Harvey margins went from 50% to -50%, until they served their own model. Why more should train? AI Autonomy

Harvey margins went from 50% to -50%, until they served their own model. Why more should train? AI Autonomy

Shout out to Ted Theodoropoulos 👉🏽 https://lnkd.in/p/e8GvFM4T

At The Beacon Collective UK I said: bless the braves who share numbers 🙏🏽

Training is a treadmill yet offers a benefit: autonomy, not just sovereignty

Custom models can run on-prem and on EU soil 🇪🇺 so...

...this opens the door to firms buying GPUs and running AI autonomously

When Harvey, Thomson Reuters and others start offering this business model...

...we'll see more firms like Latham & Watkins buy GPU's 👉🏽 Karin Johansson 😉

I did not predict negative margins, I calculated them 👉🏽 https://lnkd.in/gcp7TuBK

Tokenomics: Harvey paid ~$3.00 per million tokens and earned $2.00...own model on rented GPUs runs $0.09 to $0.25, my estimate 👇🏽