The entire AI capex cycle depends on whether next gen models (Fable-tier) unlock exponentially more token spend. Open source models are "good enough" for most day-to-day work, so closed source spend has to be a step function better AND result in more token spend.
thinking harder about this, the most likely outcome IMO: 1. strong compute shortages force cost-performance over capabilities 2. open source models continue to cross threshold of usefulness 3. more compute crunch, less revenue growth for closed models 4. frontier labs taper off

