| There are numbers in TFA that explain the current CapEx craze, but they're not called out. (Also it's a big selective in which numbers it reports, see below.) > The Bundesbank finds that about half of German firms using AI do so for 5% of working hours or less. That seems to be from [1] and matches what the St Louis Fed reports here [2]: 50%+ of American use AI weekly, but only 6% of working hours. Google's report [3] finds similar "broad but shallow use" where many people use AI for a small subset of tasks. Now consider that all the hyperscalers are already extremely crunched for compute capacity. They are drowning in demand, and have been reporting this for the past several quarters. Nothing illustrates this better than the fact that Google of all companies -- whose massive infra footprint has always been considered a killer advantage in the AI race -- had to go rent capacity from SpaceX! And this is at only ~6% usage at work; imagine what it will take to get to even 30%, let alone 100%! This is why these companies are feverishly scrambling to build more data centers even as Wall St punishes them for their insane CapEx spend. The trillion $$$ question, of course, is whether this will all be profitable. There are many sources we could consider, but let's use one from TFA itself [4] which it selectively quotes as: > According to Mr Yotzov’s study, nine in ten executives report no impact of AI on their firm’s productivity over the past three years. The same source also says: > ...these same executives predict sizable effects over the next 3 years, predicting that AI will boost productivity at their firms by an average of 1.4%, raise output 0.8%, and cut employment 0.7%. So these executives clearly plan to spend more on AI. If the 1.4% seems small, consider that labor compensation is ~50% of global GDP, or $55 trillion. In a simplistic "what the market will bear" sense, even a 1% efficiency boost is "worth" 0.5T annually. Now if 1.4% seems high, look through [5] (and also similar numbers from Germany in [1] BTW.) These are the numbers AI companies have dancing in their eyes. Their challenge, of course, is to capture all that value, but to do so you first need to capture AI usage, and for that you need compute capacity, and hence the current CapEx splurge. I do agree with TFA that the biggest impact will come from organizations "reworking their processes." However I fear what this really means is significant job losses. [1] https://cepr.org/voxeu/columns/generative-ai-german-firms-di... [2] https://www.genaiadoptiontracker.com/ [3] https://blog.google/innovation-and-ai/technology/research/un... (discussion: https://news.ycombinator.com/item?id=49020335) [4] https://www.nber.org/system/files/working_papers/w34836/w348... [5] https://aleximas.substack.com/p/what-is-the-impact-of-ai-on-... |