AI Infrastructure SimCity

AI Infrastructure SimCity A playable world for deciding where AI data centers should go and how they should be built. Built on Muse by Meta for the Meta Global AI Developer Hackathon. Gigawatt campuses are announced faster than anyone can evaluate them. Each is a billion-dollar bet on power, water, land and people, and the people who must decide rarely share one picture. AI Infrastructure SimCity gives policy makers, government officials, environmentalists, companies and investors the same model, the same evidence, and a verdict they can argue with. A live registry tracks 27 real campuses (29,600 MW); a deterministic estimate engine with fully documented methodology scores each one, and four Meta models do four distinct jobs through the Meta Model API: Muse Spark gathers (research, structured extraction, the chat planner, web search via search grounding), Muse Wonder decides (the go / conditional / no-go verdict, its conditions, risks and the build plan), SAM 3.1 looks (measures buildings, water, solar and open land from satellite imagery and feeds the hectares into the estimate), and Muse Image shows (a concept render of the facility in the landscape SAM measured). Every agent run is logged with its model and tokens (106 runs on record). A build plan opens directly in a playable simulation: spin the globe, break ground, and walk a navigable 3D site with racks, cooling, thermal view and an electrical room. Stack: Next.js 15, TypeScript, three.js, Prisma on Postgres, Tailwind. Estimates are screening-grade, not engineering or investment advice. Live app: https://ai-infrastructure-simcity-dev-rel-eng.vercel.app Technologies: Meta Model API, Muse Spark 1.3, Muse Wonder 1.0, SAM 3.1, Muse Image 1.0, Next.js, three.js, Prisma, Postgres Team: solo, Emmanuel Acheampong Built during the event; the repository was created October 26, 2026, during the hackathon window.