The Twenty Minute VCAnthropic Inference Costs Skyrocket |TikTok Deal Closes |The IPO Market:Wealthfront & EquipmentShare
CHAPTERS
Big week of tech headlines: exits, AI costs, mega-rounds, IPOs
Harry sets the agenda: Brex’s surprise $5B+ sale, TikTok’s forced divestment structure, Anthropic’s inference margin reality check, OpenEvidence’s huge step-up, IPO signals, and Salesforce’s defense win. The hosts frame the episode as a tour through how capital, competition, and AI economics are reshaping outcomes in 2026.
Brex sells to Capital One for $5.15B: great outcome vs 2021 expectations
The group agrees the Brex exit is objectively strong—building to a $5B+ outcome quickly is rare—while acknowledging the emotional hangover from raising at much higher valuations in 2021. They discuss why the market still debates ‘good vs bad’ exits and how financing cycles create distorted benchmarks.
“Hubristic financing” and the one-day tax of selling below the last round
Jason and Rory unpack why companies raise at peak prices and why that later produces awkward exits—even if fundamentals are fine. They frame it as a structural feature of boom markets: companies must fund growth, compete, and recruit, but later face ‘promise vs reality’ when growth normalizes.
Does the Brex outcome help or hurt Ramp (and other peers like Navan)?
Brex’s sale validates the spend-management category while also creating a sobering public-ish comp for Ramp’s private valuation. The hosts argue Ramp ‘won’ operationally, but the acquisition multiple and a stronger strategic buyer entering the space complicate future IPO math.
TikTok divestment deal: cheap economics, heavy geopolitics, and ‘missing’ top-tier VCs
They treat TikTok’s resolution as primarily political rather than a normal market transaction, noting the apparent bargain valuation versus revenue. Jason questions why obvious “money-minting” venture brands weren’t prominently involved, hinting that unspoken constraints or risks likely shaped participation.
Anthropic inference costs come in higher: margins improving, but token demand explodes
Anthropic’s inference costs being above expectations becomes a springboard for a broader point: unit costs may fall, but usage expands faster, keeping total inference spend high. Rory emphasizes gross margin improvement trajectory, while Jason stresses that competitive AI products will keep burning tokens.
The SaaS mid-market trap: profitable, shipped an agent… but can’t fund the inference war
Jason describes a painful scenario for $50–200M ARR companies: they did the ‘right’ things—efficiency, break-even, AI features—yet now face a new capital requirement to stay competitive. Rory agrees it’s real and says the only durable escape is delivering enough differentiated value to charge more than the token bill.
Is there an AI bubble? Watch capex commitments (TSMC as the ‘tell’)
They argue that while every semiconductor cycle eventually turns, near-term signals still point to sustained spending. Rory highlights TSMC’s capex posture as a grounded indicator because fabs are hard to unwind, while Jason notes that unless a slowdown is visible soon, most operators can’t afford to plan as if it will.
OpenEvidence’s $12B round: perfect product, but TAM expansion is the real question
OpenEvidence is framed as a near-ideal AI business: clinical decision support with credible data partnerships, compliance, and a clear monetization path via pharma advertising. Rory praises product fit and founder quality, but challenges simplistic TAM math, noting DTC advertising and pharma rep spend dynamics.
Who gets stuck holding the ‘Brex round’ in AI mega-fundraising?
They connect OpenEvidence to the earlier hubris discussion: great companies can still produce late-round 1x outcomes if the tide turns. The hard part is that in the moment, the deal always feels ‘obviously generational,’ making discipline difficult—especially when growth and momentum are exceptional.
a16z’s claim: 2/3 of private AI revenue comes from its portfolio—signal vs soundbite
Rory argues the statistic is directionally true but mechanically driven by a few giants (notably OpenAI and Databricks), making it less surprising than it sounds. Jason uses it to ask a bigger question: is venture becoming a scalable ‘asset class’ if repeatable platforms capture outsized share—while Rory warns excess capital can still ruin returns.
IPO market signals: EquipmentShare’s clean pop vs Wealthfront’s subscale struggle
EquipmentShare is presented as the archetype of an ‘effortless IPO’: large revenue scale, high growth, and profitability. Wealthfront is the counterexample—good product and mission, but small market cap creates liquidity/coverage challenges and makes the post-IPO talent/recruiting narrative harder.
Subscale IPOs and ‘clearing the logjam’: Ethos and the 2021 unicorn backlog
They discuss whether smaller IPOs are ‘capitulation’ or a healthy market-clearing mechanism for overvalued private marks. Rory argues price clears markets and that going public at a lower valuation can still be a strong outcome—especially compared with staying frozen at an unrealistic 2021 mark.
Salesforce’s $5B+ Army contract: SaaS isn’t dead, but growth faces multiple daggers
They close on Salesforce winning a massive defense contract as a rebuttal to ‘AI replaces systems of record’ narratives. Still, Jason and Rory stress that SaaS growth is pressured by seat contraction, price fatigue, and budget shifts toward AI, so incumbents must adapt without assuming agents alone restore the old NRR era.
