If you are interested in using the new and expanded Empire AI GPU compute cluster, you can apply for an allocation using the process described below. This process is open from today to Sept 17 (midnight ET).
Empire AI is a consortium that includes Cornell and eight other New York State institutions, with support from New York State and philanthropy. We are now soliciting proposals to use the extended “Alpha” machine with 144 H100 GPUs, as well as the new “Beta” machine that is expected to come online in December. Beta will consist of 288 B200 GPUs with excellent networking interconnects and additional Grace CPU nodes for data processing. More details on Alpha+ and Beta are available from Empire AI here.
Through the two forms below you can apply for an allocation of Service Units (SU) for Empire AI. An allocation is the maximum number of SU you can use between December 2025 and November 2026. One SU corresponds to one hour of H100 compute, and one hour of B200 compute corresponds to two SU (further details from Empire AI here). This cost is supported by the Provost until November 2026. After that, an SU is expected to cost $0.50, which will provide Cornell PIs with a rate that is well below a typical AWS rate (https://aws.amazon.com/ec2/capacityblocks/pricing/) and other discount rates (https://cloud.vast.ai/?gpu_option=H100%20SXM).
Any faculty member with PI status from Cornell, Cornell Tech, and Weill Cornell Medicine can apply. To apply for an allocation, please submit the *two* short forms that you find linked below. We expect that demand for Empire AI will exceed Cornell’s share of Empire AI, at least until the “Gamma” expansion of Empire AI in 2027.