Is OpenAI worth $300B?
And what are the risks?
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Happy Tuesday, everyone! Welcome to Q2. Yes, that means we’re on the cusp of earnings season — get hype! — as the year’s IPO cycle gets into gear. With CoreWeave public, Circle ramping up, and eToro and StubHub out with S-1s we’re all but feasting. Oh, and Cerebras has been unleashed!
POTUS is expected to drop a raft of new trade barriers tomorrow. Notably, Heritage has come out — in contrast with its historical economic views — in favor of Trumpish economics. Seeing the group get lit up by their own former fans is a great place to mine smiles if you have the time.
Today, on this last day before global trade is further harmed, let’s get to work. — Alex
📈 Trending Up: Musk as political foil … TikTok? … photonics … staying the course … the need for soft power … cooked books … what the fuck … the god of the sea … Transgender Day of Visibility … collective spine … lol, lmao even …
Quote of the Day: “Mr Trump is a man on a mission, determined to jack up tariffs in order to remake America’s economic model; or, more accurately, to wind it back by a century.”
Rounds of Note: $600 million for Isomorphic, $146 million for Temporal … and rumors of Cursor closing $625 million …
📉 Trending Down: Logistics … naming rights … the CHIPS Act? … inflation in Europe … merit … market sentiment … doing your expenses …
Fuck it, $40B
After Wiz sold to Google for $32 billion in cash, and X ‘sold’ to xAI for $33 billion worth of equity, who expected to see a new, largest tech transaction so quickly? Enter OpenAI with a new $40 billion raise that it pegs at a $300 billion valuation post-money.
CNBC notes that the round — led by SoftBank for $30 billion with capital from Microsoft aboard — makes OpenAI “among the world’s most richly valued private companies.”
The round is an enormous mark of trust for the AI giant, which is building an enormous new constellation of data centers, recently picked up Microsoft’s option for access to CoreWeave’s GPUs, and now has more capital than all startups in Europe ($12.5 billion) and Asia ($15.6 billion) raised in Q4 2024 combined.
Note: The Information reported in the runup to the deal that about “half of [the] capital […] will go toward Stargate, a joint venture between OpenAI, SoftBank and Oracle.” So, the headline figure is likely not all new operating cash for OpenAI.
Calling OpenAI a startup is abuse of the term, but we lack a truly better moniker for high-growth, unprofitable technology companies that remain private even as they scale into the billions of dollars worth of yearly revenue.
Is OpenAI worth $300B?
The best way to answer this question is to ask yourself what you would do with $100,000 that you had to invest right away and had access to the recent OpenAI round. Index funds, bonds, high-yield cash, or OpenAI equity?
In more numerical terms, OpenAI is probably not worth $300 billion today if we calculate its worth entirely based on known historical results. OpenAI is reported to have generated revenues last year of $3.7 billion (and losses of around $5 billion). OpenAI’s 2024 top line and new, shiny 2025 valuation put it at around 81x revenue.
Of course, if we took its Q4 revenue and converted it into a run rate, its effective multiple would fall. As it will this year in more concrete terms. The company has told investors that it expects revenues of around $12.5 billion this year. If OpenAI hits that mark on the nose, and its valuation stays static the rest or the year, OpenAI will exit this year with a trailing price/sales ratio — a far more conservative metric than an ARR multiple — of 24x. That’s not even the richest we see amongst public cloud companies.
So, is OpenAI worth $300 billion today? If the company’s 2025 plans are realistic, and the company has made a quarter’s worth of progress thereof, yes.
What are the risks?
Where to start. A rival closed-source AI giant could scoop market share from OpenAI if it launches better models with greater availability and lower prices. Anthropic, DeepSeek, Mistral — it’s a long list. Or, open-source AI models could limit market willingness to pay for closed-source AI model access. Meta could, in theory, snake a chunk of the market that OpenAI expects to win this year.
Continuing, the pace at which the world adopts AI technology could slow, limiting overall market growth. This risk could manifest amongst corporates, consumers, or both.
Other risks abound. Geopolitics could limit OpenAI’s ability to sell in certain markets, harming its ability to grow. Chinese AI could eat a larger piece of the market than expected. Domestic or foreign regulation could slow down AI progress, making it less useful for longer. Hell, copyright issues alone could prove lethal for the company.
OpenAI could also stumble in operating terms as it works on a high-wire transition from quais-lab-and-non-profit-and-for-profit Frankenstein into something sleeker and more prepared to meet a competitive future. On that note, Musk could greatly harm OpenAI by halting its ability to become more investable.
What if everything goes right?
The above-listed concerns are speculative and un-targeted. Any closed-sourced AI model company faces open-source risk and competition from its peers. All AI giants face copyright concerns. There isn’t a single model company out there that is free from geopolitics — you get the idea.
But what OpenAI does have that is unique is insane reach. In its funding post — a spare two paragraphs — reports that “500 million people [use] ChatGPT every week.” That’s a lot of folks who are using a service with a built-in paid tier. To reach its $12.5 billion revenue target, OpenAI has to secure a mere $25 per year from its weekly active userbase for one of its products to pull off its full-year growth goals.
The New York Times reports that OpenAI has more than 20 million paid accounts. At $20 per month — the lowest paid tier — and 20 million accounts, OpenAI could earn $4.8 billion from those subscribers alone this year.
OpenAI is not coasting. It’s defending its growth with new launches, something that we saw with its recent release of an image generation tool that captivated the Internet at large.
First, OpenAI said that demand for 4o’s image generator was melting its GPUs.
Then the company noted that it took five days from launch for ChatGPT to add “one million users in five days.” Yesterday, OpenAI said that it “added one million users in [one] hour.”
Finally, Lu Liu, the woman behind the new 4o image generating tool added yesterday that response to her project has been “insane,” and that she did not “expect [the recent image generator] launch could be as viral as chatgpt,” but that when “GENERAL intelligence meets people's creativity, it just goes viral very quickly.”
Hardly what you expect to see from a company that has lost its edge. Mix in regular improvements to OpenAI’s model base, and an upcoming “open source” project, and yeah, fine, if I was forced into our $100,000 question above, I’d probably give the money to Sam instead of buying more $FZROX. Fine, you’ve got me.
Closing out
CoreWeave picked a hard time to go public. The stock market is a bit of a mess at the moment as the world sorts out new trade realities. The GPU cloud wound up in headlines for shedding up to 10% of its worth yesterday, its second day of trading. Thankfully, it bounced back a bit before the close. and it is up in pre-market trading.
I don’t think any of my index funds hold CoreWeave, so don’t think that I am rooting for any single company in particular. No. I want CoreWeave to do well strictly so that other companies follow in its footsteps, and we get our hands on more S-1 filings.
More!
So much for bitcoin being a haven during troubled times. I keep half-expecting the crypto asset to appreciate during economic turmoil. But the supposedly hardest-asset-of-them-all doesn’t appear to be enjoying much upside from fiat chaos today.
Tesla deliveries are expected to drop tomorrow. If they are worse than expected, and Tesla shares drop, it could quite literally change the arc of global governance.