AI2050 is a Schmidt Sciences fellowship initiative co-chaired by Eric Schmidt and James Manyika. The 2025 cohort, announced November 5, 2025, named 21 Early Career Fellows and 7 Senior Fellows, totaling $18M in three-year awards. With this cohort, AI2050 reaches 99 fellows across 8 countries and 42 institutions across its four announced cycles.
Early Career awards target postdocs and pre-tenure faculty doing multidisciplinary, hard-to-fund research. The 2025 Early Career list includes Zongyi Li (NYU incoming, neural operators for scientific computing), Tess Smidt (MIT EECS, equivariant ML), Surbhi Goel (Penn CIS, theoretical foundations for why modern AI works and fails), Brian Hie (Stanford), Natasha Jaques (Washington), Martin Schrimpf, Lindsey Raymond (joining MIT EECS / Economics / Schwarzman College in 2026), and Ellen Zhong (Princeton).
Senior Fellows are selected by closed nomination with no application path. The 2025 Senior cohort includes Surya Ganguli (Stanford, developing a scientific foundation for explainable AI through analytical theories of diffusion models and reasoning systems), Luke Zettlemoyer (Washington, NLP), and Krishna Pillutla (IIT Madras, privacy-preserving fine-tuning and reasoning methods to prevent privacy-sensitive leakage from foundation models).
The thematic scope is broader than safety: research topics span AI scientists, safer and more reliable models, AI for biological and medical discovery, and theoretical foundations. AI2050 sits adjacent to but distinct from Schmidt Sciences' AI Safety Science program, which funds projects rather than individuals and has a tighter safety-specific frame.
For sovereignty-minded readers, AI2050 is one of the principal individual-researcher fellowships funded by a single U.S. private donor at this scale. The selection skew toward academic researchers (versus industry labs) and the multi-year structure pushes against the short-cycle, milestone-driven pattern that dominates most AI funding. The cohort year-over-year growth (now 99 fellows total) makes it a useful index of who Schmidt Sciences considers a key academic AI researcher worth multi-year support.
Recipient
Multiple researchers
Funder
Schmidt Sciences · foundation · US
Trustworthy AI science, interpretability, alignment, 'AI for science.' Multiple programs (AI2050, Safety Science, AI Safety RFP).
Primary source
https://www.schmidtsciences.org/2025-ai2050-fellows-announcement/