MI355X
AMD · CDNA 4
288 GBHBM3E
8 TB/sBandwidth
5 PFFP8 dense
$2,449 /GPU/mo$3,285 · −25%
NVIDIA · Blackwell Ultra · SXM · 8-GPU HGX
Rent the NVIDIA HGX B300 by the month.
The NVIDIA B300 is Blackwell Ultra: 288 GB of HBM3e per GPU and 13.5 PF of dense FP4, built for reasoning-model inference and very large fine-tunes. An 8-GPU HGX B300 server gives you 2.3 TB of GPU memory on a single 1.8 TB/s NVLink 5 domain.
288 GBHBM3e per GPU
8 TB/sMemory bandwidth
4.5 PFFP8 dense
13.5 PFFP4 dense
Why the B300
What the B300 is best at
- 288 GB HBM3e. Fit a 400B-parameter model at FP4 or long-context KV caches on one node.
- 13.5 PF FP4 dense. 1.5× the FP4 throughput of B200 for low-precision inference.
- NVLink 5, 1.8 TB/s. All eight GPUs share one NVLink domain: no PCIe hop for tensor parallelism.
- Reasoning and agent inference at FP4/FP8
- Fine-tuning 70B–400B models
- Long-context serving with large KV caches
Blackwell Ultra trades FP64 throughput for AI throughput: for double-precision science, B200 or MI355X are better fits.
Pricing
B300 monthly price by server size
One flat price per GPU, the same from 1 to 8 GPUs. The market median is the public on-demand list price of specialist GPU clouds × 730 hours.
| Server | CoreGPU / month | Market median / month | You save / month | You save / year | ≈ per GPU-hour |
|---|---|---|---|---|---|
| 1× B300 | $4,899 | $1,653 | $19,836 | $6.71 | |
| 2× B300 | $9,798 | $3,306 | $39,672 | $6.71 | |
| 4× B300 | $19,596 | $6,612 | $79,344 | $6.71 | |
| 8× B300 | $39,192 | $13,224 | $158,688 | $6.71 |
Prices in USD, excluding taxes. Market median checked Nov 4, 2026 across specialist GPU clouds; hyperscalers excluded. All GPU prices
Compare
B300 vs B200
Pick B300 for FP4 inference and memory (288 vs 180 GB). Pick B200 when you need strong FP64 or the lower price matters more than memory.
| B300 | B200 | |
|---|---|---|
| Architecture | Blackwell Ultra | Blackwell |
| GPU memory | 288 GB HBM3e | 180 GB HBM3e |
| Memory bandwidth | 8 TB/s | 8 TB/s |
| FP8 dense | 4.5 PF | 4.5 PF |
| FP4 dense | 13.5 PF | 9 PF |
| Interconnect | NVLink 5 · 1.8 TB/s | NVLink 5 · 1.8 TB/s |
| Power (TDP) | 1,100 W | 1,000 W |
| Price per GPU / month | $4,899 | $4,449 |
| Price per GB of memory | $17.01 | $24.72 |
| Price per FP8 PFLOPS | $1,089 | $989 |
| Availability | Only 9 left | Back Nov 20 |
Specifications
NVIDIA HGX B300 technical specs
Per-GPU figures from the vendor datasheet. Tensor throughput is dense (without sparsity).
| Model | NVIDIA HGX B300 |
|---|---|
| Architecture | Blackwell Ultra |
| Form factor | SXM · 8-GPU HGX |
| GPU memory | 288 GB HBM3e |
| Memory bandwidth | 8 TB/s |
| FP8 Tensor (dense) | 4.5 PF |
| FP4 Tensor (dense) | 13.5 PF |
| Interconnect | NVLink 5 · 1.8 TB/s |
| Power (TDP) | 1,100 W |
| GPUs per server | 1 to 8 |
| Availability | Only 9 left |
FAQ
B300 rental: common questions
Another question? Our engineers answer within one business day.
Contact usHow much does it cost to rent an NVIDIA HGX B300 per month?
$4,899 per GPU per month at CoreGPU, about $6.71 per GPU-hour, versus a market median of $6,552 ($8.98/hour × 730 h). A full 8-GPU server is $39,192 per month, taxes excluded.
Is the B300 available right now?
Yes: only 9 left in the current allocation of 64 GPUs. In-stock servers are delivered with root SSH access in under 60 seconds after payment.
How many B300 GPUs can I rent?
From 1 to 8 GPUs per server, at the same price per GPU. Servers with 8 GPUs come with the full NVLink 5 mesh. Need more? Rent several servers or ask for a private allocation.
B300 or B200?
Pick B300 for FP4 inference and memory (288 vs 180 GB). Pick B200 when you need strong FP64 or the lower price matters more than memory. See the full B300 vs B200 comparison.
What do I get with a B300 server?
Dedicated GPUs (never shared or sliced), root SSH access, your choice of a clean OS or a ready image with CUDA, PyTorch and vLLM, local NVMe scratch, and month-to-month billing with no setup fee.
Alternatives
Other GPUs in stock now
Deploy your B300 server in under 60 seconds
Pick 1 to 8 GPUs in the configurator above. Month to month, cancel anytime.