providers | $3.03$1.83–$6.507d no change | |
| MI300X192GB$0.015/GB · 3.5× spread · 8 providers | $2.83$1.71–$6.007d no change | |
| H100 80GB (form factor not stated)80GB$0.035/GB · 5.8× spread · 19 providers | $2.79$1.74–$10.127d no change | |
| V10016GB$0.168/GB · 15.9× spread · 8 providers | $2.69$0.19–$3.067d +0.8% | |
| MI325X256GB$0.009/GB · 1.7× spread · 3 providers | $2.25$2.25–$3.807d no change | |
| RTX PRO 6000 Blackwell96GB$0.023/GB · 5.0× spread · 27 providers | $2.20$1.09–$5.507d +0.3% | |
| A100 40GB40GB$0.050/GB · 3.8× spread · 16 providers | $1.99$0.89–$3.407d +0.1% | |
| A100 80GB80GB$0.023/GB · 5.7× spread · 40 providers | $1.81$0.89–$5.047d +0.1% | |
| RTX PRO 6000 Blackwell Max-Q96GB1.0× spread · 1 provider | one rate card, no median$1.80–$1.80 | |
| L40S48GB$0.029/GB · 4.6× spread · 28 providers | $1.40$0.76–$3.517d +0.2% | |
| RTX PRO 4500 Blackwell32GB$0.036/GB · 3.5× spread · 4 providers | $1.16$0.72–$2.52 | |
| V100 32GB32GB$0.035/GB · 4.1× spread · 3 providers | $1.13$0.28–$1.15 | |
| A10G24GB$0.046/GB · 1.2× spread · 2 providers | $1.11$1.01–$1.21 | |
| A1024GB$0.046/GB · 9.9× spread · 7 providers | $1.10$0.32–$3.207d no change | |
| A4048GB$0.022/GB · 3.0× spread · 4 providers | $1.04$0.49–$1.497d no change | |
| RTX 6000 Ada48GB$0.021/GB · 2.0× spread · 6 providers | $1.02$0.79–$1.577d +1.0% | |
| RTX PRO 5000 Blackwell48GB1.0× spread · 1 provider | one rate card, no median$0.96–$0.96 | |
| L4048GB$0.019/GB · 1.9× spread · 6 providers | $0.93$0.79–$1.507d no change | |
| RTX 509032GB$0.027/GB · 2.7× spread · 8 providers | $0.85$0.55–$1.517d no change | |
| RTX 5000 Ada Generation32GB1.0× spread · 1 provider | one rate card, no median$0.83–$0.83 | |
| L424GB$0.033/GB · 2.7× spread · 17 providers | $0.80$0.43–$1.177d no change | |
| Quadro RTX 600024GB$0.029/GB · 4.6× spread · 3 providers | $0.69$0.33–$1.50 | |
| RTX A600048GB$0.012/GB · 3.1× spread · 12 providers | $0.60$0.35–$1.097d +0.5% | |
| GeForce RTX 508016GB1.0× spread · 1 provider | one rate card, no median$0.59–$0.59 | |
| GeForce RTX 309024GB$0.023/GB · 2.1× spread · 4 providers | $0.55$0.35–$0.75 | |
| T416GB$0.033/GB · 3.5× spread · 9 providers | $0.53$0.23–$0.817d no change | |
| RTX 4000 Ada20GB$0.026/GB · 2.7× spread · 3 providers | $0.52$0.28–$0.76 | |
| A3024GB$0.020/GB · 4.1× spread · 6 providers | $0.48$0.33–$1.357d no change | |
| RTX 4000 SFF Ada20GB$0.024/GB · 1.1× spread · 2 providers | $0.47$0.44–$0.50 | |
| GeForce RTX 3090 Ti24GB1.0× spread · 1 provider | one rate card, no median$0.46–$0.46 | |
| RTX 409024GB$0.019/GB · 2.5× spread · 9 providers | $0.45$0.39–$0.997d no change | |
| GeForce RTX 308010GB1.0× spread · 1 provider | one rate card, no median$0.45–$0.45 | |
| RTX A500024GB$0.018/GB · 5.1× spread · 5 providers | $0.44$0.27–$1.387d no change | |
| RTX A400016GB$0.016/GB · 5.1× spread · 5 providers | $0.25$0.15–$0.767d no change |
History
No weekly history for the H100 80GB (form factor not stated) yet: the reconstructed catalog series does not carry this card, and our own series is shorter than two weeks.
Buy vs rent
No retail median from three or more sellers for the H100 80GB (form factor not stated), so no break-even is published.
Hyperscalers vs specialists
Hyperscalers do not list this card at all — the market belongs to the specialists.
The cost of not looking
Overpaying at 8 GPUs for the selected workload — dearest provider instead of cheapest. Every card against every provider: the spread table.
| H100 SXM | $54,143 | |
|---|---|---|
| H200 | $49,698 | |
| A100 80GB | $24,236 | |
| L40S | $16,060 | |
| RTX 6000 Ada | $4,555 | |
| RTX 4090 | $3,504 |
How it is computed
What we measure
Publicly listed rental prices for 53 GPU models (NVIDIA and AMD) at 63 providers — 709 current quotes across seven markets — and the retail price of the cards.
Where it comes from
Each provider's own pricing page or API; retail from reseller pages. Every figure links back to its source.
How it is normalised
Node prices are divided by GPU count, per-second rates multiplied out, a price listed in another currency shown as listed, beside its USD figure at the dated official rate (ECB; the Bank of Russia for RUB and VND). A reservation published as a total for its term is divided by the hours in that term, 8,760 per year.
What is modelled, not measured
One figure: cost per million output tokens, a listed rate divided by the throughput NVIDIA published for that card, printed with the run that produced it. Risk is not a model — it is read off the market a quote is from.
What we never do
Never estimate a price, never convert a currency silently, never publish a number a reader cannot verify on the linked page.
709 quotes · 63 providers. Seven files under CC BY 4.0; history is stored append-only and never rewritten.
As of , the cheapest listed rate for the NVIDIA H100 80GB (form factor not stated) is $1.70 per GPU-hour on CloudRift (reserved). The cheapest on-demand rate is $1.83 on Sesterce (from). Across 19 providers, the median rate card — each provider’s cheapest on-demand rate, community capacity and marketplace floors out — is $2.79 per GPU-hour, and the middle half of those rate cards sits between $2.55 and $4.22. Between rate cards, on-demand rates for the same card run from $1.74 (Gcore) to $10.12 (Cloud.ru Evolution), 5.8× apart across 19 providers. By market, the cheapest listed rates are $2.27 spot (Daytona, GPU only), $1.83 on-demand (Sesterce, from) and $1.70 reserved with no stated term (CloudRift). With 80 GB of VRAM, that is $0.023 per GB of memory per hour at the cheapest on-demand rate and $0.035 at the median.
As of , one provider prices the NVIDIA H100 80GB (form factor not stated) on-demand as a unit of 4 GPUs: $15.55 per unit-hour (immers.cloud, h100-4.16.128.120, $3.89 per GPU). 4 GPUs at the cheapest on-demand rate ($1.83 on Sesterce, from) come to $7.32 an hour, where a provider rents 4 to one account and they need not share a unit. A unit price is what the provider quotes for the 4 GPUs together; whether it also rents one is not part of the quote. 1 prices it in units of 2 — the rate table below names each quote's own unit.
Retail listings
No retail listings tracked for this card in the last 30 days.
Every rate in detail
One table per price type, never mixed: an on-demand rate and a spot rate answer different questions, and a marketplace floor is not a rate card. Prices are per GPU for one hour; the per-GB column divides by the card's 80 GB of VRAM.
On-demand
| Provider | USD / GPU-hour | 7-day change | Vs. median | USD / GB VRAM-hour | Type | Region | Min. commitment | Source, captured |
|---|---|---|---|---|---|---|---|---|
| Gcore specialist | $1.74 €1.53 listed · ECB 1.1355 on September 29, 2026 | — | −38% | $0.022 | on-demand | Everywhere Inference | — | Price page unchanged since first read, Sep 27, 2026 |
| Sesterce specialist | $1.83 | no change | −34% | $0.023 | on-demand from | — | — | Price page unchanged since first read, Sep 19, 2026 |
| GMI Cloud specialist | $2.00 | no change | −28% | $0.025 | on-demand from | — | — | Price page unchanged since first read, Sep 19, 2026 |
| Deep Infra specialist | $2.20 | — | −21% | $0.028 | on-demand | — | — | Price page unchanged since first read, Sep 27, 2026 |
| Koyeb specialist | $2.50 | no change | −10% | $0.031 | on-demand | — | — | Price page unchanged since first read, Sep 19, 2026 |
| EmpirioLabs specialist | $2.60 | — | −7% | $0.033 | on-demand | — | — | Price page unchanged since Sep 30, 2026 |
| Gcore specialist | $2.68 €2.36 listed · ECB 1.1355 on September 29, 2026 | — | −4% | $0.033 | on-demand | Virtual GPU | — | Price page unchanged since first read, Sep 27, 2026 |
| E2E Networks specialist | $2.69 | no change | −4% | $0.034 | on-demand | — | — | Price page unchanged since first read, Sep 19, 2026 |
| Massed Compute specialist | $2.73 | no change | −2% | $0.034 | on-demand | — | — | Price page unchanged since first read, Sep 19, 2026 |
| Northflank specialist | $2.74 | no change | −2% | $0.034 | on-demand GPU only | — | — | Price page unchanged since first read, Sep 19, 2026 |
| Lyceum specialist | $2.79 | — | ±0% | $0.035 | on-demand | — | — | Price page unchanged since first read, Sep 25, 2026 |
| Cerebrium specialist | $3.40 | no change | +22% | $0.042 | on-demand GPU only | — | — | Price page unchanged since first read, Sep 19, 2026 |
| Leafcloud specialist | $3.58 €3.15 listed · ECB 1.1355 on September 29, 2026 | — | +28% | $0.045 | on-demand | ams-1 | — | Price page unchanged since first read, Sep 26, 2026 |
| immers.cloud specialist | $3.89 $15.55/node ÷ 4 328.16 RUB listed · Bank of Russia 0.011844 on September 30, 2026 | — | +39% | $0.049 | on-demand | — | — | Price page unchanged since first read, Sep 28, 2026 |
| Daytona specialist | $3.95 | — | +42% | $0.049 | on-demand GPU only | — | — | Price page unchanged since first read, Sep 25, 2026 |
| fal specialist | $4.50 | — | +61% | $0.056 | on-demand | — | — | Price page unchanged since first read, Sep 23, 2026 |
| Hugging Face Inference Endpoints specialist | $4.50 | — | +61% | $0.056 | on-demand | aws | — | Price page unchanged since first read, Sep 25, 2026 |
| Replicate specialist | $5.49 | — | +97% | $0.069 | on-demand | — | — | Price page unchanged since first read, Sep 23, 2026 |
| Baseten specialist | $6.50 | no change | +133% | $0.081 | on-demand | — | — | Price page unchanged since first read, Sep 22, 2026 |
| Cloud.ru Evolution specialist | $10.12 $20.23/node ÷ 2 854.00 RUB listed · Bank of Russia 0.011844 on September 30, 2026 | — | +263% | $0.126 | on-demand | — | — | Price page unchanged since first read, Sep 28, 2026 |
Spot / interruptible
| Provider | USD / GPU-hour | 7-day change | USD / GB VRAM-hour | Type | Region | Min. commitment | Source, captured |
|---|---|---|---|---|---|---|---|
| Daytona specialist | $2.27 −$1.68 (−43%) vs its on-demand $3.95 | — | $0.028 | spot GPU only | — | — | Price page unchanged since first read, Sep 25, 2026 |
| Vultr specialist | $2.30 $18.40/node ÷ 8 | — | $0.029 | spot | — | — | Price page read from the provider’s API unchanged since first read, Sep 29, 2026 |
Reserved (committed use)
| Provider | USD / GPU-hour | 7-day change | USD / GB VRAM-hour | Type | Region | Min. commitment | Source, captured |
|---|---|---|---|---|---|---|---|
| CloudRift specialist | $1.70 | — | $0.021 | reserved | — | 1mo | Price page unchanged since first read, Sep 27, 2026 |
Every rate on this page as data: /gpu/h100/prices.json — the same fields as prices.json, narrowed to this card. CC BY 4.0, no key, no rate limit. On your own page: embed this card’s cheapest rates (/embed/gpu/h100, an iframe, no script).
Data updated: