The AI arms race just became an equity story — and retail investors are about to get their first real taste of foundation model economics.
The Summary
- OpenAI is in talks to raise at least $30 billion at a $1.4 trillion valuation, while DeepSeek is securing at least $12 billion ahead of its own IPO next year
- Both companies are moving from private funding rounds to public market preparation, signaling a maturation phase for foundation model businesses
- The real test: whether public market investors will accept compute-intensive businesses with thin margins and uncertain path to profitability at these valuations
The Signal
OpenAI at $1.4 trillion puts it roughly at Amazon's current market cap, except Amazon has 20 years of proven cash flow and OpenAI has a chatbot subscription business and an API that hemorrhages money on every call. The valuation isn't insane if you believe foundation models become infrastructure. It's bonkers if you think they become commodities.
DeepSeek's $12 billion raise is the more interesting data point. A Chinese foundation model company going public while US-China tech tensions remain high tells you two things: one, the Chinese government sees AI as critical enough to allow foreign capital exposure, and two, DeepSeek's domestic revenue must be strong enough to justify the risk. The company has been aggressively undercutting OpenAI on API pricing while claiming comparable performance on certain benchmarks.
"Foundation models are either the new cloud infrastructure or the new search engine — and the difference determines whether these valuations hold."
The IPO prep matters because it forces actual business model scrutiny. Private investors could handwave compute costs and path to profitability. Public markets won't. Consider:
- OpenAI's reported annual revenue run rate is around $11 billion, implying a 127x revenue multiple
- Inference costs remain stubbornly high despite Moore's Law progress in AI chips
- Enterprise adoption is happening, but mostly for productivity tools, not the transformational use cases that justify trillion-dollar valuations
Jamie Dimon's comments on data centers and AI risks from London add context. JPMorgan is simultaneously building its own models and investing in AI infrastructure. When the largest US bank says data center capacity is a constraint, that's not analysis, that's a buying signal. Banks don't publicize constraints unless they're already positioning to profit from solving them.
The Implication
Watch the S-1 filings when they drop. The revenue mix, gross margins, and customer concentration will tell you whether these companies are infrastructure plays or just very expensive API wrappers. If enterprise revenue is growing faster than consumer, that's bullish. If compute costs aren't declining as a percentage of revenue, that's a red flag.
For anyone building with these models: IPO pressure means pricing will change. Either prices go up to satisfy public market margin expectations, or open source models get good enough that foundation model companies get squeezed. Plan for both.