Crusoe recognized as an NVIDIA Exemplar Cloud
Crusoe Cloud has earned NVIDIA Exemplar Cloud validation for large-scale AI training on NVIDIA HGX B200, running 8 frontier models across multiple precisions on 512 NVIDIA GPUs, powered entirely by renewable energy.
Crusoe has achieved NVIDIA Exemplar Cloud validation for training on NVIDIA HGX B200 systems, confirming that Crusoe Cloud delivers the consistent, end-to-end performance value that large-scale AI training demands. This milestone reflects work across every layer of the stack: facility design, GPU orchestration, and networking, all engineered to ensure customers extract optimal performance from NVIDIA's advanced Blackwell-based platform.
What is the NVIDIA Exemplar Cloud?
The NVIDIA Exemplar Cloud initiative sets a transparent benchmarking standard for cloud providers. Rather than testing isolated components, it evaluates real AI training and inference workloads end-to-end using NVIDIA's Performance Benchmarking suite, measuring throughput across large-scale LLM training scenarios with production-grade software stacks.
Crusoe's validation ran on 512 NVIDIA Blackwell GPUs across a broad set of frontier models, including GPT OSS 120B, Nemotron4 340B, Llama 3.1 70B, Llama 3.1 405B, Qwen3 235B, DeepSeek V3 671B, Grok1 314B, and Nemotron-H 56B, using both NeMo and Megatron-Bridge frameworks and covering FP8, BF16, and NVFP4 precisions. That breadth matters: it's not a single benchmark on a favorable workload, but consistent, validated performance across model families, sizes, and precision formats that real AI teams actually use.
This reflects a broader reality: NVIDIA is the only platform that runs every frontier and open-source AI model in production today (from Llama and DeepSeek to Qwen, Grok, and Nemotron) across the precision formats and frameworks AI teams actually deploy.
Passing this evaluation means Crusoe Cloud delivers workload-level performance aligned with NVIDIA's reference architecture. For AI teams evaluating infrastructure, it's an objective, NVIDIA-validated signal that training jobs will run efficiently and predictably on Crusoe, with no guesswork required.
Unlocking the full potential of the NVIDIA Architecture
NVIDIA HGX B200 is a generational leap in AI compute. Built on NVIDIA's Blackwell architecture, it delivers up to 4x faster LLM training and 15x higher inference performance compared to Hopper-generation GPUs, while improving energy efficiency by 12x — translating directly into higher compute per watt, lower token cost, and a longer useful life for every GPU deployed.
These gains are driven by the second-generation Transformer Engine with FP4 and FP8 precision support, purpose-built for large-scale generative AI and mixture-of-experts models.
But hardware alone doesn't guarantee performance. Extracting the full capability of NVIDIA Blackwell GPUs requires infrastructure engineered to match its power density, direct liquid cooling (DLC) requirements, and high-bandwidth networking demands. Crusoe's AI-optimized data centers and cloud platform are purpose-built to do exactly that.
Powered by renewable energy in Norway
The NVIDIA HGX B200 system used for Crusoe's Exemplar Cloud validation ran in Norway, powered entirely by hydroelectric power, making it 100% renewable. This matters beyond sustainability: the naturally cool Nordic climate reduces the facility-level cooling load required to support HGX B200's direct liquid cooling (DLC) infrastructure, improving overall system efficiency and uptime.
Achieving NVIDIA Exemplar Cloud validation on renewable-powered infrastructure demonstrates that sustainable energy and peak AI performance aren't a tradeoff. They go hand in hand.
What this means for AI teams
For customers building on Crusoe Cloud, this validation delivers practical confidence across three dimensions:
Performance you can count on. NVIDIA Exemplar Cloud validation isn't a spec sheet. It's a real-world benchmark. Your AI work will perform consistently and predictably at the scale you need. This suite of training workloads taxes the system at scale in preparation for both pre-training and RL post-training work.
Infrastructure built for Blackwell from the ground up. Crusoe's platform is engineered around the power density, DLC requirements, and architecture of NVIDIA Blackwell GPUs, and plugs directly into NVIDIA's full developer ecosystem — CUDA, NeMo, Megatron-Bridge, and the broader catalog of frameworks and libraries millions of AI developers already use. So you get the full capability of the hardware without compromise.
Direct access, real support. Crusoe Cloud provides access to NVIDIA HGX B200 systems today, with trial runs available to validate your workloads before any long-term commitment. When you need help, you work with solutions engineers who specialize in AI infrastructure, not a support ticket queue.
As NVIDIA continues to advance its accelerated computing platforms, Crusoe's vertically integrated approach ensures infrastructure keeps pace, so customers can push the boundaries of what's possible without being slowed down by the underlying stack.
Get started now
To get started with NVIDIA HGX B200 on Crusoe Cloud, visit crusoe.ai/cloud or contact our team.
