how dependent are AI valuations on a maximalist version of AI build-out and uptake which, in terms of scale, speed, and freedom from scrutiny, might be less plausible under a different administration?