Hi everyone, thank you for today’s call (2025.05.22). Special thanks to @icarus for leading the session! This is the generated summary (short version, please find the long version here ): Today’s DeAI Working Group call for the Internet Computer focused on evaluating GPU server configurations (e.g., Supermicro, Gigabyte) capable of supporting high-powered GPUs (up to 600 watts), discussing detailed documentation of AI model requirements for agentic AI, and emphasizing strategic alignment of infrastructure investments with clearly defined use cases and economic viability. The group explored targeted decentralized AI applications, recommending a focus on practical mid to small-scale models (≤70B parameters). Next steps include collaboratively creating detailed comparison spreadsheets, initiating hardware tests, and formulating key strategic questions, alongside upcoming interactive sessions and workshops.
Links shared during the call:
- GPU Database | TechPowerUp
- Benchmark MLPerf Inference: Datacenter | MLCommons V3.1
- wip spreadsheet on accelerators and AI models: https://docs.google.com/spreadsheets/d/1a-NzZeh6D9smxovW7ByYOx7Ls1tRqBJyaS9gXVzRy-Y/edit?gid=0#gid=0
- Introduction by new member: www.metame.com