Catalog
Datacenter NVIDIA: A100, H100, H200, Blackwell as listed. We show on-demand first. Reserved is labeled, not mixed in as a fake hour.
Provider review / verified 2026-09-05
LambdaGPU review
Lambda sells on-demand GPU cloud, reserved clusters, and the boxes under desks. The cloud is for people who want NVIDIA machines that boot like workstations, not a YAML career. On-demand exists. Spot mostly does not. That is a feature.
Models listed6
From /GPU/hr$0.99
Regions in sample1
Business modelSpecialist GPU cloud
Our reading
Who this provider is for
I send people to Lambda when they are done playing marketplace and not ready for CoreWeave procurement. The hour is rarely the floor. The hour is honest: on-demand, ML-shaped images, 1-Click clusters if you are actually training. If you need the cheapest H100 on earth, leave. If you need an H100 that behaves like a Lambda workstation you already trust, stay.
Rent it if Labs and startups training for real, days, not demos, who want on-demand or reserved NVIDIA without AWS archaeology.
Skip it if Bargain hunters, spot-native batch, and anyone who needs a hyperscaler’s IAM, PrivateLink, and a three-year EDP.
How we compare hoursStrengths
- On-demand GPU cloud with a workstation-grade software stack (Lambda stack, not mystery CUDA)
- Reserved clusters when the job leaves the ‘pod’ mental model
- No spot cosplay: you are not comparing an interruptible bid to a published hour
- A100 has historically been a reason to open Lambda even when H100 is not the cheapest
Limits
- You will pay more than Vast.ai and often more than RunPod Community for the same GPU name
- Not a serverless inference toy. Idle is your problem unless you shut it off
- Capacity can waitlist on the GPUs everyone wants. Brochure ≠ stock
- Hyperscaler-adjacent enterprise features are not the point of the product
Same standard, every cloud
What we measured on Lambda
Four checks we apply to every provider. This is not a score, and it is not a paid ranking.
Price signal
Lambda on-demand sits in managed-neocloud territory. It wins when you count failed marketplace jobs, not when you screenshot a 4090.
Product shape
Instances and clusters. Workstation DNA. SSH. The stack is the pitch. Spot is largely absent, compare like for like.
Still verify
Confirm the SKU (PCIe vs SXM), region, and whether you are looking at on-demand or a reserved quote.
- Model
- Specialist GPU cloud
- Headquarters
- San Jose, US
- Setup
- Cloud instances, clusters, workstations
- Billing grain
- Hourly on-demand; reserved contracts
- Spot
- Not the product
- Cluster
- Yes, 1-Click / reserved
- Founded
- 2012
- Best known for
- ML workstations + GPU cloud
Catalog listings
Lambda catalog sample
Published on-demand hours first. Spot and waitlists are labeled. Confirm the live rate before you provision.
GPUsRegionBilling typeInterconnect/GPU/hr
Lambda is what you graduate to
There is a path I see on Discord every week: Vast.ai, then RunPod, then someone pages you because the run is now the company. Lambda is that next door. You lose the thrill of a $0.29 4090. You gain a machine that looks like the workstations Lambda has shipped since before this boom had a name.
I do not pretend it is cheap. I pretend less about uptime.
On-demand without the spot fog
A lot of competitors look cheaper because the number on the blog post was interruptible. Lambda’s on-demand is on-demand. When we put them on this site, that is the number that ranks. If a reserved cluster is cheaper, it will say reserved.
If you need spot, go somewhere that sells spot and say so. Do not launder it through a Lambda comparison.
Clusters vs pods
A pod is a weekend. A cluster is a training run with a gossip schedule. Lambda’s cluster story is why I still open them when the interconnect matters. RunPod can give you an H100. Lambda is trying to give you several that can speak.
If you are still on one GPU, you may be overpaying for a brand. Rent the GPU page, not the mythology.
How I actually rent it
- 01
Match the SKU
PCIe vs SXM, GPU count, region. The name H100 is not a SKU.
- 02
On-demand or reserved, never both in one cell
If you need weeks, ask about reserved. Do not steal an on-demand hour and hope.
- 03
Shut it off
Lambda will happily bill an idle instance. So will everyone. Set a calendar reminder like an adult.
FAQ
Lambda questions
Is Lambda cheaper than RunPod?+
Usually not on Community. Sometimes competitive on Secure/on-demand datacenter cards. Lambda is buying fewer surprises, not the floor.
Does Lambda have spot GPUs?+
Spot is not how Lambda sells the cloud. Compare their on-demand to other people’s on-demand.
Lambda vs CoreWeave?+
Lambda if you want self-serve ML cloud with workstation culture. CoreWeave if you are buying serious reserved AI capacity and can stand a heavier vendor.
Who is Lambda for?+
Teams training models who are tired of marketplace entropy and not yet a hyperscaler account.
Does Lambda sell H100 and H200?+
Yes, as cloud SKUs when in stock. Confirm SXM vs PCIe and the live hour on the model page.
Sources
Where these numbers come from
Prices and SKUs are from the GPUBeacon catalog sample. Ranking rules are on the methodology page. Outbound links may be affiliate. They never change the hour we display.
GPUs
Keep comparing
Other providers to compare
Same fields, different product. Compare the GPU hour before you lock a brand.
