AI flagship · Blackwell Ultra
NVIDIA B300 servers New
NVIDIA’s Blackwell Ultra flagship, with 288 GB of HBM3e on every GPU.
- 288 GBHBM3e per GPU
- 8 TB/smemory bandwidth
- NVLink 5GPU interconnect
- 1–8 GPUsper server
Price per GPU · monthly
$4,019/mo
−30%Market median $5,745≈ $5.51 per GPU-hour
Below every on-demand list price we found
Server size
Configure a B300 serverUSD per server per month · prepaid · pay in BTC, ETH, USDT, USDC, LTC
Why the B300.
The B300 is the Blackwell Ultra generation of NVIDIA’s data-center GPU. It keeps the Blackwell design and fifth-generation NVLink of the B200 and raises memory to 288 GB of HBM3e per GPU: 2.3 TB in an 8-GPU server. Pick it when the model, the context or the batch no longer fits comfortably on B200 or H200.
- 288 GB HBM3e per GPU
The most memory per GPU in our range, tied with the AMD MI355X.
- NVLink 5 · 1.8 TB/s
All eight GPUs of a server talk to each other at full bandwidth through NVSwitch.
- FP4 and FP8 tensor cores
Blackwell’s second-generation Transformer Engine runs NVFP4 and FP8 natively.
From 1 to 8 B300 GPUs.
Same price per GPU at every size. vCPU, DDR5 RAM and NVMe scale with the number of GPUs.
| Server | GPU memory | vCPU | RAM | NVMe | Market median | CryptGPU | Action |
|---|---|---|---|---|---|---|---|
| 1× B300 | 288 GB | 32 | 384 GB | 3.84 TB | $4,019/mo | Configure 1× B300 | |
| 2× B300NVLink 5 | 576 GB | 64 | 768 GB | 7.68 TB | $8,038/mo | Configure 2× B300 | |
| 4× B300NVLink 5 | 1,152 GB | 128 | 1,536 GB | 15.36 TB | $16,076/mo | Configure 4× B300 | |
| 8× B300NVLink 5 | 2,304 GB | 256 | 3,072 GB | 30.72 TB | $32,152/mo | Configure 8× B300 |
NVIDIA B300 specifications.
Figures from NVIDIA’s published specifications. Tensor figures are peak dense throughput unless marked otherwise.
GPU and memory
- Architecture
- Blackwell Ultra, TSMC 4NP
- Transistors
- 208 billion (two dies)
- Form factor
- SXM module (HGX B300)
- GPU memory
- 288 GB HBM3E
- Memory bandwidth
- Up to 8 TB/s
- GPU-to-GPU link
- NVLink 5, 1.8 TB/s per GPU
- Host link
- PCIe Gen6 x16
- Multi-Instance GPU
- Up to 7 instances
Tensor throughput (dense)
- FP4
- 13.5 PFLOPS
- FP8 / FP6
- 4.5 PFLOPS
- FP16 / BF16
- 2.25 PFLOPS
- TF32
- 1.1 PFLOPS
- FP32
- 75 TFLOPS
- FP64
- 1.25 TFLOPS
Cores and media
- Streaming multiprocessors
- 160
- Tensor Cores
- 640, 5th generation
- Video decoders
- 7 NVDEC
Server, per GPU
- vCPU
- 32
- RAM
- 384 GB DDR5
- NVMe
- 3.84 TB
- GPUs per server
- 1, 2, 4, 8
Per-GPU figures are NVIDIA’s HGX B300 eight-GPU totals divided by eight. NVIDIA lists 288 GB of HBM3E per GPU in its Blackwell Ultra technical blog and DGX B300 user guide, while its HGX and DGX B300 product pages give 2.1 TB for eight GPUs. NVIDIA publishes no per-GPU power figure for HGX B300.
Which models fit on B300 servers.
Smallest server that holds each model, by precision. Rule of thumb with headroom for the KV cache; long contexts and large batches need more.
| Model (total parameters) | FP16 | FP8 | 4-bit |
|---|---|---|---|
| Qwen3.5-9B | 1× | 1× | 1× |
| Gemma 4 31B | 1× | 1× | 1× |
| Llama 3.3 70B | 1× | 1× | 1× |
| gpt-oss-120b | 2× | 1× | 1× |
| DeepSeek-V4-Flash | 4× | 2× | 1× |
| Qwen3.5-397B-A17B | 4× | 2× | 1× |
| DeepSeek-R1 (671B) | 8× | 4× | 2× |
| Kimi K2.6 (1T) | — | 8× | 4× |
Estimate: FP16 ≈ 2.4 bytes, FP8 ≈ 1.2 bytes and 4-bit ≈ 0.65 bytes per parameter, with 92% of GPU memory usable. — = does not fit in the largest B300 server. Several recent models ship natively in FP8 or 4-bit formats; the columns show what each precision needs. Try the sizing helper for other models and for fine-tuning.
B300 prices across the market.
Monthly price per GPU. Public on-demand list prices collected on 23 Sep 2026; hyperscalers shown for reference only.
Per GPU, 730 hours a month. Hyperscalers (AWS, Google Cloud, Azure, Oracle) are excluded from the median. How we compare
B300 questions.
More in the full FAQ.
How is the B300 different from the B200?
Same Blackwell architecture and NVLink 5, but 288 GB of HBM3e per GPU instead of 180 GB, and more low-precision throughput. If your model fits in 180 GB per GPU, the B200 costs less.
Can I rent fewer than eight B300 GPUs?
Yes. B300 servers come with 1, 2, 4 or 8 GPUs, at the same price per GPU for every size.
Is the B300 a good choice for FP64 simulation?
No. Blackwell Ultra gives up most FP64 throughput for AI formats. For double precision, the H200 or the AMD MI355X are better choices.
Build your B300 server.
Pick 1 to 8 GPUs and see the exact monthly price. $4,019 per GPU, the same at every size.