Hofvarpnir Studios is a 501(c)(3) nonprofit founded in 2021 by Shauna Kravec and Nova DasSarma with the mission of bridging the gap between academic AI safety researchers and high-performance computing. The organization designs, builds, and maintains GPU clusters that support a range of machine learning workloads, from fine-tuning large language models to CPU-intensive reinforcement learning tasks. Its primary partners include the Center for Human-Compatible AI (CHAI) and the Steinhardt Lab at UC Berkeley, as well as researchers at UT Austin, MILA, and Redwood Research. By providing shared compute infrastructure, Hofvarpnir enables safety researchers to investigate complex neural networks that would otherwise be out of reach.
Funding Details
- Annual Budget
- $453,562
- Monthly Burn Rate
- $37,797
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $1,443,540
- Fiscal Sponsor
- -
Theory of Change
Hofvarpnir believes that academic AI safety research is constrained by lack of access to compute. By providing shared GPU infrastructure to university labs and independent researchers, it removes a key bottleneck, enabling more researchers to run experiments on frontier-scale models, publish safety-relevant findings, and ultimately contribute to the development of AI systems that are aligned with human preferences and behave as expected. Greater research output from underfunded academic groups expands the overall technical AI safety field.
Grants Received
from Open Philanthropy
Projects
No linked projects.
People
No linked people.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Apr 2, 2026, 9:53 PM UTC
- Created
- Mar 20, 2026, 2:34 AM UTC