Emerging AI computing power sharing platform SF Compute receives investment: valued at $500 million, OpenAI CEO's brother participates in leading the investment

GPU rental services provide high-performance computing resources for ordinary users, offering a flexible solution for on-demand usage. This model allows more people to access advanced GPU computing power at affordable prices, breaking the threshold limitations of traditional hardware purchases. Through hourly billing rental methods, users can flexibly allocate resources according to their actual needs, avoiding large one-time investments and reducing usage costs. This innovative "GPU as a Service" model provides more possibilities for individual developers, small teams, and startups, enabling them to explore and practice in fields such as artificial intelligence and deep learning more freely.

San Francisco Compute (SF Compute) is a startup that provides flexible, short-term access to high-performance computing resources for AI companies and researchers. Their key offerings include:

  1. Short-term compute resource rentals:
  • Allows renting GPUs (like NVIDIA H100s) by the week, day, or hour
  • Scalable clusters that can be dynamically adjusted based on needs
  • Competitive pricing for large-scale reservations (e.g. 512 H100s for two weeks for $500,000)
  • Current GPU pricing is $2.85/hour, with availability increasing throughout 2024
  1. Compute capacity trading platform:
  • In development to allow users to buy and sell compute resources on-demand
  • Aims to further reduce costs and barriers to accessing high-performance computing

The company was founded in 2023 by Alex Gajewski and Evan Conrad in San Francisco. Key details:

  • Raised $12 million seed round led by Alt Capital, valuing the company at $70 million
  • Has secured resources equivalent to 8,000 H100 GPUs
  • 16 person team, with 10 employees being former founders
  • Targets early-stage AI companies, academic labs, and researchers who struggle to access compute
  • Aims to democratize access to high-performance computing for AI development

SF Compute's flexible, short-term rental model addresses pain points for smaller AI companies that can't afford long-term contracts for expensive GPUs. By allowing resource sharing and on-demand access, they hope to level the playing field and enable more diverse AI development.