What if a few AI companies end up with all the money and power?

Noah Smith · Noahpinion · April 13, 2026 at 00:50 · ⏱ 5 min read  | Read on Substack ↗
Summary
Noah Smith argues that the AI industry is no longer in a bubble because agentic coding has become a massive revenue driver, with Anthropic overtaking OpenAI in enterprise sales. He warns that a cybersecurity arms race could create a powerful moat for leading AI companies, potentially leading to extreme economic concentration and inequality, rather than the widely feared risks of superintelligence or job displacement.
  • Agentic coding — where AI builds apps and runs analysis autonomously — has become the killer app, generating enormous revenue.
  • Anthropic has likely surpassed OpenAI in revenue due to its enterprise focus, while OpenAI focused on consumers.
  • Anthropic's computing costs are much lower than OpenAI's, putting it on track to turn a profit faster.
  • Anthropic delayed its new model Mythos because it was too good at hacking; the model found critical vulnerabilities missed for decades.
  • Cybersecurity is inherently adversarial, forcing defenders to pay top dollar for the latest AI models, creating a profit moat.
  • The expansion of AI into adversarial fields like quant trading, litigation, and fraud prevention could create additional revenue sources and moats, leading to extreme inequality if a few companies dominate.
Read time 5 min
Length 5,205 chars
Category macro
More from Noahpinion