Rent the B200 if…
you train, or serve at high throughput where FP8 or FP4 compute is the limit, or your model needs more than 141 GB per GPU.
Deploy a B200 · $2,690/mo
NVIDIA B200180 GB HBM3e · Blackwell
vs
NVIDIA H200141 GB HBM3e · Hopper
NVIDIA B200 vs H200: Blackwell or Hopper with more memory?
Both answer the memory problem of the H100. The H200 does it with 141 GB on the same Hopper compute; the B200 adds Blackwell's throughput, 180 GB and 8 TB/s — for 1.6× the monthly price.
- B200, dedicated$2,690per month · $3.68/h effective
- H200, dedicated$1,690per month · $2.32/h effective
- Price ratio1.6×the B200 costs 1.6× the H200 per month
The short answer
Which one should you rent?
If your job is limited by memory and not by compute — large-model inference at moderate batch sizes, long contexts — the H200 gets you 141 GB at 4.8 TB/s for $1,690 a month, the lower price per GB of the two. If it is limited by compute too, or you can serve at FP4, the B200 is the faster GPU and usually the cheaper one per result: 180 GB, 8 TB/s and about 2.3× the dense tensor throughput, for 1.6× the price.
- Compute. About 2,250 TFLOPS of dense FP16/BF16 on the B200 against about 989 on the H200, plus FP4 at about 9,000 TFLOPS.
- Memory. 180 GB at 8 TB/s against 141 GB at 4.8 TB/s: 28% more capacity and 67% more bandwidth on the B200.
- Price per GB. $14.94 per GB a month on the B200 against $11.99 on the H200: the H200 is the cheaper memory, the B200 the faster.
- Software. Hopper is the most mature target in every framework; Blackwell needs CUDA 12.8 or later, and FP4 needs libraries that support it.
Rent the H200 if…
you serve 70B-class models with long contexts at moderate concurrency, want the most memory per dollar on Hopper, or need the most mature software stack.
Deploy an H200 · $1,690/moSpecifications
B200 vs H200: the numbers side by side
NVIDIA's published figures for each part, tensor throughput without sparsity. The last column is how the B200 compares, as a multiple of the H200.
| Specification | NVIDIA B200 | NVIDIA H200 | B200 vs H200 |
|---|---|---|---|
| Architecture | Blackwell · 2024 | Hopper · 2024 | — |
| GPU memory | 180 GB HBM3e | 141 GB HBM3e | 1.3× |
| Memory bandwidth | 8 TB/s | 4.8 TB/s | 1.7× |
| FP16 / BF16 tensor (dense) | ~2,250 TFLOPS | ~989 TFLOPS | 2.3× |
| FP8 tensor (dense) | ~4,500 TFLOPS | ~1,979 TFLOPS | 2.3× |
| FP4 tensor (dense) | ~9,000 TFLOPS | Not supported | Only the B200 |
| Maximum power | Up to 1,000 W | Up to 700 W | 1.4× |
| Form factor | SXM (HGX B200) | SXM5 | — |
| GPU-to-GPU link | NVLink 5 · 1.8 TB/s | NVLink 4 · 900 GB/s | — |
| Multi-Instance GPU (MIG) | Up to 7 instances | Up to 7 instances | — |
| ECC memory | Yes | Yes | — |
| Video engines | 7 NVDEC · 7 NVJPG | 7 NVDEC · 7 NVJPG | — |
| CUDA compute capability | 10.0 | 9.0 | — |
| Dedicated host here | 32 vCPU · 384 GB RAM · 4 TB NVMe | 24 vCPU · 256 GB RAM · 3 TB NVMe | — |
| Price per month here | $2,690 | $1,690 | 1.6× |
Sources: NVIDIA datasheets and product pages for each GPU. Tensor figures are dense (NVIDIA often prints them doubled “with sparsity”). Prices are this site's flat monthly rates for a dedicated server with the host listed, in USD excluding VAT.
Value
What each dollar buys
Price per month divided into the two resources most jobs run out of first — memory and memory bandwidth — and what the same GPU costs kept on for a month at the market-median hourly rate.
Per GB of GPU memoryB200 $14.94 · H200 $11.99per month, dedicated host included
Per TB/s of bandwidthB200 $336 · H200 $352per month — what decode-heavy inference pays for
Effective hourly rateB200 $3.68/h · H200 $2.32/hflat monthly price over 730 hours
Same GPU on demand, 24/7B200 $5,220 · H200 $3,270market-median hourly rate × 730 h
Model sizes
What fits on one B200 and one H200
The largest model whose weights fit on a single GPU, keeping 10% of the memory for the runtime, activations and a small KV cache. Long contexts and large batches need more — size them with the GPU memory calculator.
| Weights precision | NVIDIA B200 | NVIDIA H200 | Typical use |
|---|---|---|---|
| 16-bit (BF16 / FP16) | up to ~81B parameters | up to ~63B parameters | Full-precision inference; the base weights of a LoRA fine-tune |
| 8-bit (FP8 / INT8) | up to ~160B parameters | up to ~130B parameters | FP8 serving with vLLM or TensorRT-LLM on GPUs that support it; INT8 weights elsewhere |
| 4-bit (GPTQ, AWQ, Q4) | up to ~290B parameters | up to ~230B parameters | Quantised inference on a budget; QLoRA fine-tuning |
Parameters in billions, weights only: 2 bytes per parameter at 16-bit, 1 at FP8 or INT8, about 0.55 at 4-bit (GPTQ, AWQ, GGUF Q4). Mixture-of-experts models need memory for all their experts, not just the active ones.
Prices
B200 and H200 price per month, by term
Per dedicated GPU server, in USD excluding VAT — CPU cores, RAM, NVMe, 20 TB of transfer and 24/7 support included.
| Term | NVIDIA B200 | NVIDIA H200 | Difference per month |
|---|---|---|---|
| Monthly | $2,690 | $1,690 | $1,000 |
| 3 months · −5% | $2,556 | $1,606 | $950 |
| 6 months · −10% | $2,421 | $1,521 | $900 |
| 12 months · −15% | $2,287 | $1,437 | $850 |
| Dedicated host | 32 vCPU · 384 GB RAM · 4 TB NVMe · 25 Gbps | 24 vCPU · 256 GB RAM · 3 TB NVMe · 25 Gbps | included |
B200 vs H200: questions answered
Is the B200 better than the H200 for inference?
For high-throughput serving, usually: more bandwidth (8 against 4.8 TB/s), more compute and FP4. For a single 70B-class model at modest concurrency, the H200's 141 GB at $1,690 a month is often enough, and cheaper.
Which has more memory, the B200 or the H200?
The B200: 180 GB of HBM3e against 141 GB. The B300 goes further, with 288 GB per GPU.
Can I rent eight B200s or eight H200s in one node?
Yes. 8-GPU NVLink nodes of both are on the clusters page, priced per node per month: $18,900 for eight B200s and $10,990 for eight H200s.
Are the B200 and the H200 in stock?
8 of 12 B200 units were free at the last stock update; 9 of 32 H200 units were free at the last stock update (18 September 2026). Both are provisioned automatically, in under 10 minutes after payment.
Can I rent the B200 or the H200 without KYC?
Yes. Both are rented with an email address and paid from a prepaid balance in Bitcoin, Ether or USDT — $2,690 and $1,690 a month, dedicated, with no identity check and no card.
Other GPU comparisons
Both are in stock, by the month
Dedicated, single-tenant and live in under 10 minutes: the B200 at $2,690 or the H200 at $1,690 a month, paid in BTC, ETH or USDT with no KYC.