a16z · Oct 4, 2026 · 55m · 40m saved
With David George, General Partner at Andreessen Horowitz, where he leads the Growth fund
OpenAI, Anthropic and SpaceX's AI business together bring in about $120 billion of revenue, David George said, and add more revenue each month than every hyperscaler except Amazon. Most of it comes from about 30 million coders.
The usual framing asks which part of the AI stack wins: frontier models or open source, labs or applications. George, who runs Andreessen Horowitz's Growth fund, said the question is wrong.
"The answer in AI is probably and." — David George
David George, General Partner at Andreessen Horowitz, on the a16z Show, has led the firm's Growth fund since it was set up, writing its first check into Databricks at a $6 billion valuation. The conversation first ran on Uncapped, the podcast of a Benchmark partner, who interviewed him.
The full interview is covered here so you can skip it. 55 minutes of audio, 15 minutes of reading.
Here are the 12 arguments that matter.
Key Takeaways
AI has diffused into less than 5% of the business economy, George said, and almost all revenue comes from about 30 million coders
There are 1.5 billion knowledge workers who barely use it yet
Data center capacity is sold out until 2028, and he sees no plausible way for new capacity to go unused
The frontier labs take 95% or more of AI dollars today, but he expects open models and apps to grow alongside them
Legal AI is about 12 months behind coding, and clients now demand that their law firms use Harvey
Consumer AI is still in "skeuomorphic mode": most people use ChatGPT the way they used a search engine
Self-driving could grow ride-hailing 10x again, as Uber and Lyft did to San Francisco taxis
Throwing money at an AI company makes it better, which was not true in the software era, so the power law is stronger
This product cycle rates 9 or 10 out of 10, he said, while the capital cycle is only about a 6
1. It's All Going to Work
The host opened with a remark from his Benchmark partner Eric Vishria on another podcast: in AI, "it's all going to work," whether the question is open source or frontier models, Nvidia or new chip companies.
"So I am very, very closely aligned with Eric on this point that it's all going to work." — David George
George said spending on AI infrastructure has just passed the railroad build-out as a share of GDP
One of them put it at about 3% of GDP. George expects it to grow about 10x over many years, which he hopes will drive GDP growth rather than simply take a larger share of it.
Almost all of AI's revenue comes from a small group of heavy users
"So it's like 30 million coders in the world who are the vast majority of the revenue." — David George
There are about a billion consumer users and 1.5 billion knowledge workers, he said, and inside the 30 million it is a steep power law, with the top million spending heavily. Coding and writing are what work in the enterprise today. Legal is taking off and is about 12 months behind.
"It's like less than 5% diffused into the B2B economy." — David George
