Microsoft Support for the National AI Research Resource Pilot
Microsoft’s commitment to the NAIRR pilot
Microsoft is delighted to contribute as a lead supporter of the National AI Research Resource (NAIRR) pilot. (opens in new tab) The mission of the NAIRR pilot aligns with our commitment to broaden AI research and spur innovation by providing greater computing resources to AI researchers and engineers in academia and non-profit sectors. As part of our commitment to the NAIRR pilot (opens in new tab), Microsoft is offering $20 million in Microsoft Azure compute credits. Read below for more information on what we’re offering researchers, and check out the tabs above for exciting research directions, eligibility details, and how to apply.
What’s available for researchers
High-performance computing resources
We provide high-performance computing resources to meet various workload requirements.
For larger workloads: NVIDIA A100 GPUs with InfiniBand (opens in new tab) connectivity are available for select NAIRR Pilot projects to develop world-class AI for science. Researchers can request access to NDm_A100_v4 series (opens in new tab) virtual machines (VM) designed for high-end Deep Learning training and tightly coupled scale-up and scale-out HPC workloads. Deployments with up to 256 GPUs will be considered with preference given to projects ready to onboard immediately that can run the GPUs at full utilization for 2-3 months.
For smaller workloads: Researchers can request NCads_H100_v5 (opens in new tab) SKU series powered by NVIDIA H100 NVL GPU and 4th-generation AMD EPYC™ Genoa processors; NC_A100_v4 (opens in new tab) SKU series powered by NVIDIA A100 PCIe GPU and third generation AMD EPYC™ 7V13 (Milan) processors; and Azure HBv2 (opens in new tab) and HBv3 (opens in new tab) series VMs. We recommend the NC series GPUs for AI training workloads where the model can be trained on either a single NC_A100_v4 instance with a maximum of 4 80G A100 GPUs (320 GB GPU memory), or a single NCads_H100_v5 instance with a maximum of 2 94G H100 GPUs (188 GB GPU memory). Proposals can request multiple NC_v4 or NC_v5 instances to run parallel workloads.
Leading-edge models
We’re dedicated to delivering frontier and open models to further research, including popular large language and vision foundation models. Researchers can access these models through our Azure OpenAI Service (opens in new tab) and the Azure AI model catalogue (opens in new tab). The Azure AI model catalog features over a thousand models (opens in new tab) including models curated by Microsoft, OpenAI, Hugging Face, and more. (Access is available for all catalog models, except those listed as requiring Azure Marketplace.)
Domain-specific resources
We recognize the benefit of domain-specific resources for some research projects. For this reason, we offer health and life science models (opens in new tab), innovative tools for chemistry and materials science research via Azure Quantum Elements (opens in new tab), and tools for research and development on AI fairness, accuracy, reliability, and interpretability (opens in new tab). Researchers can also access datasets including (opens in new tab)Fields of the World (opens in new tab).
Scientific support
We’re dedicated to helping researchers make the most of these resources. To support domain experts who are new to machine learning, researchers can apply for a research collaboration with Microsoft’s scientists and engineers, including researchers at Microsoft Research and Microsoft’s AI for Good Lab.