Affiliate disclosure: HostFleet may earn a commission if you sign up through links on this page. That never changes the recommendation. Read the live HostFleet about page for methodology and affiliate-policy context. This is a source-backed price guide: every listed rate comes from an official provider page or the dated HostFleet dataset. It is not a benchmark or an inventory guarantee.

Rates checked: July 29, 2026
HostFleet comparison data: July 23, 2026 (Lambda H100 row added July 27)

H100 rental price per hour in 2026: cloud GPU rates compared

An H100 costs about $2.89 to $6.50 per GPU-hour in the current HostFleet comparison for public self-serve and managed deployment options. That range is useful only after you separate the products behind it: a dedicated RunPod Pod, a Lambda VM, and a serverless or managed deployment are not interchangeable purchases.

The lowest comparable entry in this table is RunPod Secure Cloud H100 PCIe at $2.89/hour. Lambda’s one-GPU H100 PCIe instance is $3.29/GPU-hour. Modal’s H100 is $0.001097/second, or $3.9492/hour before rounding. Those are published rates, not a promise that capacity will be available in a particular region or that the final bill contains only GPU time.

For the wider GPU market and the live data source behind this page, use HostFleet’s GPU pricing table. If the real decision is dedicated Pods versus workers that can return to zero, read RunPod Pods vs Serverless pricing before choosing from a rate card.

H100 price per hour: current comparison

The table is a live-data snapshot from gpu-pricing.json. Rates are USD per GPU-hour. A listed hourly rate is not automatically a like-for-like comparison: the scope column names what is being bought.

Important scope note: RunPod’s public page also shows a $1.99/hour Community Cloud rate for an H100 PCIe. HostFleet uses the $2.89/hour Secure Cloud rate in its like-for-like Pods row. Community and Secure Cloud have different availability and operating tradeoffs, so treating $1.99 as a universal market floor would be misleading.

Fal’s $3.99 number remains visible as a dated dataset entry rather than a newly verified rate. Its official public pricing page blocked an unauthenticated fetch on July 29. That is a source-access limitation, not evidence that the price changed; still, re-check it in a browser before using Fal in a purchase decision.

What a warm H100 costs per month

An hourly GPU price becomes a serious commitment when the deployment stays allocated. The following are GPU-only estimates, not provider quotes.

monthly GPU estimate = published hourly GPU rate × 720 hours

The assumption is one GPU held for 30 days, with no CPU, RAM, storage, network, taxes, or support charges. Under that assumption:

The Modal calculation uses its unrounded per-second rate: $0.001097 × 3,600 × 720 = $2,843.42. The other rows use the displayed rate multiplied by 720 and rounded to the nearest dollar. These examples explain the cost of keeping capacity allocated; they do not predict a workload that genuinely scales down between requests.

The cheapest number is not always the cheapest deployment

RunPod: the lowest Secure Cloud rate, with more infrastructure to own

At $2.89/hour for a Secure Cloud H100 PCIe, RunPod is the lowest currently verified rate in this table’s dedicated, self-serve scope. A Pod is a container-shaped machine: you choose the hardware and remain responsible for the image, server process, networking, observability, and stopping it when work ends.

That makes it a sensible starting point for a team that needs a continuously available H100 and can operate the endpoint. It is not a scale-to-zero default. The current public pricing page distinguishes Pods from Serverless, and it distinguishes Secure Cloud from Community Cloud. Use RunPod for inference APIs and jobs for the product-shape tradeoff, not just the rate.

Lambda: a VM rate with a defined instance shape

Lambda’s $3.29/GPU-hour figure is for a 1x H100 PCIe instance, not a bare GPU abstraction. Its public table attaches the GPU to an instance with CPU, RAM, and local storage. That can make planning simpler for an operator who wants a conventional VM, but it also means the number should not be compared as if every provider supplies the same surrounding resources.

Lambda also lists H100 SXM variants and larger multi-GPU instances at different per-GPU prices. This page deliberately uses its 1x H100 PCIe row because it is the closest match to the other single-GPU H100 entries.

Modal, Replicate, Baseten, and Fal: pay for a serving surface, not only a card

Modal and Replicate publish per-second H100 rates. Baseten publishes a per-minute managed-deployment rate. Those billing units are easy to convert to hourly numbers, but the customer is buying a deployment surface with its own container, scaling, and platform rules—not simply renting a VM.

• Modal is the cleanest fit when a Python-native workload can actually scale down. A warm H100 still consumes roughly $2,843 of GPU time in a 30-day month, so its serverless label does not make an always-ready endpoint cheap. See Modal pricing explained for the billing model.
• Replicate lists $5.49/hour for a private H100 deployment. Its published rate should be evaluated with the deployment’s scaling settings and model-serving constraints, not compared only with a raw VM price.
• Baseten lists $6.50/hour for an 80 GiB H100 deployment. The higher rate can be rational when its managed serving layer removes work that would otherwise be carried by an infrastructure team; it is not the budget choice for an idle endpoint. See Baseten pricing explained.
• Fal lists a $3.99/hour H100 rate in the dated HostFleet dataset, plus an advertised committed-use floor of $1.89/hour. Do not rank the committed-use figure against on-demand prices: it has a different commitment requirement, and the current public price page needs a fresh browser check.

For a broader explanation of why provider billing model matters as much as GPU hourly rate, see HostFleet’s serverless GPU pricing matrix.

Which H100 option fits the job?

Choose by operating model first, then compare the correct rate:

• You want a dedicated H100 and can run the stack: start by checking a RunPod Secure Cloud Pod. It is the lowest verified dedicated entry in this table, but its GPU bill continues while the Pod runs.
• You want a conventional GPU VM: compare Lambda’s 1x H100 PCIe instance with the actual CPU, RAM, storage, region, and capacity you need.
• You have bursty GPU jobs: compare Modal and RunPod Serverless using expected billable GPU-seconds and cold-start tolerance. Do not use 720-hour math for a job that truly returns to zero.
• You need a managed model-serving control plane: compare Baseten, Replicate, and Fal after defining minimum replicas, request patterns, and support needs. Their rate cards describe a different product than a self-managed Pod.

Verdict

For a continuously allocated, self-managed H100, RunPod Secure Cloud’s $2.89/hour H100 PCIe is the current lowest verified entry in this comparison. Lambda at $3.29/GPU-hour is the next clear single-instance alternative. If the workload can genuinely scale down, a serverless provider may cost less overall despite a higher hourly equivalent; if it needs a warm endpoint all month, the GPU-only floor is already roughly $2,000 to $4,700 before supporting infrastructure.

Treat that range as a capacity-budget decision, not a benchmark result. Verify exact GPU variant, cloud type, minimum-replica setting, storage, and regional availability before committing.

Sources

• RunPod pricing — Secure Cloud, Community Cloud, and Serverless H100 rates; checked July 29, 2026
• Modal pricing — H100 per-second rate; checked July 29, 2026
• Lambda GPU instances — 1x H100 PCIe and H100 SXM instance rates; checked July 29, 2026
• Baseten pricing — H100 80 GiB per-minute and hourly rate; checked July 29, 2026
• Replicate pricing — private H100 per-second and hourly rate; checked July 29, 2026
• Fal pricing — H100 list and committed-use pricing source; HostFleet source check July 23, 2026; unauthenticated access challenge observed July 29, 2026
• HostFleet GPU pricing dataset — /opt/hostbot-v2/src/data/gpu-pricing.json, refreshed July 23, 2026, with Lambda H100 data added July 27, 2026

Need a self-managed H100 endpoint? Using our affiliate link supports HostFleet’s testing budget at no extra cost to you: RunPod (+$5 credit on your first $10). Links are labeled, and source citations in this article are never affiliate links.