KoyebEU-WestOn-demand
$3.00/ GPU / hr
Open provider
Cloud GPU · verified 2026-09-05
NVIDIA H200
Hopper141 GB
Hopper with a bigger tank. Same SM family as H100, 141 GB of HBM3e at 4.8 TB/s. You rent it when memory is the bill, not because the badge says 200.
VRAM141 GB
From /GPU/hr$3.00
ArchitectureHopper
VendorNVIDIA
Sponsored
Featured H200 offers
The four cheapest in-stock on-demand H200 hours in this sample. Each card is a sponsored placement.
Packet.aiUS-WestOn-demand
$3.29/ GPU / hr
Open provider
RunPodUS-EastOn-demand
$3.39/ GPU / hr
Open provider
LambdaUS-WestOn-demand
$3.49/ GPU / hr
Open providerSponsored does not change ranking. Commissions never move a row, a GPU, or a displayed price.
In brief
What to know before you rent
- Compute is H100-class. The product is 141 GB HBM3e and 4.8 TB/s.
- Worth it when the KV cache or optimizer states do not fit 80 GB.
- Still 700 W SXM and NVLink 900 GB/s, not a Blackwell node, not a 4090.
Catalog listings
H200 Pricing and Availability
Published on-demand rates first. Spot and waitlists are labeled. Confirm the live rate before you provision.
ProvidersRegionBilling typeInterconnect/GPU/hr
Our reading
Who this GPU is for
I rent H200 when the extra 61 GB pays for itself in fewer GPUs or fewer spills. I do not rent it as a default “better H100.” If the job is compute-bound inside 80 GB, the H200 hour is a vanity tax.
Rent it if 70B+ serving with real context, training states that overflow H100, single-GPU work that would otherwise become a messy 2×H100 split.
Skip it if Batches that already fit H100, budgets that only work at Vast.ai floor H100, or anyone who needed NVLink 5 and FP4.
How we compare hoursDatasheet
Technical specifications
NVIDIA H200 · Per GPU · SXM
Compute · dense
- FP8
- 1,979 TFLOPS3,958 with sparsity, Hopper, same class as H100 SXM
- FP16 / BF16
- 989.5 TFLOPS1,979 with sparsity
- TF32
- 494.5 TFLOPS989 with sparsity
- INT8
- 1,979 TOPS3,958 with sparsity
- FP32
- 67 TFLOPS
- FP64
- 34 TFLOPS67 TFLOPS Tensor Core
- Precision support
- FP8, FP16, BF16, TF32, FP32, FP64, INT8
Memory
- Capacity
- 141 GB HBM3e
- Bandwidth
- 4,800 GB/s
- ECC
- Yes
Silicon
- Architecture
- Hopper
- Process
- TSMC 4N
- Form factor
- SXM (typical cloud SKU)
Fabric and host
- GPU interconnect
- NVLink 900 GB/s
- Host interface
- PCIe 5.0 x16
Power
- Board power
- 700 WSXM class
Platform
- Partitioning
- MIG, up to 7 instances
H200 is not a new architecture. It is H100’s compute with more, faster memory. If a listing shows 80 GB and “H200,” it is mislabeled. Walk away.
Trade-offs
Pros and cons of H200
What this GPU does well, and when the hour is the wrong buy.
Strengths
- 141 GB is the difference between one GPU and a sharded headache for many 70B stacks
- 4.8 TB/s HBM3e moves KV cache without pretending bandwidth is free
- Same Hopper software path as H100, fewer “new architecture” surprises
- MIG still exists if the operator exposes it
Limits
- You still have Hopper Tensor cores. FP4 lives on Blackwell
- The hour is often ~2× H100. Memory has to earn that
- Stock is thinner than H100. Waitlists happen on the good SKUs
- SXM vs PCIe still matters. The name still lies
When 141 GB is the product
H200 exists because 80 GB started losing to context length. KV cache is a memory problem. Optimizer states are a memory problem. Hopper FLOPs were already there.
I treat the H200 hour as a memory purchase. If I cannot point at the bytes that did not fit on H100, I do not upgrade. The blog post that says “H200 is better” without a VRAM sentence is an ad.
Versus H100, not versus B200
The honest comparison is H100. Same family, same NVLink generation, same MIG story. Different tank. B200 is a different conversation: FP4, 1.8 TB/s NVLink, 1 kW, and a catalog that often says waitlist.
If you need Blackwell, rent Blackwell. Do not launder that decision through an H200 page because the number in the name is bigger.
How I actually rent it
- 01
Measure the spill
What does not fit on 80 GB? If you cannot name it, stay on H100.
- 02
Match the node
SXM, GPU count, NVLink. H200 in a lonely PCIe slot is a weird way to buy bandwidth.
- 03
Price the alternative
Two H100s vs one H200. Sometimes the split is cheaper. Sometimes it is a software tax.
FAQ
H200 questions
Who has the cheapest H200?+
As of September 5, 2026, the cheapest verified in-stock on-demand H200 tracked by GPUBeacon is $3.00 per GPU per hour from Koyeb.
How much does the H200 cost per month?+
As of September 5, 2026, at 720 hours per month, one H200 costs an estimated $2,621 at the median on-demand price in this sample. Cheapest verified in stock: $2,160 per month on-demand.
Where can I rent an H200?+
GPUBeacon currently tracks H200 configs from 10 providers. The cheapest verified in-stock on-demand configs come from Koyeb, Packet.ai, RunPod and Lambda. See the full price comparison above for every provider and config.
Are H200 prices going up or down?+
GPUBeacon publishes a catalog sample, not a 90-day price index. As of September 5, 2026, the median on-demand H200 in this sample is $3.64 per GPU per hour. The cheapest verified in-stock hour is $3.00.
Why choose the H200?+
70B+ serving with real context, training states that overflow H100, single-GPU work that would otherwise become a messy 2×H100 split.
When is the H200 not a good fit?+
Batches that already fit H100, budgets that only work at Vast.ai floor H100, or anyone who needed NVLink 5 and FP4.
Is H200 just an H100 with more VRAM?+
Mostly yes: Hopper compute, 141 GB HBM3e, higher bandwidth. That is the point. Do not expect Blackwell TFLOPS.
Will H200 always beat H100 on a 70B?+
No. If you were already compute-bound inside 80 GB, you bought headroom. If you were swapping or shrinking context to survive, H200 can drop GPU count and win the bill.
Keep comparing
Alternatives to NVIDIA H200
Same job, different envelope. Compare VRAM, bandwidth and the hour before you lock a badge.
NVIDIA H100
80 GB HBM3 · 30% less bandwidth than the H200
From$1.73/ GPU / hr
Read the verdictNVIDIA B200
180 GB HBM3e · 67% more bandwidth than the H200
From$3.75/ GPU / hr
CompareNVIDIA A100
80 GB HBM2e · 58% less bandwidth than the H200
From$1.19/ GPU / hr
Compare
AMD MI300X
192 GB HBM3 · 10% more bandwidth than the H200
From$1.99/ GPU / hr
CompareSources
Where these numbers come from
TFLOPS and TDP are from the vendor datasheet. Hourly prices are from the GPUBeacon catalog sample. Open the originals, then confirm the live SKU with the provider before you rent.