KERALA RISING · FROM EXPORTING PEOPLE TO EXPORTING IDEAS core team, ops, and the tiny inference bill, the institution is self-funding, and because inference is a rounding error, that crossover depends almost entirely on the revenue ramp, not on the cost of the AI. The central finding Stated as plainly as possible: the artificial intelligence is not where the money goes, and it is not what determines whether the institution becomes self-funding. Verification is so cheap that it disappears into the rounding of the operating budget, and it gets cheaper every year. The build is a one-time, milestone-gated capitalisation. And the shared revenue lines, growing with every program and every corridor, are what carry the institution to operating self-sufficiency. The state is asked to capitalise the machine once, against milestones, with a published unwind protocol if the first program’s independent assessment is negative. It is not asked to fund the AI in perpetuity, because the AI barely costs anything, and it is not asked to subsidise the institution forever, because the platform is built to pay for itself. A worked illustration, with the caveat stated first The numbers that follow are by intent round and explicitly illustrative; their only purpose is to show 253