OpenAI is now worth $852 billion on paper. That makes it more valuable than Berkshire Hathaway, more valuable than Visa, more valuable than Johnson & Johnson — a company that makes actual physical products that people cannot live without.
OpenAI makes chatbots.
Let me be precise: OpenAI makes genuinely impressive, widely used, occasionally transformative chatbots. But the gap between what it is and what $852 billion says it is represents one of the most dangerous financial fictions since the dot-com era.
Let’s Do the Math Nobody Wants to Do
OpenAI surpassed $25 billion in annualized revenue. That sounds enormous. But investors valued it at 34x revenue — in an industry with negative margins on its core product.
Running large language models at scale costs more than selling access to them. For every dollar OpenAI collects from ChatGPT subscriptions and API calls, it spends more than a dollar on compute. The path to profitability requires either dramatically higher prices, dramatically lower compute costs, or a completely different business model than the one they’re currently running.
Amazon, Nvidia, and SoftBank didn’t invest $122 billion because they believe in the mission. They invested because they each have enormous financial interests in AI infrastructure being worth a lot of money. Nvidia sells the chips. Amazon sells the cloud compute. SoftBank has been burned by tech bets before and desperately needs a winner. Their participation in the round is not a signal of OpenAI’s value — it’s a signal of their own need for the AI narrative to hold.
The Revenue Is Real. The Business Model Isn’t.
$25 billion in revenue is impressive. But here’s what it doesn’t tell you:
How much of it is sustainable? A significant chunk of OpenAI’s enterprise revenue comes from companies in early AI adoption phases — pilot programs, proof-of-concept deployments, experimental integrations. Many of those contracts will not renew at the same scale once companies figure out what AI actually does for them versus what they hoped it would do.
How much is structural? Enterprise software businesses live or die by churn. If Microsoft — which owns a significant stake in OpenAI — starts competing more aggressively with its own Copilot products, OpenAI’s enterprise pipeline shrinks overnight.
What happens when competition arrives at scale? Gemini 3.1 is already leading 13 of 16 major AI benchmarks at one-third the API cost. Grok 5 has 6 trillion parameters. Anthropic is approaching $19 billion in revenue. OpenAI’s moat — the thing that justifies a premium valuation — is eroding in real time.
The IPO Setup
OpenAI is reportedly taking early steps toward a public listing, potentially as soon as late 2026. The $122 billion funding round at $852 billion valuation was not primarily about capital. OpenAI doesn’t need $122 billion to operate.
It was about setting a reference price.
When OpenAI goes public, investment banks will point to the most recent private valuation. Retail investors — people reading headlines, people with 401(k)s, people who want exposure to “the AI thing” — will buy shares at or above that reference price. The early investors, the VCs, the executives with equity — they will sell.
This is the same structure that produced the dot-com bubble. Pets.com was real. Webvan was real. They had real users, real revenue, real venture backing. They were also worth nothing close to what their valuations implied, and the people who bought at the top lost everything.
What $852 Billion Actually Requires
For OpenAI’s valuation to be justified at a reasonable 25x earnings multiple, the company would need to generate approximately $34 billion in annual profit. It currently generates negative profit on its core product.
To reach $34 billion in profit, OpenAI would need to either transform into a fundamentally different company than it is today — or the entire global AI market would need to consolidate around OpenAI in a way that gives it pricing power it currently does not have and shows no signs of achieving.
Neither of those things is happening by the time they go public.
Invest in AI. Use AI. Build with AI. But don’t confuse the genuinely transformative nature of the technology with the genuinely unjustifiable price of the companies selling it. The technology is real. The valuations are a story. And stories have endings.
Wow