INTERACTIVE TOOLS
Benchmarks
Not the usual LLM benchmark. These measure the economics of delivering a token — the silicon, the building it sits in, and the policy wrapped around both.
GPU · Utilization
At what utilization do GPUs turn a profit?
Rent accelerators, serve an open model, sell the tokens. Eight open models on NVIDIA, AMD, and Google silicon across five regions — find the break-even point.
Interactive
DC · Buildout + Servicing
What does it cost to build the hall and serve the token?
Facility capex per MW, silicon spend, power, PUE, and depreciation — roll it all into a fully-loaded $/GPU-hr and a cost per million tokens.
Interactive
Gov · Policy
What does government control do to the economics?
Procurement EOs, export controls, residency mandates, energy regulation — move the levers and watch break-even shift on the same hardware.
Interactive