, June 14, 2026

Wall Street Can't Read and That's OpenAI's Problem Now


When OpenAI and Anthropic make their IPO prospectuses available to the public, investors are going to have to learn about a whole new economy.

  •   1 min read
Wall Street Can't Read and That's OpenAI's Problem Now

Table of content

Wall Street analysts spent the last three years pretending they understood transformers and attention mechanisms. They didn't. They still don't. Now they have to learn what a token is before OpenAI's S-1 drops and reveals their entire revenue model is denominated in something that sounds like Chuck E. Cheese currency.

SpaceX already runs on this. Employees get compensated partially in tokens that represent computational resources. Wall Street looked at this and nodded politely like someone's aunt listening to an explanation of Bitcoin at Thanksgiving. Zero comprehension. Pure panic behind the eyes.

The token economy means companies don't sell software anymore. They sell compute by the fractional unit. Input tokens, output tokens, context window tokens. It's a metered utility model except the utility is making a chatbot tell you how to fix your garbage disposal. Analysts who still describe SaaS companies as having "sticky recurring revenue" are going to have to explain to their clients why OpenAI charges by the syllable.

Anthropic's IPO will be worse. Their Constitutional AI approach means they're selling tokens that promise not to teach your teenager how to make a pipe bomb. That's the product. Well-behaved tokens. Premium pricing for computational goodness. Try building a discounted cash flow model around that.

Every investment bank is scrambling to hire someone under 30 who can explain API calls without using the word "basically" fourteen times. They're failing. The talent pool is busy building wrappers around GPT-4 and calling themselves founders.

Retail traders think they're ready because they bought NVDA at $180 and felt smart. They have no idea what a context window is. They think latency is something their wife complains about. When they see "cost per million tokens" on an income statement, they're going to assume it's a typo and buy anyway.

Wall Street wanted AI exposure. Careful what you wish for, because now you have to learn computer science from a prospectus written by lawyers who also don't know what a token is.

Photo by on Unsplash

Related Posts

If you got something out of this, Phil McCandlestick accepts tips. No pressure — the chart was free.

Leave a Tip