Many users praised charts on AI training data startups reaching $8.5B revenue while others dismissed the $100B valuations as inflated or unsustainable given margins and multiples.
Based on 26 visible X reactions from 185 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@deedydas That $8.5B headline includes 60-70% contractor payouts. Strip those out and I don't see margins that justify anything close to $100B.
@deedydas Eight and a half billion in revenue from training data is huge.
@deedydas Eight and a half billion in revenue from training data alone is huge.
@deedydas $100B on $8.5B revenue. Inflated.
Scale, Surge, Mercor, and Handshake control over 75% of the market.
@deedydas That $8.5B headline includes 60-70% contractor payouts. Strip those out and I don't see margins that justify anything close to $100B.
@deedydas This list is gold.
@deedydas ❤️
jokes aside data in the next wave of ai is going to be even more valuable imo than in previous eras of ML more and more tasks are going to be offloaded to agents => verification, testing and any kind of optimization now happens over a data distribution instead of via static tests => complicated shapes and distributions of data are ever more critical => more $$ and time on data not just big labs but companies deploying agents at scale will need this model of checking / optimizing their agents as well
funny how this tweet led to all data/env vendors going no actually we make way way more trust 😩 https://twitter.com/deedydas/status/2076124392711696455
Many users praised charts on AI training data startups reaching $8.5B revenue while others dismissed the $100B valuations as inflated or unsustainable given margins and multiples.
Based on 26 visible X reactions from 185 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@deedydas ❤️
jokes aside data in the next wave of ai is going to be even more valuable imo than in previous eras of ML more and more tasks are going to be offloaded to agents => verification, testing and any kind of optimization now happens over a data distribution instead of via static tests => complicated shapes and distributions of data are ever more critical => more $$ and time on data not just big labs but companies deploying agents at scale will need this model of checking / optimizing their agents as well