Human-Centered

Marlowe Enables Next-Gen AI and Data-Driven Discovery at Scale

248 NVIDIA H100 GPUs across all seven Stanford schools along with a team to help accelerate science — a shared computational instrument for the research that wasn't possible before.

Stanford campus from above Stanford Oval · aerial
248H100s
Online now
87%
Utilization · 24h
190+groups
Research groups
7schools
Across campus
2.1MGPU-hours
Delivered
Latest Updates All updates →
talks seminars Apr 22 NVIDIA & Marlowe: Training Robots with World Foundation Models talks seminars Mar 11 NVIDIA New AI Models: Open Models — Open to Build maintenance Feb 20 Planned Maintenance Window: February 24-25 talks seminars Feb 18 NVIDIA & Marlowe: Post-training Language Agents with NeMo RL and NeMo Gym research milestone Feb 7 Milestone: Training POC for 80B-Parameter Brain Model on Marlowe Completed talks seminars Jan 28 Compute Resources @ Stanford and Beyond talks seminars Nov 12 NVIDIA PhysicsNeMo: Community Models and Dataset Integrations talks seminars Oct 29 Building Scalable, End-to-End Generative AI with NVIDIA NeMo Framework on Marlowe talks seminars Sep 24 Graph Neural Networks & LLMs in PyG on Marlowe talks seminars Aug 13 Using CuPyNumeric and the Legate Ecosystem for Multi-GPU Scaling on Marlowe education Aug 7 DataSci 211: Accelerating Research with Marlowe talks seminars Jul 16 NVIDIA Clara for AI-Enabled Healthcare and Life Sciences on Marlowe talks seminars Jun 25 NVIDIA Nsight Systems for Profiling Code on Marlowe talks seminars May 21 Distributed Training & Marlowe Multi-GPU Best Practices talks seminars Apr 29 Marlowe Featured at 2025 Stanford Data Science Conference talks seminars Apr 23 Marlowe GPU Computing Foundations: Architecture, Applications, and Acceleration press Dec 15 Stanford Welcomes First GPU-Based Supercomputer
Upcoming

Events

All events
Spotlight

Researcher spotlight

Research Showcase →
Dan YaminsDan Yamins

Dan Yamins on training brain-scale neural networks on Marlowe.

Validated 80-billion-parameter brain-inspired neural network training on Marlowe, then launched PSI2-30B — a counterfactual world model — in the first hero run on the system, across 24 nodes.

Associate Professor of Psychology and Computer Science

Watch

Marlowe on film

All videos
Access

From access request to first Marlowe job.

Faculty PIs sponsor accounts; postdocs and students may be flexibly added. New PIs start with 5,000 free GPU-hours.

  1. PI submits a sponsorship formA short statement of research area and expected GPU-hours.
  2. Marlowe Team provisions your accountTypically within a week of submission.
  3. PI receives an onboarding email from the Marlowe teamSlurm, storage, and first-job workflow.
  4. Submit your first jobThe Marlowe team helps troubleshoot first-job issues and early workflow questions.
Apply for access
Training

Onboarding & training

Workshops and seminars run throughout the year — see what's coming up above, and find recordings and step-by-step guides in the documentation.