@JaredSleeper Not entirely sure...
Reasoning models and agents shifted compute toward test-time, and KV caching for long-context agents, which is exactly why memory demand isn't just a training story
AI labs can't serve all the inference demand they already have, they're compute-limited
@capybaraReborn 2) Made moves in up the stack. It acquired Eigen AI for inference optimization, which is about speeding up and improving efficiency for open-source models, opening revenue pockets for diversification beyond its core business. Ofc brings along integration / execution risks
@capybaraReborn 1) your housing analogy undersells the scale. It's like saying contractors would have no business if the whole world needed building. We're early on AI's S-curve; training has driven demand so far, but inference will take it to a new level
@NonnoInterista@farted94 Ma quanti anni ha sto essere patetico?๐๐๐๐ se arrivo a 70 anni e faccio sti video da sfigato 12enne con una voce da pervertito sopprimetemi
@86_longo Non ti giocare la credibilitร per qualche like/retweet.
Se hai una notizia BOMBA ok.
Se devi scrivere che Renato Sanches ha l'asciugamano del Milan per fare il bidet allora no.