Nvidia validates Nebius HGX B300 racks for AI cloud training

Nvidia granted Nebius Exemplar Cloud validation for its production HGX B300 racks on October 8, 2026, confirming the systems Nebius rents to customers match Nvidia's reference architecture for training workloads. Nebius, which trades on NASDAQ as NBIS, also posted the fastest single-node B300 times in the closed division of MLPerf Training 6.0, covering Llama-3.1-8B and GPT-OSS 20B. Each HGX B300 node carries 8 Blackwell Ultra GPUs with approximately 2.1 TB of HBM3e memory and fifth-generation NVLink.

Photo: Matheus Bertelli / Pexels Nvidia validates Nebius HGX B300 racks for AI cloud training The AI cloud provider earned Nvidia Exemplar Cloud status for its Blackwell Ultra systems and posted the fastest single-node B300 times in MLPerf Training 6.0 Nvidia https://cryptobriefing.com/markets/nvidia/ has signed off on Nebius’s production HGX B300 racks, granting the AI cloud provider its Exemplar Cloud validation for training workloads. The approval, announced on October 8, 2026, confirms that the racks Nebius runs for paying customers match Nvidia’s own reference architecture. In a market where every cloud claims to be fast, this is the chipmaker itself vouching for the setup. Nebius, which trades on NASDAQ under the ticker NBIS, paired the news with a second result. Its systems posted the fastest single-node B300 times in the closed division of MLPerf Training 6.0. What Nvidia actually checked Under the program, cloud operators run a set of standardized benchmarking recipes on their live production hardware. The results are then measured against reference targets that Nvidia defines. Nebius cleared that bar after an engineering review and an analysis of its benchmark results. The point is to show that the hardware customers actually rent performs the way Nvidia says it should. Inside the HGX B300 The hardware at the center of this is Nvidia’s HGX B300 platform. Each node packs 8 Blackwell Ultra GPUs . AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. The platform carries approximately 2.1 TB of HBM3e memory. It also uses fifth-generation NVLink, Nvidia’s high-speed interconnect. Nvidia designed the system for two main jobs: training large models and running high-throughput inference. Nebius’s validation specifically targets the training side. On the benchmarking front, Nebius’s MLPerf Training 6.0 results covered two large language models: Llama-3.1-8B and GPT-OSS 20B. MLPerf’s closed division is the stricter category. Participants have to train the same models under the same rules, which makes the results more of an apples-to-apples comparison. A pattern, not a one-off This is not Nebius’s first round with Nvidia’s inspectors. The company previously earned Exemplar status on other Nvidia platforms, including the H200 and the GB300 NVL72 configurations. Nebius is also far from alone in deploying this hardware. The broader HGX B300 ecosystem includes server makers such as Supermicro, ASRock Rack and ASUS, along with cloud providers like CoreWeave. Validation results for other cloud providers sit outside this announcement. So the B300 crowd is growing, but Nebius is the one holding this particular certificate for its training setup. Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .