Rubin vs H200 SXM
Rubin has 288 GB of HBM4 at 22 TB/s per GPU, against 141 GB of HBM3e at 4.8 TB/s on H200 SXM. H200 SXM rents from $2.60/GPU-hour on demand (4 October 2026); no cloud we track publishes an hourly Rubin price yet.
Specifications, side by side
| Specification | Rubin | H200 SXM |
|---|---|---|
| Exact GPU | Rubin | H200 SXM |
| Memory per GPU | 288 GB | 141 GB |
| Memory generation | HBM4 | HBM3e |
| Memory bandwidth per GPU | 22 TB/s | 4.8 TB/s |
| NVLink bandwidth per GPU | 3.6 TB/s | 900 GB/s |
| FP16/BF16 compute per GPU, dense | 4 PFLOPS | Not published by NVIDIA as a dense figure (its H200 page lists 1,979 TFLOPS with sparsity) |
| Availability | Announced; NVIDIA says partners offer it from the second half of 2026 · No public hourly price (3 October 2026) | Rentable now: 33 offers from 24 providers (4 October 2026) |
Rubin has 2.04 times the published memory capacity. Calculation: 288 ÷ 141. Specifications updated 2026-10-05.
These are NVIDIA's published peak figures, not measured speed on any workload.
Recorded on-demand rental rates
| Rate | Rubin | H200 SXM |
|---|---|---|
| Lowest recorded on-demand | No recorded on-demand rate | $2.60/GPU-hour (recorded 4 October 2026) |
Every rate comes from our daily price collection and shows the date it was recorded.
Check Rubin cloud announcements and availability.
Current H200 SXM rental offers
Provider, memory & availability
| GPU / provider | Billing | $/GPU-hour | GPUs / total per hour | Availability / details |
|---|---|---|---|---|
| H200 (SXM)GMI Cloud141 GB VRAM | on-demand | $2.60per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Daytona141 GB VRAMH200; GPU rate only, CPU/memory/disk billed separately | spot / interruptible | $2.61per GPU-hour30-day comparison unavailable | Node minimum and total not reported | Not reportedView provider2026-10-03Details
|
| H200 (SXM)Hyperstack141 GB VRAM | reserved / committedTerm and minimum volume not published in this table; confirm reservation quoteStarting rate; confirm terms | $2.79per GPU-hour30-day comparison unavailable | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Together AI141 GB VRAM | spot / interruptible | $2.99per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Koyeb141 GB VRAM | serverless | $3.00per GPU-hour$0.000833/sec+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Impossible Cloud141 GB VRAM287-H200; 16 GPUs quoted; 8 GPUs/node; availability by request | reserved / committed12 monthsStarting rate; confirm terms | $3.11per GPU-hour30-day comparison unavailable | Node minimum and total not reported | Waitlist2026-10-03Details
|
| H200 (SXM)Spheron141 GB VRAM | spot / interruptible | $3.12per GPU-hour+21.4% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Prime Intellect141 GB VRAM | on-demand | $3.14per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)RunPod141 GB VRAM | community cloud | $3.59per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 NVLMassed Compute141 GB VRAM | on-demand | $3.62per GPU-hour+0.0% over 30 days | 1× $3.62per node-hour | Not reportedView provider2026-10-04Details
|
| H200 NVLRunPod143 GB VRAM | secure cloud | $3.79per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Hyperstack141 GB VRAM | on-demand | $3.99per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Cerebrium141 GB VRAM | serverless | $4.20per GPU-hour$0.001166/sec+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Crusoe141 GB VRAM | on-demand | $4.29per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)DigitalOcean141 GB VRAMgpu-h200x1-141gb | on-demand | $4.47per GPU-hour30-day comparison unavailable | 1× $4.47per node-hour | Not reportedView provider2026-10-04Details
|
| H200 (SXM)DigitalOcean141 GB VRAMgpu-h200x8-1920gb | on-demand | $4.47per GPU-hour30-day comparison unavailable | 8× $35.76per node-hour | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Modal141 GB VRAM | serverless | $4.54per GPU-hour$0.001261/sec+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Daytona141 GB VRAMH200; GPU rate only, CPU/memory/disk billed separately | on-demand | $4.54per GPU-hour30-day comparison unavailable | Node minimum and total not reported | Not reportedView provider2026-10-03Details
|
| H200 (SXM)RunPod141 GB VRAM | secure cloud | $4.59per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Jarvislabs141 GB VRAM | on-demand | $4.59per GPU-hour+15.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Spheron141 GB VRAM | on-demand | $4.80per GPU-hour-3.2% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)DataCrunch141 GB VRAM | on-demand | $4.88per GPU-hour+21.9% over 30 days | 1× $4.88per node-hour | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Nebius141 GB VRAM | on-demand | $5.40per GPU-hour+20.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Replicate141 GB VRAM | serverless | $5.49per GPU-hour$0.001525/sec+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Google Cloud141 GB VRAM | spot / interruptible | $5.59per GPU-hour+3.9% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)RunPod141 GB VRAMFlex worker | serverless | $5.93per GPU-hour$0.001647/sec30-day comparison unavailable | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Together AI141 GB VRAM | on-demand | $5.99per GPU-hour30-day comparison unavailable | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)fal141 GB VRAM | serverless | $6.00per GPU-hour$0.001667/sec+33.3% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)OVHcloud141 GB VRAM | on-demand | $6.20per GPU-hour+0.0% over 30 days | 8× $49.56per node-hour | Not reportedView provider2026-10-04Details
|
| H200 (SXM)CoreWeave141 GB VRAM | on-demand | $6.30per GPU-hour+0.0% over 30 days | 8× $50.44per node-hour | Not reportedView provider2026-10-04Details
|
| H200 (SXM)AWS141 GB VRAMp5en.48xlarge | on-demand | $7.91per GPU-hour30-day comparison unavailable | 8× $63.30per node-hour | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Fireworks AI141 GB VRAM | serverless | $8.00per GPU-hour$0.002222/sec+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
| H200 (SXM)Google Cloud141 GB VRAM | on-demand | $9.31per GPU-hour+0.0% over 30 days | Node minimum and total not reported | Not reportedView provider2026-10-04Details
|
No offers match these filters.
Try another provider or reduce the minimum memory.
Tables rank by price, lowest first. Commercial relationships never affect ordering. How we collect and rank prices.
Monthly view assumes 730 hours of continuous use. Daily view assumes 24 hours.
Specifications and sources
Rubin values come from NVIDIA's individual GPU specification table. H200 SXM capacity comes from its NVIDIA specification. Updated 2026-10-05. Quoted evidence is listed under Sources in the footer.
H200 SXM memory type · NVIDIA DGX B200 · NVIDIA HGX B200/B300 · Blackwell Ultra specifications · NVIDIA H200
NVIDIA publishes two sets of per-GPU Rubin figures. Its HGX table lists 22 TB/s of memory bandwidth and 3.6 TB/s of NVLink bandwidth; its Vera Rubin NVL72 table lists 19.2 TB/s and 3 TB/s, for an at-scale AI factory. Ask the provider which system it rents before comparing performance.
H200 SXM rental prices and history · Rubin specifications and cloud availability · All GPU comparisons