OpenAI closed a $122 billion funding round at a post-money valuation of $852 billion this week — the largest private fundraise in the history of capital markets. SoftBank co-led; the rest of the syndicate is the usual suspects (a16z, MGX, T. Rowe-advised accounts, others). The proceeds will fund compute and data-center commitments that have already been announced.
Sierra raised $950 million the same week. Moonshot AI raised $2 billion. Cerebras filed for an IPO at $26 billion.
I have been around long enough to remember when WorldCom was worth $180 billion, when Cisco was worth more than Microsoft, when Yahoo turned down $44 billion from Microsoft because it wasn't enough, and when Lehman Brothers was a perfectly fine counterparty. Some of those valuations were validated; most weren't. The difference, in retrospect, was always the same boring thing: the cash conversion math.
OpenAI at $852 billion implies, very roughly, that the market expects something like $50–80 billion of stable annual operating profit from the entity within a decade. To get there, OpenAI has to capture something like 20–30% of the global enterprise AI software market, at GAAP gross margins north of 50%, against rivals that include Anthropic, Google, Microsoft (which is also OpenAI's largest infrastructure provider, awkwardly), and a Chinese cohort that has independently demonstrated frontier-class models on far less compute.
That is possible. It has been possible before. Some of the names I listed above made the math work. Most didn't.
What's new this cycle is that the capital base is real. The dotcom bubble at peak had Pets.com burning $147 million of paid-in capital before it folded. OpenAI is burning sums per quarter that exceed the entire VC-funded loss of every dotcom startup combined. If the math works out, the upside is enormous and historic. If it doesn't, the bezzle is going to be the largest private-capital writeoff anyone has ever seen.
I am not a forecaster. I am a guy who has watched several of these cycles and has noticed that, in retrospect, the people who said "this time is different" were sometimes right and were always more often wrong. The honest position right now is "we will know in three years."
What I am willing to say is that anyone telling you they know whether $852 billion is the right number is lying, selling something, or both.
The bull case is that this isn't really a tech valuation; it is a valuation of an emerging civilizational input. If AI does to white-collar productivity what electricity did to manufacturing, the addressable market isn't the global software industry — it is the global wage bill. At that scale, $852 billion is a rounding error. I take the argument seriously even though I don't bet on it.
- The largest private fundraise in history is now an OpenAI round, beating the previous record (also OpenAI) by a wide margin
- Sierra's $950M Series at $15B is the second-largest enterprise-AI raise of the week
- Cerebras's IPO filing at $26.6B includes Q4 2025 revenue of $510M (up 76% YoY)
Don't treat $852 billion as a fact. Treat it as a price the market is willing to pay today for a story about the future. Stories change. If you are operating, build like the AI bet is real but the OpenAI-specific bet is not necessarily — diversify your model providers, watch your costs, don't tie a critical dependency to a single vendor's ten-year cap-table outcome.
— Hank