The Twenty Minute VCAnthropic Inference Costs Skyrocket |TikTok Deal Closes |The IPO Market:Wealthfront & EquipmentShare
At a glance
WHAT IT’S REALLY ABOUT
Big tech deals, soaring AI costs, and shaky IPOs reshape markets
- Brex’s $5.15B sale to Capital One is framed as a strong founder outcome but also a case study in “hubristic financing” after raising at $12B in 2021.
- The TikTok divestiture is viewed as primarily geopolitical rather than economic, creating a potentially “cheap” asset for the chosen U.S. owners while leaving key algorithm control questions unresolved.
- Anthropic’s higher-than-expected inference costs spark a debate: model provider margins are improving, but application-layer companies may face rising token burn that becomes existential without pricing power.
- OpenEvidence’s jump from a $1B to $12B valuation in a year is praised as a high-quality AI business, but investors question whether the true doctor/pharma TAM can support the implied long-term revenue scale.
- Public markets appear selectively open: EquipmentShare exemplifies the “scale + growth + profitability” IPO, while Wealthfront highlights the dangers of subscale listings and post-IPO talent/compounding challenges.
IDEAS WORTH REMEMBERING
5 ideasA $5B exit can be both a win and a warning.
The panel celebrates Brex as a heroic outcome from zero to $5B, while noting that raising at peak prices creates a “weird tax” at exit when outcomes are judged against prior paper marks.
Peak valuations force companies into bigger promises—and bigger reputational hangovers.
“Hubristic financing” rewards momentum and competitiveness in the moment, but it also commits founders/employees/customers to an implied path (own the category) that may be impossible once growth normalizes.
Brex’s deal resets valuation expectations for spend-management peers.
Using the reported ~7x revenue framing, the discussion suggests Ramp’s much higher private multiple deserves scrutiny, even if Ramp is operationally “winning,” because real-money buyers just anchored a lower comp.
Capital One’s Discover rails may make Brex worth more inside a bank than standalone.
A key strategic angle is interchange economics: owning a closed network (Discover) could let Capital One capture more economics and compete with structural advantages versus independent fintechs.
AI inference costs may rise for app companies even as per-token prices fall.
The group argues that better capabilities drive more usage (more tokens), so total spend can increase; founders should model inference costs going up, not down, especially as agents run more continuously.
WORDS WORTH SAVING
5 quotesThe bad feelings last for a day, and the 5 billion lasts forever , right? So you'll get over it.
— Rory O’Driscoll
At some point, NVIDIA puts will be a great buy 'cause every semiconductor cycle for the last 40 years has ended up in a massive downswing. I ain't buying them today.
— Rory O’Driscoll
We will actually burn an infinite amount of tokens if we can.
— Jason Lemkin
If you walked into your board meeting and said, "Hey, good news, guys. Inference costs are, are going down 30% this year 'cause our, our IT team's really good at managing costs," I would, I would throw my mouse at the monitor.
— Jason Lemkin
We talk about is SaaS dead or what's going on. I'm, I worry this is the next final act. I think it's the final nail.
— Jason Lemkin
High quality AI-generated summary created from speaker-labeled transcript.
