NVIDIA HGX B300 on Nebius AI Cloud
Train and serve your most demanding AI models on NVIDIA’s latest Blackwell Ultra platform — deployed in production on Nebius AI Cloud.
What makes HGX B300 on Nebius different
GPU performance, optimized in-house
Our GPU clusters are optimized across every layer of the stack, from in-house designed servers to a software layer validated against NVIDIA benchmarks, so your AI workloads run at the highest achievable performance.
AI without operational overhead
We take care of the infrastructure, at any scale. Run containerized experiments with Serverless Jobs or deploy large-scale training clusters on Slurm, without touching drivers, networking, or cluster configuration.
Security and reliability by design
Every Nebius cluster comes with auto-healing that detects and recovers from hardware failures with minimum possible interruption. Our platform is also built on industry security and compliance standards, so your data and workloads always stay secure.
What teams run on NVIDIA HGX B300
Large-scale LLM workloads
Train, fine-tune, and serve frontier language models on a platform built for memory-intensive workloads. HGX B300 carries significantly more memory per GPU than its predecessor, enabling larger models, longer context windows, and higher concurrency at production scale.
Multimodal AI
Build and deploy AI systems that combine text, images, audio, or video. HGX B300 architecture efficiently balances compute and memory bandwidth, making it a practical platform for developing and running next-generation multimodal models.
AI reasoning and agentic AI
Develop and scale reasoning models and agentic systems that generate significantly more tokens per request. HGX B300 memory capacity and compute profile sustain the throughput these workloads require, without the constraints that limit smaller-memory platforms.
NVIDIA HGX B300 specifications
Specification
HGX B300
Form factor
8× NVIDIA Blackwell Ultra GPUs
FP4 Tensor Core (sparse | dense)
144 PFLOPS | 108 PFLOPS
FP8/FP6 Tensor Core (sparse)
72 PFLOPS
INT8 Tensor Core (sparse)
307 TOPS *
FP16/BF16 Tensor Core (sparse)
36 PFLOPS
TF32 Tensor Core (sparse)
18 PFLOPS
FP32
600 TFLOPS
FP64/FP64 Tensor Core
10 TFLOPS
Total memory
2.1 TB HBM3e
Total memory bandwidth
62 TB/s
NVIDIA NVLink
Fifth generation
NVIDIA NVLink Switch
NVIDIA NVLink™ 5 Switch
NVLink GPU-to-GPU bandwidth
1.8 TB/s
Total NVLink bandwidth
14.4 TB/s
Networking bandwidth
1.6 Tb/s
Source: nvidia.com/en-us/data-center/hgx and NVIDIA Blackwell Ultra datasheet. Projected performance subject to change.
Supercomputer-class performance for your workloads
NVIDIA HGX B300 systems on Nebius are connected via NVIDIA Quantum-X800 InfiniBand at 800 Gb/s per GPU. Having this high-speed interconnect enables massive parallel computation across thousands of GPUs, utilizing the full potential of NVIDIA accelerated compute, so your distributed workloads scale without becoming bottlenecked by the network fabric.
NVIDIA Exemplar Cloud Validation
NVIDIA Exemplar Cloud Validation
Nebius is a Reference Platform NVIDIA Cloud Partner and holds NVIDIA Exemplar Cloud validation across multiple GPU generations — from NVIDIA H200 to NVIDIA GB300 NVL72. Exemplar Cloud is awarded to providers that demonstrate real-world training performance against NVIDIA’s benchmarking standards, not just peak specifications.
Frequently Asked Questions
The NVIDIA HGX B300 is an 8-GPU server platform built on NVIDIA Blackwell Ultra architecture. Each GPU carries 270 GB of HBM3e memory and fifth-generation NVLink interconnects, designed for large-scale AI training and high-throughput inference workloads.
Get started with NVIDIA RTX PRO 6000 Blackwell Server Edition on Nebius
Launch your first instance in minutes or talk to our team to find the right setup for your workload.