Nebius Joins the PyTorch Foundation as a Platinum Member

The PyTorch Foundation announced that Nebius has joined as a Platinum member, the foundation's top tier of membership. Nebius is an AI cloud company listed on Nasdaq under the ticker NBIS and headquartered in Amsterdam. Ryan Hanrui Wang, Senior Vice President of Applied AI at Nebius, will take the company's seat on the Governing Board and also serve as its representative on the Technical Advisory Council. Mark Collier, Executive Director of the PyTorch Foundation, welcomed the addition.
Platinum membership carries one seat on the Governing Board. That board sets foundation policy through its bylaws and mission and vision statements, which define the scope of foundation initiatives as well as its technical vision and direction. The Technical Advisory Council is the foundation's technical governance body, so holding positions on both means Nebius will have a voice in both the strategic and technical tracks of the organization. The PyTorch Foundation describes itself as a vendor-neutral home for the open source intelligence layer used to train, optimize, serve and run models, and it operates under the Linux Foundation. Its project portfolio spans the core PyTorch framework plus vLLM, DeepSpeed, Ray, Helion and Safetensors.
Nebius comes into the role with existing ties to several of those projects. Startups and enterprises already use PyTorch to train and fine-tune models at scale on Nebius AI Cloud. The company has a long record of supporting vLLM with compute for development, testing and benchmarking, including multi-node work on large models such as DeepSeek R1. Its Applied AI team contributes inference and post-training optimizations to open source projects, and that work has already been incorporated into the PyTorch Ecosystem. vLLM is a widely used open source engine for high-throughput LLM inference, and it has become a common target for teams that want to serve models efficiently on their own hardware or in the cloud. Post-training refers to the stages after pretraining, such as fine-tuning and alignment, which is where a lot of production engineering effort now sits.
As a Platinum member, Nebius says it plans to contribute to PyTorch and vLLM performance on the latest hardware platforms and to share what it learns from running inference and post-training at production scale. That is a meaningful commitment in practice: performance work on new accelerators is typically where ecosystem fragmentation shows up first, because kernels, compilers and serving stacks must be tuned per chip. A cloud provider with real production workloads is usually one of the few parties positioned to find those problems early. This suggests the practical payoff for developers may be less about a new feature announcement and more about incremental tuning and fixes landing upstream in projects they already use.
Wang's background lines up with that mandate. He leads inference, post-training and agentic systems at Nebius, and he previously co-founded and served as CEO of Eigen AI, which Nebius acquired in 2026. He holds a Ph.D. in Computer Science from MIT, where he created the SpAtten sparse-attention framework and Hardware-Aware Transformers. His open source work has passed one million downloads and is part of the PyTorch Ecosystem, and he is an MIT Technology Review Innovator Under 35. Sparse attention and hardware-aware model design are both areas directly relevant to making large models cheap and fast to serve, so his dual board and TAC roles give Nebius a technically grounded position in foundation discussions.
In his statement, Wang framed the membership as a way to collaborate more deeply with the community and contribute expertise in high-performance infrastructure and applied AI, and said the goal is to help developers and enterprises move more easily from experimentation to production. Collier described the ecosystem's strength as collaboration on AI's toughest scaling challenges, and said Nebius's direct participation advances projects like vLLM and PyTorch for both high-throughput inference and large-scale training.
The wider trend here is that foundation membership has become a way for infrastructure vendors to invest in the software layer their customers depend on. Nebius sells compute and a full-stack platform for training and deployment, so improvements to PyTorch and vLLM that make models run faster or train more reliably make its own hardware more useful, and its upstream contributions become a shared asset rather than a proprietary edge. A likely trade-off is that open source performance work is public by design, so competitors who run similar workloads can benefit from the same optimizations. For an infrastructure provider, the differentiation then has to come from operations, capacity and tooling rather than from hidden kernel tricks.
No timeline or specific deliverables were attached to the announcement. The stated intent is broad: performance work on current hardware and knowledge-sharing from production deployments. Readers tracking this should watch for upstream patches and benchmarks in PyTorch and vLLM, plus any documentation or guidance that comes out of the production-scale experience Nebius says it will share.
Why it matters: developers running training or inference on Nebius AI Cloud, or on PyTorch and vLLM generally, now have a major cloud vendor formally in the foundation's governance, which in my reading raises the odds that performance fixes for new hardware reach upstream projects faster. Product teams evaluating serving stacks may see more vendor participation in benchmark and optimization work, though the practical effect depends on what actually lands in the repositories. Companies considering foundation membership can also read this as a template: contribute compute and engineering, and gain board-level influence over direction.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
The PyTorch Foundation, the vendor-neutral home for the open source intelligence layer, is announcing today that Nebius has joined as a Platinum member. Nebius, the AI cloud company, is building... The post Nebius Joins the PyTorch Foundation as a Platinum Member appeared first on PyTorch .