And would you believe it, OpenAI is not going to dry up and blow away? #
**Thursday. **Wish me luck. I’m taking my oldest to a park playdate today for her upcoming nursery school cohort. The goal will be a stiff upper lip and a bag full of pouches. Today, we’re looking at OpenAI’s early Q3 numbers, the Startup Media Collective, and then rounding off our coverage of OpenRouter’s sale by looking at Stripe’s investment memo. To work! — Alex
📈Trending Up: Goingzerofortwoin market interventions …Moderna shares, on news of a cancer breakthrough…cryptocurrency prices…the value of crypto stocks…the Clarity Act? …non-financial EU tech penalties…Musk as the GOP piggybank…custom AI chips…Azure’s growth rate? …📉Trending Down:YouTube-Netflix relations…being shocked…Alibaba shares, despite quick AI growth…shock, surprise…A/B testing…
**Keep an eye on this: **Nikkei reports that China is “restricting or delaying exports of key aerospace and optical materials to Taiwan,” including critical inputs for fiber optics, photonics, and semiconductor manufacturing. The stuff that Taiwan excels at, in other words. We’ve generally considered the CCP threat to global chips through the lens of an invasion; economic pressure, however, appears to be the lever du jour.
OpenAI, so dead it’s going to go public
Yesterday, we examined the idea that OpenAI is going to zero. Our verdict? Not likely. The company’s $122 billion fundraise from earlier this year, paired with reaccelerating growth after a slack Q2 (OpenAI grew just 18% from the first quarter to the second), gives the company more than enough inertia to stay afloat. Not merely that: OpenAI’s recent results could give it license to go public as soon as the fourth quarter.
Quickly, OpenAI CFO Sarah Friar told employees that while Anthropic may take the wrapper off its S-1 filing shortly and list in September, her company will take a little more time. OpenAI’s mega fundraise in March provides “flexibility” regarding timing, the executive said.
Why might it take OpenAI one quarter longer than Anthropic to go public? Anthropic is coming off two legendary quarters; OpenAI is not. But in the fourth quarter, OpenAI will have its third-quarter results to highlight, and with the AI giant posting 35% total run rate growth and 50% expansion in its enterprise sales (per Friar), the company should have a third quarter worth boasting about early in the fourth. A perfect time to go public, apart from the holidays.
Recall that OpenAI has already reached a $40 billion run rate. That’s up sharply from its second-quarter results. So, Anthropic should list in the final month of Q3, and OpenAI either in early Q4 or the first quarter of 2027. Unless something changes. A hallmark of the AI era is slow change, right? Surely our timing prognostications will be to the minute.
Tech’s media network is maturing
Turner Novak, one of my favorite venture capitalists, published this banger yesterday:
If that reads like Old Church Slavonic to you, let me help. Novak is recounting the list of podcasts and livestream shows that startup founders and venture capitalists regularly appear on. Even more, the groups often have explicit connections to technology power-nexuses or, at a minimum, the ability to secure marquee names from the tech industry. OpenAI bought TBPN (one of its co-founding hosts was previously an entrepreneur at Founders Fund). MTS Live is backed by a16z and the firm’s Erik Torenberg in particular; Torenberg is a partner at the venture firm and part of its New Media initiative. Sourcery is an interview show that covers high-profile startups; it’s sponsored by Brex, a former venture darling. MOTS is * Members of Technical Staff, *a podcast that has its own, you guessed it,
fitness challenge(and interviews with technology operators). Dwarkesh is tech’s public intellectual, and someone I thank for introducing me to
Sarah Paine. Senra references the Founders podcast by David Senra, which went viral earlier this year for its
discussion with a16z co-founder Marc Andreessen about introspection. And
Invest Like The Bestis a Patrick O’Shaughnessy joint that interviews tech’s leading lights. Not to mention All In, my former podcasting home This Week in Startups, ETN, Novak’s own Peel Pod –the list goes on and on and on. Which is to say that the technology industry has done it. It has its own, owned media network. This has been a long project. Who among us fails to recall Future.com and its early demise? But if you have enough money, early missteps can’t slow you down. And today you can consume purely tech-approved media in a circuit, just without (mostly) needing to answer pesky, uncomfortable questions. Why not simply sit in a circle and encourage one another?
The now-confirmed Stripe-OpenRouter deal is worth $7.5 billion, with $1.5 billion going to the startup’s founders and $6 billion to its backers. Notably, a quick scroll of X indicates that every venture capitalist I have ever met wrote the first check into OpenRouter, believed in it from the get-go, and is very happy to have taken a contrarian bet on a founder who had already grown a startup to a $13 billion valuation.
I get it. DPI is rare. Getting some is a win. Getting a lot is a big win. And venture capitalists are in the sales business, so chest-thumping after a score is to be expected. Ring the gong. I am not going to frown at joy.
Anyway. Stripe’s investor letter discussing the OpenRouter deal leaked — I assume it was written for public consumption — detailing a very interesting argument for why OpenRouter fits neatly into the Collison world. We’ll quote at length to preserve tone (emphasis added):
[W]e see capital and intelligence are becoming the two digital flows undergirding every business.Up until now, every developer has needed a straightforward and reliable way to manage their revenue pipeline, and serving this need gave rise to Stripe. Going forward, however,every developer will also need a straightforward and reliable way to manage their intelligence pipeline. This observation first led us to OpenRouter. […]We think that there are deeper reasons to pursue integration [with OpenRouter] besides convenience, however. Our experience in working with our customers has led us to realize that
intelligence is special: It is expensive, heterogeneous, and constantly changing.As with financial capital, businesses must reason about cost and return of every unit in a deliberate and granular way. How valuable is this task? With which models can it best be handled? Who will pay, and when, and what is the time-value of that delay?
Boiling business down to capital and intelligence is reasonable. We might have phrased the concept as ‘financial capital and human capital’ in the pre-AI era. Today, synthetic intelligence products can replace elements of human knowledge and labor, shifting our framing.
Regardless, providing dead-simple tooling to both drive maximal financial capital while ensuring minimal intelligence capital *spend *is an attractive pairing. As the AI industry is already building on Stripe — “88% of the Forbes AI 50 (including OpenAI and Anthropic) are building on Stripe (most of the remaining 12% are pre-monetization)” — bringing the two critical “digital flows” under one roof makes good sense.
The deal makes directional sense, mind, but you can still quibble on price without running afoul of Stripe’s argument. Nuance. It’s possible!
There’s more in the letter, including that Stripe thinks the world reached the singularity at the start of the year, a deliberately “fuzzy” marker. What does the onset of the singularity mean for the payments giant? That it “saw a large inflection in long-run trends (for example, a huge increase in the rate of new firm creation),” which makes the AI takeoff seem rather welcoming instead of worrisome.
Your definition will vary. But increasingly competent agents, ever-better (and cheaper) AI models, and the use of AI tools to accelerate the development of their next generation (we’re still a ways off from RSI, but today’s sparks ignite tomorrow’s fire) do put us in at least a period of faster-tempo technology change.
“The singularity appears to be accelerating our core business,” Stripe writes. In numerical terms, the company disclosed in its investor letter (earnings report?) that H1 2026 revenue grew 41% compared to the first half of 2025, with free cash flow rising 43% over the same time period. The payments giant claims that by the end of the year, ten of its products will “generate more than $100 [million] of net revenue,” while the “profitability” of its core payments business allows it to buy other companies sans dilution, with its share count lower today than “three years ago.”
Using free cash flow to purchase smaller companies while shrinking float is investor catnip. Forget dilution; Stripe is offering its shareholders concentration. That’s usually something that doesn’t occur until after a company has gone public. It’s almost as if Stripe is a public company larping as a private shop. Almost. Here’s to more of this, now brought to you by Stripe:
Speaking of “expensive, heterogeneous, and constantly changing” intelligence: Callosum just announced a $100 million Seed round to provide “heterogeneous compute” to the world, pairing chips (Cerebras, Rebellions, Axelera AI) and models against decomposed AI workflows to serve up the “orders-of-magnitude leaps in capability and cost of intelligence that the economics of AI now demand.” Callosum frames its work as the engine that brings AI-self-sufficiency to the world, arguing that compute “[h]eterogeneity is also how sovereignty becomes reality rather than aspiration: frontier-level intelligence running on whatever silicon a nation can access, with no single supplier holding all the keys.”
- Atomic led the round, with the UK Sovereign AI Fund participating. Per the company, this is the fund’s first-ever investment. Given the sovereignty part of Callosum’s pitch, I’m not surprised.