NVIDIA B300 Cloud GPU Specs
Deploy the NVIDIA B300 on HexGrid Cloud, billed per minute. The B300 carries 288GB of HBM3e memory on the Blackwell Ultra architecture.
HexGrid Cloud price
Contact
Billed per minute
VRAM
288GB
HBM3e
FP32
—
Compute performance
Market average
—
Other clouds, list price
Running the B300 on HexGrid Cloud
What you get when you deploy with us, beyond the hourly rate.
Live in under a minute
Pick a region, hit deploy, and SSH in. No quota requests, no sales call.
Per-minute billing
You pay for the minutes you use. Stop the instance and billing stops with it.
Ready for training day one
CUDA, PyTorch and the usual drivers are preinstalled, or bring your own image.
No egress charges
Move checkpoints and datasets out without a surprise bandwidth bill.
About the NVIDIA B300
What the B300 is built for, and where it falls short.
The NVIDIA B300 is a Blackwell Ultra accelerator designed for AI reasoning and next-generation inference workloads. It increases memory capacity to 288GB of HBM3e per GPU while maintaining up to 8 TB/s of memory bandwidth and increasing low-precision AI compute over B200.
Best suited for
The B300 Blackwell Ultra substantially increases GPU memory capacity to 288GB while increasing AI compute for reasoning and large-context inference workloads.
- AI reasoning
- Large LLM inference
- LLM training
- Long-context models
- Mixture-of-Experts models
- AI factories
Strengths
- 288GB HBM3e memory
- Up to 8 TB/s memory bandwidth
- Improved AI reasoning performance
- FP4 and FP8 acceleration
- Fifth-generation NVLink
Limitations
- Extremely high infrastructure requirements
- High power consumption
- Premium cloud pricing and limited availability
NVIDIA B300 specifications
Full technical specifications for the NVIDIA B300.
Memory
- Memory
- 288 GB
- Type
- HBM3e
- Bandwidth
- 8,000 GB/s
- ECC
- Yes
Compute
- Tensor generation
- Gen 5
Power & physical
- TDP
- 1100 W
- Form factor
- SXM
- Cooling
- Server-cooled
Interconnect
- Interface
- SXM
- NVLink
- Yes
- NVLink bandwidth
- 1,800 GB/s
- Infinity Fabric
- No
Virtualization
- MIG
- Yes
- vGPU
- Yes
- SR-IOV
- Yes
Software
- CUDA
- Yes
- ROCm
- No
Precision support
- FP64
- Yes
- FP32
- Yes
- TF32
- Yes
- BF16
- Yes
- FP16
- Yes
- FP8
- Yes
- FP6
- Yes
- FP4
- Yes
- INT8
- Yes
B300 frequently asked questions
Common questions about NVIDIA B300 pricing, specifications and workloads on HexGrid Cloud.
How much memory does NVIDIA B300 have?
The NVIDIA B300 provides 288GB of HBM3e memory per GPU.
What is B300 memory bandwidth?
NVIDIA documents up to 8 TB/s of HBM3e memory bandwidth per B300 GPU.
What is the difference between B200 and B300?
B300 Blackwell Ultra increases memory capacity from 180GB to 288GB and adds additional AI compute aimed especially at reasoning and large-context inference.
Does B300 support FP4?
Yes. Blackwell Ultra supports low-precision AI formats including FP4.
How much does a B300 cost per hour?
B300 cloud pricing depends on provider, server configuration, GPU count, and availability. HexGrid compares available B300 offers.
Other GPUs on HexGrid Cloud
If the B300 isn't the right fit, these are also available to deploy.
NVIDIA B200
Available on HexGrid Cloud
Compare with B300NVIDIA H200
Available on HexGrid Cloud
Compare with B300NVIDIA H100 SXM
Available on HexGrid Cloud
Compare with B300Start training on the B300 today
Billed per minute, running in about a minute.