Warm GPUs should be available to everyone.

Warm capacity goes to whoever can reserve it. A team that can keep a GPU busy around the clock gets one that is always ready. Everyone else chooses between cold starts and paying for the hours their model sits idle. The hardware is not the scarce part. The way it gets allocated is.

Boardwalk pools work from models that get busy at different times. When one model gets quiet, another can use the room. One GPU stays warm across many tenants, so being ready does not depend on any single tenant being large enough to fund it alone.

We would rather be accurate than impressive. Warmth is a likelihood, not a guarantee, and we say so plainly instead of selling around it. Open models are worth running well, and that should not depend on the size of the budget behind them.