Top Takeaways from Benedict Evan’s: AI Eats the World

OpenAI raised $6.6Bn from Microsoft, Softbank and Thrive Capital on a valuation of US$157Bn. It took Microsoft 20 years to get to US$20Bn.
The top use cases for LLMs in production are: Software/IT, customer service & marketing
Debate around whether LLM scaling is going to continue or slow down - no conclusion yet
Today’s LLMs cost around $1Bn to build according to Anthropic CEO. Future could be $5Bn to $10Bn
In 2024, the Big Four cloud players - Alphabet, AWS, Meta, Microsoft - spent US$220Bn on Capex
LLM-based software no longer has near-zero marginal cost: meaning, “launch free, go viral, work out revenue later” does not work
LMAREMNA - as a measure of model quality
“Everyone in tech is giving someone else’s business model away for free”
Air Canada pay damages after chatbot lies to grieving passenger about discount
“AI gives you infinite interns”
“internet made distribution free, LLM makes creation free”?
Accenture booked US$1Bn in Aug 2024 for GenAI related work
“There are two ways to make money - you can bundle or unbundle” - Jim Barksdale
AI is just what doesn’t work yet: AI —> Smart —> Auto —> Just software
Shopify powered $270Bn of ecommerce in last 12 months, 35% of AMZN GMV
Software eats cars: Robotaxi trips in CA reached 300k in Aug 2024