Nebius achieved NVIDIA Exemplar Cloud validation on the GB300 platform for training workloads. The certification targets hyperscale AI training, not inference, distinguishing it from broader cloud GPU certifications.
Key facts
- Nebius is first cloud provider with GB300 Exemplar Cloud for training.
- Certification validates hyperscale AI training performance on GB300.
- Nvidia invested $2B in Marvell for NVLink Fusion partnership.
- Microsoft began validating Nvidia's Vera Rubin NVL72 system.
Nebius announced it has been validated as an NVIDIA Exemplar Cloud on the NVIDIA GB300 platform for training workloads. The company claims it is the first cloud provider to receive this specific certification from NVIDIA for hyperscale AI training. [According to Nebius achieves NVIDIA Exempla]
Why this matters more than the press release suggests
This validation signals a shift in how cloud providers compete for AI training workloads. Unlike the NVIDIA Partner Network (NPN) cloud certifications that cover inference and general GPU usage, the Exemplar Cloud designation on GB300 specifically validates performance for large-scale training jobs. Nebius can now claim verified performance metrics that enterprise AI teams require when selecting infrastructure for multi-billion parameter model training.
What the certification covers
The GB300 platform, part of NVIDIA's Blackwell architecture, targets next-generation AI training. The Exemplar Cloud validation includes benchmarks for training throughput, scalability across nodes, and network performance. Nebius did not disclose specific benchmark numbers or the size of its GB300 deployment. [According to Nebius achieves NVIDIA Exempla]
Competitive landscape
Microsoft, Google Cloud, and AWS each hold various NVIDIA cloud certifications, but none have yet announced Exemplar Cloud status on GB300 for training. Microsoft recently began validating NVIDIA's Vera Rubin NVL72 system, a next-generation architecture beyond GB300. [According to MSN] The race for training-specific cloud validation is intensifying as hyperscalers compete for AI training contracts worth hundreds of millions of dollars.
Nvidia's $2B Marvell investment
Separately, Nvidia invested $2 billion in Marvell to deepen the NVLink Fusion partnership. NVLink Fusion is critical for scaling GPU clusters in training clouds, directly benefiting providers like Nebius that rely on Nvidia's interconnect technology. [According to MSN]
What to watch
Watch for benchmark disclosures from Nebius on GB300 training performance, and whether Microsoft, Google Cloud, or AWS announce their own GB300 Exemplar Cloud training certifications in the next 60 days.
[Updated 01 May via gn_gpu_cluster]
Nebius reported strong Q1 for FY2026 alongside the certification, indicating financial momentum as it scales GB300 deployment [per Nebius]. The company did not disclose specific benchmark numbers or deployment size, but the quarterly strength suggests growing enterprise adoption of its training cloud.
[Updated 01 May via gn_gpu_cluster]
In a parallel move, Nebius acquired AI infrastructure startup Eigen AI for $643 million to accelerate its GB300 deployment and expand its training cloud capabilities, according to a separate filing [per Nebius]. The acquisition bolsters Nebius's engineering talent and proprietary software stack for large-scale model training, potentially shortening its time-to-market for enterprise customers.
