20VC: You Need $1BN To Do Seed Today: The Untold Truths of Venture | Who Wins and Who Loses in the Next Decade of VC | Why Margin and Price Matter Less in a World of AI with Sameer Dholakia, Bessemer

1. Key Themes

Intelligence as a Market Bigger Than Oil

Sameer anchors his entire investment posture on a TAM argument: if intelligence is demanded by every human and institution, then precision about market mix is irrelevant. He frames this as the third "magic" moment of his 31-year career, each of which was followed by a trillion-dollar supercycle (internet, virtualization/cloud, now AI).

"Intelligence is going to 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: consumers, prosumers, SMB, mid-market, enterprise, government." 00:00:00

"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, and it doesn't happen very often." 00:00:00

Underwriting Forward Revenue: Look Through the Windshield, Not the Rearview Mirror

Sameer's pricing framework is NTM-based rather than current-ARR-based. He cites investments that "would make most venture capitalists throw up" on trailing multiples: Abridge at $5M ARR / ~$700-800M valuation, Anthropic at ~$100M ARR in the $15B round (~150x), and Legora at $20M ARR in a $1.8B round. His rule of thumb is 10-15x NTM for hypergrowth assets, with a ceiling on how far forward he'll underwrite.

"If your view... was that, yeah, but on an NTM basis, and that's what I care about. I'm looking at, what do I think in the next 12 months this thing can be, and do I feel like this is a reasonable multiple in the next 12 months?" 00:16:53

"Something that's in the 10 to 15 times NTM, I don't think is crazy. Especially if there are these businesses that are growing at rates that have never been seen in the history of business." 00:19:01

Capital Is a Moat and Kingmaking Is Real

Sameer agrees with Harry that capital is a moat in this cycle: scarce compute makes the biggest war chest a necessary (but not sufficient) condition to win, and capital also creates brand, which attracts talent and customers in a reinforcing flywheel.

"I absolutely believe capital is a moat. I have always said that markets pick winners, and then it becomes a self-fulfilling prophecy." 00:00:00

"It is a necessary condition for you to win the market, that you have the biggest war chest... because the compute is so scarce, and therefore the compute is so expensive, that if you don't have that, you just can't play." 00:25:08

The Barbell: Giant Platform or Extreme Specialist, Nothing in Between

Looking at venture "like any other industry," Sameer argues that professionalization and scale always produce consolidation plus specialists, with the middle squeezed. This applies directly to billion-dollar seed funds and to retail LP access.

"In most industries, when they professionalize and they get to scale, you either have to be on this end of the spectrum, which is a giant global platform... or you need to be an extreme specialist... If you're in between, good luck." 00:29:36

"Not being a giant global platform or an incredible specialist who can add differentiated value... I think it's a hard world to make money." 00:00:00

Cloud Redux Mental Model for AI Market Structure

Sameer's original Anthropic thesis: capital intensity consolidates the infrastructure layer to a handful of players (like AWS/Azure/GCP), while hundreds of purpose-built apps sit on top. Anthropic's AWS relationship supplied compute and distribution at an early stage.

"I think this market is going to be a cloud redux... it was massively compute intensive and therefore capital intensive. And so it was going to consolidate to a handful of players. But just like in the cloud era... there would be hundreds of apps built on top of them." 00:07:59

Open Tokens and Closed Tokens: "Genius of the And"

Sameer updated his view in the last 12 months on open-model economics, and Bessemer now holds both Anthropic (closed) and Fireworks and fal (open/infra). He believes the closed/open mix is irrelevant at this TAM, and that a US-based open-weight leader would be a real business.

"In the last 12, I would say just my deep appreciation for how big the open token economy is going to be. I just didn't see it 12 months ago." 01:08:00

"You can build a very large business and get venture returns if you are the leading provider of a open-source solution... most enterprises don't want to be responsible for the open source, they're going to want some provider with the right talent behind it." 00:13:33

Vertical AI Across Professional Services

Bessemer has built a thesis around "AI for [profession]": doctors (Abridge), lawyers (EvenUp, Legora), auditors (Fieldguide). Sameer sizes the legal TAM at roughly a trillion dollars of global legal services spend, so winners need minimal share to build giant businesses, and Jevons paradox expands demand from unserved legal needs.

"There's about a trillion dollars of legal services spend in the world... they don't have to take any meaningful share of that to go build a massive business." 00:22:00

"We've done AI for doctors with Abridge and AI for lawyers with EvenUp and Legora and... Fieldguide is AI for auditors." 00:38:20

Sovereign Models and "Own Your Intelligence"

Bessemer has invested in Sarvam (India) and is investing in Mistral. The Fireworks pitch for enterprises is that they will want to own, not rent, intelligence, with sovereign models serving government and regulated use cases.

"Enterprises are not going to want to rent intelligence from somebody else, they're going to want to own their own. I just think it's like the simplest and really compelling tagline for any enterprise CIO buyer." 00:59:23

"One of the things that makes India such a ripe opportunity for intelligence local to their market is the Indic languages... And that's multimodal. It's voice, it's text." 01:00:17

Culture as Competitive Advantage: Teammates Beat Colleagues

Sameer's SendGrid playbook: hard-coded values (happy, hungry, humble, honest), pronoun-counting interviews, and whole-company off-sites. The payoff is behavior under stress: teammates solve problems rather than point fingers or leave.

"Teammates and friends will beat colleagues and coworkers every day and twice on Sunday." 00:45:49

Founder Secondaries, DPI, and IPO Dynamics

Founders are rational about taking chips off the table; Sameer encourages it. He predicts permanent structural staying-private-longer, followed by an IPO unlock later next year and 2028 driven by public-market demand for pure-play AI exposure.

"I believe that it is a permanent structural shift that companies will stay private longer." 00:52:03

"There will be an obvious appetite from investors in the public markets to get pure-play AI exposure, and they don't want it from just a few companies. They're going to want it in dozens or hundreds of companies." 00:53:00

2. Contrarian Perspectives

Capital Is a Moat, and Billion-Dollar Seed Funds Are Rational

Conventional wisdom says capital isn't a moat and seed funds should be small and nimble. Sameer says the opposite: seed pricing is "factors higher" than the $3-5M checks on $20-40M valuations of his early career, so you need many shots on goal. Bessemer's own $1.75B seed-and-early fund puts this into practice.

"Seeing the prices that I'm seeing... if you just think about the number of shots on goal you need to have to find the winner, I don't know how you would have a small seed fund and be successful. I don't." 00:28:24

Margins Matter Less Than Share and Position

In a world of rapid scale, gross margin at investment is a poor guide. Sameer points to businesses whose margins improved by "factors" within 12-24 months, though he warns you must be able to defend the path to 60-70%.

"I can remember most of the businesses I invested in, gross margins at the time we invested, in this AI world, were meaningfully lower than they are at scale today. I mean, meaningfully lower, like factors improvement inside of 12 to 24 months." 00:20:28

Waymo Is Not a Ride-Hailing Company, So Uber's Market Cap Is the Wrong Comp

When Harry challenged him that Waymo's valuation (~$150B) exceeds Uber's (~$120-125B), Sameer argued the ride-hailing market is the wrong frame: Waymo can license its driver into personal car ownership, trucking, and logistics.

"Waymo is going to be an extraordinary, I think a trillion-dollar business, easy." 01:10:56

"If I could have bought that car with an option to have a Waymo driver on it and pay them a subscription every single month on that car... it would be a default purchase." 01:10:04

Don't Bother Predicting the Open vs. Closed Token Mix

While most guests argue about 99% of tokens vs. 99% of dollars, Sameer calls the exercise irrelevant at this TAM and says his own bull case on the frontier missed by "factors."

"When you're in a market that big, predicting with any precision the mix between closed tokens and open tokens is irrelevant." 00:11:28

"My bullish bull case was that we could get to some massive valuation, and I'll just say that I missed the mark on how big of a business could be created at the frontier by factors." 00:11:28

Perverse Entry Price Logic: Pay 150x ARR Because You're Underwriting a Different Growth Regime

The Anthropic round was "the lowest-voted deal in the history of the firm" per Byron Deeter, at ~150x current ARR, and Sameer's regret is not writing a bigger check.

"I just wish we had written a bigger check, that's all." 00:10:33

"According to Byron, he said it was the lowest-voted deal in the history of the firm of the time he had been there." 00:10:10

3. Companies Identified

Anthropic

Frontier AI lab (Claude). Bessemer invested in the Series D (~$15B pre / $18B post) alongside Menlo and Lightspeed. Sameer says the firm's prior growth fund (Century II) put about 20% into the name. He cites an AWS-based compute and distribution relationship as pivotal and regrets not writing a bigger check.

"Our last growth fund, our Century II, I think we had 20% into one name, the name was Anthropic. Feel pretty good about it." 00:35:59

Fireworks AI

Open-model inference platform; led by Lin Qiao with George Hu as president. Sameer and Harry both invested at ~$15B. Pitch: enterprises will want to own their intelligence.

"Every business is going to want to own some of their intelligence, for sure." 00:59:23

fal

Gen-media and infrastructure platform (open tokens). Bessemer is an investor, cited as one reason Sameer changed his mind on open tokens.

"Fal on the gen media side and the infrastructure side there and they're going up and down the stack. They're an amazing, amazing business, amazing team." 01:08:00

Legora

Legal AI company led by Max; Bessemer led a $1.8B round at ~$20M ARR; Sameer says the company went from 20 to 200 (ARR) since.

"The company just went, you know, 20 to 200... They're on fire, and I feel pretty good about that 1.8 billion." 00:19:01

Harvey

Legal AI competitor to Legora; Harry says both will do well. Cited by Sameer as one of two names big-law buyers recognize as the major legal AI companies.

"If you ask anybody at a big law firm, who are the major legal AI companies, they would name those two." 00:24:02

Abridge

AI for doctors (clinical documentation). Bessemer invested at ~$5M ARR and a $700-800M valuation.

"When we invested in Abridge, they were at 5 million of ARR, and we were paying whatever it was, 700, 800 million dollar valuation." 00:16:53

EvenUp

AI for lawyers (personal injury); cited as part of Bessemer's vertical-AI portfolio.

"We've done AI for doctors with Abridge and AI for lawyers with EvenUp." 00:38:20

Fieldguide

AI for auditors; invested early-growth at sub-$10M ARR and a sub-$200M entry price. Sameer regrets not backing up the truck; says capability jumps in early 2026 lifted adoption.

"They're just ripping. They're growing really fast... Their inflection has been extraordinary." 00:38:47

EliseAI

AI for housing and property operations; Bessemer announced a ~$200M investment.

"We just announced a round with EliseAI, where... we were going to be investing nearly $200 million." 00:33:59

Waymo

Autonomous driving; Bessemer invested nearly $200M. Sameer thinks it becomes a trillion-dollar business across ride-hailing, personal vehicles, trucking, and logistics.

"We invested nearly $200 million into Waymo, and I just think that... autonomy and what that changes and how that changes the way we work and live, I think is awesome." 01:08:33

Mistral

European sovereign model lab; Bessemer is investing.

"We are invested in or investing in Mistral right now." 00:59:54

Sarvam

Indian sovereign AI lab (Indic languages, multimodal); Bessemer is an investor.

"We're invested in Sarvam in India." 00:59:54

SendGrid

Sameer's prior company (email API); culture case study; IPO'd. Sold to Twilio.

"We had our 4Hs: happy, hungry, humble, and honest." 00:42:51

Twilio

Developer communications API; Bessemer led seed/A/B; Sameer was in the Twilio exec team post-SendGrid acquisition.

"We led the seed and... A and B in Twilio, led the Series B in SendGrid." 01:07:01

Jasper

AI marketing app on GPT-3; the demo from Dave Rogenmoser in winter 2021 gave Sameer his "magic" moment.

"He shows me a demo of Jasper, that's this AI marketing app sitting on top of GPT-3... I was like, magic." 00:05:00

VMware

Virtualization pioneer; Sameer's second "magic" moment, which enabled the cloud supercycle.

"VMware created this layer of software that disintermediated the hardware from the OS." 00:05:00

OpenAI

Sameer and Harry both passed at ~$30B and regretted it; its growth from ~$40M to well north of $1B in 2023 stretched Sameer's imagination for Anthropic.

"OpenAI had gone from like 40 million to... something well north of a billion, billion and a half, maybe approximating two." 00:15:38

Stripe

Bessemer's biggest rue: "the developer API venture capital firm" that missed Stripe. Also cited by Harry for private-company optionality (an $8B OpenRoute acquisition).

"We were the developer API company and we weren't in Stripe. That hurts." 01:06:39

Rocket Lab

Space company; David Cowan found it when "everybody thinks is crazy."

"He finds stuff that everybody thinks is crazy like Rocket Lab that then becomes... an enormous business." 01:05:58

Auth0

Developer identity API company led by David Cowan at Bessemer.

"We had Auth0 that David Cowan led." 01:06:39

Thinking Machines

Lab valued at ~$40B per Harry; used as a comparator for a US open-model business.

"If you look at the Thinking Machines that are $40 billion price today, is there a $500 billion open model US to be built?" 00:13:19

Vercel

Mentioned for publishing data on open-model usage shift.

"A lot of data that's being produced by Vercel and Guillermo, who publishes it." 00:11:03

Base10 and Modal

Inference competitors to Fireworks, which Harry thinks will do well.

"I think Base10 and Modal will do well, too. I think Fireworks will win." 00:23:01

Cognition and Factory

Coding-agent companies Harry flags as possible generational companies.

"Cognition or Factory could be." 00:39:10

Canva

Cited by Harry as a company whose public-style disclosures reset expectations.

"Canva release their numbers... you have similar kind of reactions." 00:54:20

Adyen

Public payments company Harry says is structurally disadvantaged versus private Stripe.

"Stripe is able to do a lot of things because of their private nature that Adyen is not able to." 00:54:20

Oura

Consumer brand where Harry argues IPO makes sense.

"For things like Oura, I think it also totally makes sense, a consumer brand, one where retail wants to be part of it." 00:55:52

Tesla

Referenced as the FSD subscription analog for Waymo's OEM opportunity.

"Just like Tesla, and Elon's got it for his FSD on his cars." 01:10:04

Machina and MicroAGI

Robotics data companies Harry names; Sameer says Bessemer isn't in them that he knows of.

"We saw Machina just get done by Sequoia yesterday. You've got MicroAGI in Europe." 01:09:16

Red Hat

Open-source business model precedent.

"Whether you're going from Red Hat to, you know, fill in the blank in the database market." 00:13:33

AWS

Supplied Anthropic early compute and distribution; internal Nova model "not particularly distinctive."

"AWS... had an internal model, I think it was called Nova, that was not particularly distinctive, I'll be generous." 00:07:59

Blitzy, Flax, Vanta (sponsors)

Autonomous enterprise code platform, a financial platform for business owners, and a compliance platform. Harry names Ramp, Harvey, and Writer as Vanta customers.

"Fortune 500 enterprises are shipping enterprise-grade code five times faster." 00:01:19

Menlo, Lightspeed, Accel, USV, Sequoia, Benchmark

Funds referenced: Menlo and Lightspeed in the Anthropic round; USV's $900M fund; Accel's growth practice (Harry's "Accel deals" such as 1Password and Qualtrics, and the "Leaders Fund").

"We just saw, literally earlier today, USV announced their $900 million funds." 00:28:00

4. People Identified

Byron Deeter (Bessemer)

Partner; co-led Anthropic with Sameer, "the picker" and the best closer.

"Byron Deeter every time, for sure, 100%. That guy is a force of nature." 01:05:58

Lin Qiao (Fireworks CEO)

Harry calls her generational; Sameer notes she leads a technical team from Meta.

"You have the brilliance of Lin, strategic thinker, deep technologist, unbelievable technical team from Meta." 00:58:38

George Hu (Fireworks President)

Former Twilio executive; Sameer calls him "one of the very best software executives in the industry."

"He knows how to build a go-to-market engine, Harry, like very few people in technology in the market today know how to build." 00:58:38

David Cowan (Bessemer)

Three-decade partner and contrarian picker (Rocket Lab, Auth0).

"He finds stuff that everybody thinks is crazy like Rocket Lab." 01:05:58

Max (Legora CEO)

Global ambition, a company on fire.

"He has global ambitions." 00:22:00

Jeff Lawson (Twilio)

Extraordinary leader with whom Sameer worked at Twilio.

"In the same exec leadership team room with Jeff Lawson, who's extraordinary." 00:58:15

Dave Rogenmoser (Jasper)

Showed the GPT-3 demo that gave Sameer his AI "magic" moment.

"Dave Rogenmoser over at Jasper said, 'Hey, can you spend some time with me...'" 00:05:00

Matt Murphy (Menlo)

Led the Anthropic round and called Byron.

"Matt called up his buddy Byron and said, 'Hey, we're looking to build the syndicate here for this startup named Anthropic.'" 00:07:59

Venky Ganesan (Menlo)

Mutual friend; gets stick for calling early stage "option bets" but is "intellectually 100% correct."

"What he meant intellectually is 100% correct. It is an option." 00:32:50

Ravi Mhatre (Lightspeed)

Sameer's dream add; wrote a billion-dollar check into Anthropic's $60B round.

"Ravi is incredible, deep thinker... not just really good, but he's willing to take these giant risks." 01:07:10

Dario Amodei (Anthropic)

Sameer references "Machines of Loving Grace" on compressing a half-century of biology into a decade.

"I really do believe what Dario says on that in Machines of Loving Grace." 01:11:43

Adam Fisher and Jeremy (Bessemer)

Named among Bessemer's top pickers (Adam in Israel).

"Byron, Jeremy, Adam Fisher in Israel and David Cowan out here." 01:04:25

Talia

Bessemer team member who told Harry to ask about the teammates line.

"Talia told me to ask you about teammates and friends beat colleagues and coworkers." 00:40:40

Guillermo Rauch (Vercel)

Publishes the open-model data Harry cites.

"Vercel and Guillermo, who publishes it." 00:11:03

Eddie Izzard

Comedian ("cake or death") invoked by Harry.

"You ever seen Eddie Izzard? He goes, 'Cake or death?'" 00:27:34

Eminem

His lyric ("Success is my only motherf***ing option") is used as a hiring-culture reference.

"We'd quote the Eminem song." 00:47:19

5. Operating Insights

Pronoun-Count Interviews Hire "We" People

Sameer's SendGrid interviewing hack: ask for a proudest accomplishment and count "I" vs. "we." It operationalizes the "humble" value into a concrete test.

"We would sit back and do a pronoun count. How often does Harry say we? How often does Harry say I? I want the people that say we." 00:43:42

Define Each Value in Writing, Because Words Are Ambiguous

Sameer had to document what each "H" meant ("happy" meant optimistic and problem-attacking, not just joyful), so values screen out real behaviors like negativity.

"I did find at SendGrid, I had to write down, what do we mean by each of these 4Hs?" 00:48:10

Whole-Company Offsites Pay for Themselves in Crisis

Flying the entire company to Mexico each January, with a second day of cross-office social activities, builds the trust that prevents finger-pointing.

"Teammates and friends in hard moments... don't start pointing fingers. They come together." 00:45:49

Enforce Culture Against High Performers

"You can't have the asshole that's a great performer that doesn't abide by the rules," and everyone must feel empowered to call out deviations.

"Everybody has to be empowered on the team to be like, 'Nope, I'm going to call that out if you are deviating from our culture and values.'" 00:42:19

Use Secondaries to Make Founders Swing Harder

Encourage founders to take some liquidity at late stages so they think long-term and take bigger swings.

"I tell them to. I'm like, 'Shoot. You should probably sell a little bit here, bud, to make sure that... you're financially secure for your family.'" 00:51:19

6. Overlooked Insights

Early-Stage Vertical AI Inflected After a January-February Model Leap

Sameer says Fieldguide's growth accelerated because "post this January, February magical time when the models took a leap forward... their solution became wildly more compelling." The implication: model capability steps unlock vertical apps with a lag, so companies that look like slow burners can flip to hypergrowth after a model release, which is the right moment to back up the truck.

"Post this January, February magical time when the models took a leap forward, all of a sudden, their solution became wildly more compelling and the uptake is showing it." 00:38:47

Public-Market Scarcity Premium for Pure-Play AI Exposure

Offhand, Sameer argues that LPs are moving to private because most of the run-up is already captured before IPO, and that public-market demand for "dozens or hundreds" of pure-play AI names will lift IPO pricing in 2027-2028. This frames today's late-stage entries as buying ahead of a coming public-market supply-demand imbalance, not just private momentum.

"They can't reap the rewards of this generational boom if they're waiting for them in the public markets. Most of that run-up will have already happened in the private markets." 00:52:03