“The computational performance of leading AI supercomputers has doubled every 9 months“.
“Power requirements and hardware costs of leading AI supercomputers have doubled every year.“
“If the observed trends continue, the leading AI supercomputer in June 2030 will need 2 million AI chips, have a hardware cost of $200 billion, and require 9 GW of power.“
“We find that the share of private sector compute rapidly increased from less than 40% in 2019 to about 80% in 2025.” (see chart).
Exhibits are uploaded daily giving unparalleled insights into the inner workings of the company.
From Ben Evans – “There’s a lot of email on the daily struggles of running social networks (people worrying that US growth is slowing (2022), or that WhatsApp is overtaking Messenger (2013), and I didn’t know that Meta tracks consumer sentiment towards Meta the company. There also a few tantalising ideas – a project for a paid, premium version of Facebook from 2020, for example. And there are charts of how the ad load has doubled. But really, what you see across all of these documents is a relentless focus from the top on making the current thing work better while worrying about the next thing.“
Moore’s law stands out for its speed against improvements in other industrial processes.
“The speed of intercontinental travel rose from about 35 kilometers per hour for large ocean liners in 1900 to 885 km/h for the Boeing 707 in 1958, an average rise of 5.6 percent a year. But that speed has remained essentially constant ever since—the Boeing 787 cruises just a few percent faster than the 707. Between 1973 and 2014, the fuel-conversion efficiency of new U.S. passenger cars (even after excluding monstrous SUVs and pickups) rose at an annual rate of just 2.5 percent, from 13.5 to 37 miles per gallon (that’s from 17.4 liters per 100 kilometers to 6.4 L/100 km). And finally, the energy cost of steel (coke, natural gas, electricity), our civilization’s most essential metal, was reduced from about 50 gigajoules to less than 20 per metric ton between 1950 and 2010—that is, an annual rate of about –1.7 percent.“
What stands out from this deep analysis is the renewed optimism in China on advanced semiconductor design and manufacturing.
“Ren [Huawei founder] further said that he is leading a network of more than 2,000 Chinese companies who are collectively working to ensure that China achieves self-sufficiency of more than 70 percent across the entire semiconductor value chain by 2028. These predictions should be taken seriously.“
Stripe did $1.4 trillion payments volume in 2024 – it is worth reading their annual letter.
Topics include AI Economy (their data shows top 100 AI firms reach $5m ARR in 24 months vs. 37 months for their top SaaS firms in 2018), vertical SaaS, stablecoins, Europe.
Want a guide to what happened in the world of ecommerce in 2024? Read here.
The most fascinating development continues to be a move from made in China, to sold by China to now, with the rise of Temu and Shein, promoted by China.
Retail media has taken over from traditional TV in terms of advertising dollars.
“OpenAI’s new o3 system – trained on the ARC-AGI-1 Public Training set – has scored a breakthrough 75.7% on the Semi-Private Evaluation set at our stated public leaderboard $10k compute limit. A high-compute (172x) o3 configuration scored 87.5%.“
“This is a surprising and important step-function increase in AI capabilities, showing novel task adaptation ability never seen before in the GPT-family models. For context, ARC-AGI-1 took 4 years to go from 0% with GPT-3 in 2020 to 5% in 2024 with GPT-4o. All intuition about AI capabilities will need to get updated for o3.“
