The Twenty Minute VCWho Wins in the Next Decade of VC | Margin and Price Matter Less in a World of AI | Sameer Dholakia
At a glance
WHAT IT’S REALLY ABOUT
AI reshapes venture: capital moats, frontier winners, open-model upside
- Sameer Dholakia frames AI as the largest “supercycle” of his career, with intelligence becoming a market bigger than oil because it amplifies every human and organization.
- Bessemer’s Anthropic investment was underwritten as a capital-intensive, hyperscaler-like winner-take-most arena where partnerships (notably AWS) meaningfully de-risk compute and distribution.
- He rejects a binary open-vs-closed narrative, arguing both frontier labs and open-model infrastructure will be huge, while sovereignty and provenance will meaningfully influence adoption.
- Dholakia explains modern growth underwriting as looking forward (NTM revenue, revenue quality, margin trajectory) rather than anchoring on today’s ARR multiples, especially for hypergrowth AI businesses.
- He argues venture is splitting into mega-platforms and extreme specialists, with capital intensity pushing up fund sizes, check sizes, and the need to concentrate into generational winners.
IDEAS WORTH REMEMBERING
5 ideasAI is a bigger supercycle than cloud or the internet because intelligence is a universal input.
Dholakia argues that AI is a “magic” moment like the internet and virtualization—except larger—because intelligence touches every user segment (consumer to government) and becomes a foundational input to nearly all work.
Frontier labs resemble cloud hyperscalers: capital intensity drives consolidation into a few platforms.
Bessemer approached Anthropic like a hyperscaler bet: training frontier models is compute- and capital-intensive, likely concentrating into a handful of winners; the AWS relationship was pivotal for both compute and early distribution.
Open models won’t “kill” frontier labs; both can be massive in a market this large.
He frames “open vs closed” as an ‘and’ decision: the TAM is so large that arguing exact token/dollar splits is less important than holding exposure across the ecosystem (e.g., Anthropic + Fireworks).
Model provenance and sovereignty will shape enterprise/government buying, not just model quality.
He expects governance/provenance concerns to limit direct adoption of Chinese-origin models by some CIOs, creating space for US/EU “sovereign” or trusted open models (e.g., Mistral; Sarvam for Indic-language needs).
In AI growth investing, underwriting shifts from current ARR multiples to near-term (NTM) scaling and long-duration compounding.
He still believes entry price matters, but underwriting should center on forward revenue power (NTM) and compounding duration; he’s willing to pay eye-watering current ARR multiples if NTM multiples normalize quickly under hypergrowth.
WORDS WORTH SAVING
5 quotesIntelligence is gonna be bigger as a market than oil because oil powers all the machines. Intelligence powers all the humans. Not some of the humans, all the humans.
— Sameer Dholakia
I refer to it as, like, magic, right? Like, I've been in software for 31 years now. And the handful of other moments where I have felt magic, a trillion-dollar supercycle followed every single time.
— Sameer Dholakia
According to Byron, he said it was the lowest-voted deal in the history of the firm, uh, the, of, of the time he had been there.
— Sameer Dholakia
By the way, I, I absolutely believe capital is a moat.
— Sameer Dholakia
Teammates work harder for each other. They're not working for themselves. They're working for their teammates, and they're working for their team, and you never wanna effing let down your teammates.
— Sameer Dholakia
High quality AI-generated summary created from speaker-labeled transcript.
