"New AI models use enormous amounts of computer power, more than many technology experts anticipated even a year or two ago" [a year or two ago]
The quote establishes that even expert forecasters failed to anticipate compute scaling within a 1-2 year horizon, directly supporting the claim that institutional budget cycles cannot keep pace with frontier model compute demands.
Reasoning from this article
The article frames the $9B request as a catch-up measure while simultaneously noting 'even larger sums will likely be needed in the future,' indicating this is a recurring structural deficit rather than a one-time gap. The pattern generalizes to any government or large institution whose procurement cycles operate on 2-5 year horizons while AI compute requirements scale on 12-18 month chip release cycles. This creates a permanent structural lag for bureaucratic actors relative to commercial AI deployers.