Lambda Labs GPU Pricing (September 2026)
Quick answer: in the September 12, 2026 snapshot, single-GPU A100 40GB is $1.99/hour, H100 PCIe is $3.29/hour, and H100 SXM is $4.29/hour. At 730 hours, that is $1,452.70, $2,401.70, or $3,131.70 before applicable tax—not a discounted monthly plan.
The illusion is that "Lambda A100 price" or "Lambda H100 price" refers to one interchangeable GPU. It does not: memory capacity, PCIe versus SXM, and GPU count change both the hardware and the published per-GPU rate.
Lambda's public instance page was reviewed on September 12, 2026; the rates below were unchanged from our July snapshot. Listed pricing does not confirm regional stock.
| Lambda instance | Published price per GPU-hour | Configuration note | Snapshot |
|---|---|---|---|
| 1x A100 SXM 40GB | $1.99 | Single-GPU instance | September 12, 2026 |
| 1x A100 PCIe 40GB | $1.99 | Single-GPU instance | September 12, 2026 |
| 8x A100 SXM 80GB | $2.79 | Per GPU; the instance contains eight GPUs | September 12, 2026 |
| 1x H100 PCIe 80GB | $3.29 | Single-GPU instance | September 12, 2026 |
| 1x H100 SXM 80GB | $4.29 | Single-GPU instance | September 12, 2026 |
The A100 80GB rate must not be presented as a single-GPU instance price: the public table exposes it in an 8x A100 SXM configuration and labels the price per GPU-hour. Renting that configuration costs 8 × $2.79 = $22.32 per instance-hour.
For simple 730-hour planning:
| Lambda example | Hourly basis | Approx. 730-hour cost |
|---|---|---|
| 1x A100 40GB | $1.99/GPU-hr | $1,452.70 |
| 8x A100 SXM 80GB | $22.32/instance-hr | $16,293.60 for the instance |
| 1x H100 PCIe 80GB | $3.29/GPU-hr | $2,401.70 |
| 1x H100 SXM 80GB | $4.29/GPU-hr | $3,131.70 |
What This Page Does and Does Not Claim
This page is intentionally narrow:
- it uses provider-published public pricing
- it does not freeze RunPod or Vast.ai marketplace quotes as if they were list prices
- it does not treat a hyperscaler whole-VM price as identical to a managed single-GPU listing
- it does not invent a Lambda rate for a GPU SKU that is not clearly published on the page reviewed
That is the cleanest way to answer Lambda pricing questions without letting the page turn into a stale quote graveyard.
Why Lambda Is Easier to Quote Than a Marketplace
Lambda is useful in pricing comparisons because it publishes fixed public list prices for self-serve GPU instances. That is very different from a marketplace where individual hosts can expose different offers at different times.
When people compare Lambda with RunPod or Vast.ai, they often flatten unlike-for-unlike pricing models into a single table. That is where bad content starts.
Lambda vs Public AWS and Google Cloud References
The following September 12, 2026 references use different pricing models. AWS examples are for US East (N. Virginia), not a global rate:
| Provider | Published example | Price | What the number actually is |
|---|---|---|---|
| Lambda | 1x A100 40GB | $1.99/hr | Fixed self-serve instance price |
| Lambda | 8x A100 SXM 80GB | $2.79/GPU-hr | Per-GPU rate inside an eight-GPU instance |
| AWS | p4d.24xlarge A100 Capacity Block | $1.475/hr per accelerator | Reserved-capacity example |
| Lambda | 1x H100 PCIe 80GB | $3.29/hr | Fixed self-serve instance price |
| Lambda | 1x H100 SXM 80GB | $4.29/hr | Fixed self-serve instance price |
| AWS | p5.4xlarge H100 Capacity Block | $5.191/hr per accelerator | September 2026 reserved-capacity example |
| Google Cloud | A2/A3 accelerator-optimized VM | Check calculator or console | Whole-VM price varies by machine type, region, and provisioning model |
Google Cloud prices A2 A100 and A3 High/Mega H100 configurations as accelerator-optimized VMs. Use the applicable machine shape, region, and provisioning model in its pricing table or calculator. This review does not establish one comparable fixed Google Cloud quote. For the cross-provider comparison, use the cloud GPU price index; for host-marketplace versus tier pricing, use Vast.ai vs RunPod.
Monthly Cost Examples Based on Published Rates
For the September 12, 2026 snapshot, multiply the published rate by planned hours. A 40-hour week on a single A100 40GB is $79.60; on a single H100 PCIe it is $131.60 before tax. At 730 hours per month:
| Published example | Approx. monthly cost |
|---|---|
| Lambda A100 40GB at $1.99/hr | $1,452.70/mo |
| Lambda 8x A100 80GB at $22.32/instance-hr | $16,293.60/mo for the instance |
| Lambda H100 PCIe at $3.29/hr | $2,401.70/mo |
| Lambda H100 SXM at $4.29/hr | $3,131.70/mo |
| AWS A100 Capacity Block example at $1.475/hr | $1,076.75/mo |
| AWS H100 Capacity Block example at $5.191/hr | $3,789.43/mo |
These are arithmetic projections, not universal invoices or monthly reservations. Lambda notes that applicable tax can be added. The AWS A100 example is a per-accelerator allocation inside an eight-GPU instance; the whole p4d.24xlarge at $11.80/hour projects to $8,614 for 730 hours. AWS Capacity Blocks have their own purchasable durations and reservation terms, so a 730-hour multiplication is not an offered block quote.
Price History
| Snapshot | A100 40GB | A100 80GB | H100 PCIe 80GB | H100 SXM 80GB |
|---|---|---|---|---|
| April 2026 | $1.99/hr | Not recorded | $3.29/hr | $4.29/hr |
| July 30, 2026 | $1.99/hr | $2.79/GPU-hr in 8x instance | $3.29/hr | $4.29/hr |
| September 12, 2026 | $1.99/hr | $2.79/GPU-hr in 8x instance | $3.29/hr | $4.29/hr |
Does Lambda Publicly List RTX 4090?
The public Lambda pricing page reviewed for this update is centered on data-center GPU pricing. If your actual job is an RTX 4090-class workload, do not force a Lambda comparison where it does not belong.
Instead:
- check an RTX 4090 cloud rental guide for the 24GB class
- check the Cloud GPU Pricing Comparison 2026 for source-backed A100 and H100 references
Bottom Line
Lambda is one of the cleanest public baselines for self-serve A100 and H100 pricing because it publishes simple fixed hourly rates.
If you need:
- a clear public self-serve number, Lambda is useful
- a hyperscaler reference, use Google Cloud VM pricing or AWS Capacity Block examples and label them carefully
- a live marketplace rate, verify it in the deployment flow on RunPod or the live market on Vast.ai instead of copying an old blog quote
Related guides:
- Cloud GPU Pricing Comparison 2026
- RTX 4090 Cloud Rental: Is It Worth It?
- RTX 4090 vs A100 vs H100 Comparison
- GPU Cost Calculator
Official sources reviewed September 12, 2026:
