Today, we're parsing a few new funding rounds, and then going deep on the Stripe-OpenRouter deal. #
- Welcome to *. Cautious Optimism, a newsletter on tech, business, and power. Modestly upbeat
Monday. Nvidia is backing more data centers while the public turns against the AI infrastructure build out. Such are the times we live in.
Looking ahead, this week is going to be full of interesting progress reports. We’ll be treated to Baidu, Klarna, and Pony AI’s earnings on Tuesday; Bill.com, Webull, Weibo, and Opera report Wednesday; and Thursday will bring numbers from Alibaba and NetEase. Next week, we’ll hear from a host of SaaS companies, as well as chip giant Nvidia.
Today, we’re parsing a few new funding rounds, and then going deep on the Stripe-OpenRouter deal. To work! — Alex
📈Trending Up:Local AI…arguing for one’s place in the trough…influencing AI…AI regulation woes…ad Astra…the German far-right…corruption…tech political spending…📉Trending Down:US-South Korea relationsShein’s anticipated IPO worth…consumer spending growth? …nerd goggles…Japanese GDP growth viz. expectations…Moldovan GDP expectations…Thai GDP growth
The Rundown #
** Anthropic generates $126.4M/day in Q2:** With its revenue reportedly crossing $11.5 billion in the second quarter, it looks like Anthropic saw sales of around $5.3 million per hour in the period. The company seems to have more than doubled its Q1 revenues of $4.73 billion. Holy shit.
- Anthropic also reported “positive adjusted operating income in the second quarter.” That’s a country mile from GAAP net income, but it’s still good to see the AI labs start to generate profit — even if it’s heavily adjusted right now.
** Wispr raises $280M:** The popular AI dictation tool has now raised $361 million. Wispr makes a great product. I’ve used it. I once tried to train Dragon NaturallySpeaking to help it better understand my voice back in the Windows 98 era. It didn’t work. But voice dictation
doeswork today.
As part of its funding round, Wispr announced the “preview of [its] first proprietary speech model, Canto.” The news sounds like a quickly growing startup upping its product quality (Wispr claims Canto dramatically reduces transcription errors) while expanding its margins (it’s cheaper to serve your own model than to pay someone else’s margins).
I think in terms of untapped AI market, voice is pretty damn large given how much the average office worker types. The bull case makes sense here.
- Wispr is now worth $2 billion, which is a lot, but given its
impressive growth, I wonder how quickly its effective multiple will halve, and then halve again. ** Higgsfield raises $400M:** Higgsfield is now worth $5.4 billion against run-rate revenue of $700 million. Investors are therefore buying into the multimedia AI platform at less than 7x its run rate.
The modest multiple implies Higgsfield is operating in a market where margins are thin and switching costs low. Despite those headwinds, it has grown from a run rate of $500 million (June), and $50 million (September, 2025). Impressive.
It’s getting chippy out there: I was once told that a good sign of a bubble’s peak is rising levels of fraud. But what about pique? The Resentment Theory of Bubble Timing states that the more public mud-slinging you see, the deeper we are into a period of intense speculation that is causing the less fortunate to lash out.
Resentment is the mode du jour in tech these days. If you didn’t get an early gig at Anthropic, you may feel that there’s nothing you can do to catch up financially to your peers. This theoretical disconnect is part of what is driving the permanent underclass discourse.
Similarly, rancor about VCs as gatekeepers/kingmakers bubbled over in the form of aired dirty laundry back in June, when a host of founders shared their (partly anonymized) fundraising stories involving misbehaving investors.
That’s now come back the other way with a vengeance. Last week, we were presented with a host of fresh complaints:
- VCs are annoyed at foundersfor stretching the truth. - Founders are firing back that VCsare full of it, too. - Everyone’s arguing that VCs are handing off their thinking to AI models, driving groupthink. - And complaining that job-hopping investors are
leaving portfolio companies in the cold. In each example, we see people treating other people poorly as they scramble for a piece of the spoils while the getting is good. Long-term thinking is being subsumed by near-term greed. History sees itself becoming as it always has.
The more everyone else is getting rich or bending the rules, the more you might want to as well. Bubbles create massively disparate near-term outcomes. I do not yet know if our brand new Resentment Theory will prove prescient, but I think it’s a fair barometer to keep an eye on.
*The Resentment Theory of Bubble Timing also explains why so many tech folks hate Anthropic, despite it being a part of their very own world.*They backed the wrong horse and will thus lose out on massive gains!
I dig the Stripe-OpenRouter deal #
The Stripe-OpenRouter deal is a go. Bloomberg reports that the payments giant will pay $7 billion for the popular AI model router provider, which is a dramatic premium over the $1.3 billion valuation the latter accrued when it raised $113 million in May.
The transaction has generated quite a bit of skepticism, but reading through the punditocracy this morning, it seems founders and venture capitalists are more thrilled at the exit and the implied return for LPs than ready to throw stones at a potential head-scratcher.
As an OpenRouter customer and fan (its public datasets are brilliant) let’s consider the deal from a few different angles.
Here are three ways to think about the deal from a financial perspective:
Subscribe to Cautious Optimism to unlock the rest.
Become a paying subscriber of Cautious Optimism to get access to this post and other subscriber-only content.