Ask HN: Why doesn't OpenAI start 100 companies?
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
A Hacker News user asked why OpenAI doesn't start companies using its own models instead of selling tokens.
The thread weighs token sales against building products as AI prices fall.
Background
The question was posted as an Ask HN thread by user ex-aws-dude, who called the title "a bit of an exaggeration" and framed the real issue as whether OpenAI or Anthropic could earn more by building companies that use their models than by selling tokens. It drew a handful of comments and a follow-up analysis by AI Pricing Guru, a pricing-site editorial team, which stressed the thread is a discussion, not an OpenAI announcement. The debate lands as model prices are widely described as commodifying and declining.
How it unfolded
- Oct 22, 2026User ex-aws-dude posts the Ask HN thread, which draws 16 points and 4 comments.
- Oct 22, 2026Commenters including CoolestBeans, digitaltrees, paxy and Rodmine respond with competing takes.
- Oct 23, 2026AI Pricing Guru publishes an analysis, noting the thread is a discussion and not evidence OpenAI has tested autonomous businesses.
Who’s saying what
- Support
- CoolestBeans says OpenAI is already making "a galaxy of attempts at application," but that selling tokens through long-running agents is the only thing provably working for revenue growth, and near-term revenue matters given borrowing costs.
- Support
- digitaltrees says OpenAI is moving into the app layer and pushing into consulting, and that it avoids starting service companies mainly because those still need people on the ground.
- Caution
- paxy argues running one business is easier than running 100, and that a company needs far more than good software to succeed.
- Analysts
- Rodmine invokes "during a gold rush, sell shovels," saying OpenAI's heavy in-house custom silicon investment makes it better off letting others compete on applications.
- Analysts
- The AI Pricing Guru editorial team argues strong models don't automatically create profitable companies, since they don't supply paying customers, distribution, reliable operations or accountability.
Still unverified
CoolestBeans' assertions about OpenAI's runway and what its decision makers believe about a Jevons Paradox demand effect are commenter claims; digitaltrees' claim that OpenAI bought a friend's consulting firm is also a commenter assertion. AI Pricing Guru states its profitability reading is its own economic interpretation, not a claim about OpenAI's private strategy.