I built an AI datacenter simulator from scratch inside @WorkBuddyAI
using their Hy4 model and wanted to see how far I could push browser game dev

you start with $10M and have to build a profitable AI inference datacenter from zero

buy H100 / H200 / B200 racks, accept customer workloads, manage power + cooling and try to hit $1M in monthly revenue before you either:

run out of cash
melt the datacenter

customers range from AI coding startups to voice + video companies, each with different compute requirements and economics

and every workload pushes up:

GPU utilization
power usage
temperature
revenue

pretty fun little way to learn how inference economics actually work while also testing how far these models are getting with Three.js + game dev

prompt you can try yourself:

"Build me a polished browser game called AI DATACENTER SIMULATOR. Start me with $10M and let me buy H100, H200 and B200 GPU racks, accept AI workloads, manage power + cooling and scale toward $1M in monthly revenue. Make heat the main mechanic, with the 3D datacenter visibly getting hotter as utilization increases. Add a few simple random events, proper win/lose states and keep the whole thing visual, polished and easy to understand."

@WorkBuddyAI @TencentHunyuan