Sep 11, 2025 · 1h 34m · 20vc
OpenAI’s $10BN Secondary Sale, Ramp Hits $1BN ARR & Brex Hits $700M · 20VC with Harry Stebbings
Rory O'Driscoll · 35m spoken Jeff Lawson · 25m spoken Jason Lemkin · 14m spoken Harry Stebbings · 8m spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20VC, host Harry Stebbings leads a panel of tech veterans—Jeff Lawson, Jason Lemkin, and Rory O'Driscoll—to dissect current industry trends, including massive AI valuations, the transition from traditional SaaS seat-pricing to automated AI models, and the changing dynamics of corporate governance and VC diligence.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 10.1% of the talking time here. How this is scored →
Harry as informed peer 3.7 Guest teaching 3.9 Guest disagreement 2.2 Harry pushing back 2.7
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:26:46 Jason demands criminal prosecution for startup fraud
Jason forcefully argues that founders committing financial misrepresentations should go to prison to stop rampant ecosystem greed and dishonesty.
Hardest push from Harry ▶ 53:11 Harry challenges national sovereignty argumentsHarry directly challenges Rory's premise that national sovereignty can successfully power commercial technology companies without free market dynamics.
Biggest teaching moment ▶ 1:12:37 Jeff explains the structural limits of Twilio's APIJeff educates the panel on why a simple three-field API restricts continuous value-add compared to broader infrastructure control planes like Cloudflare.
Harry holds his own ▶ 28:53 Harry calculates Greenoaks' fund concentration riskHarry demonstrates keen venture capital expertise by calculating the exact percentage check size Sierra represents relative to Greenoaks' overall fund.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome and Guest Introductions: Jeff Lawson Joins the Panel | 3 | 2 | 1 | 1 | Harry introduces guest Jeff Lawson and kicks off the discussion on Elon Musk's pay package. Rory breaks down the proxy statement metrics while Jeff explains Twilio's simple compensation philosophy. | |
| Boardroom Leverage: Founder Control and the Threat of Walking | 2 | 4 | 2 | 1 | The panel analyzes founder control in boardrooms, with Jeff highlighting the downside risk that Tesla's valuation depends heavily on Elon Musk's cult of personality. | |
| Missionary vs. Mercenary Founders & AI Acqui-hires | 5 | 4 | 3 | 5 | Harry pushes back on Jeff's moral critique of mercenary founders by highlighting how LP endowments and universities benefit financially from acqui-hires like Scale AI. Rory candidly describes acqui-hire corporate pretenses. | |
| The Fintech Boom: Valuing Ramp and Brex | 4 | 3 | 2 | 3 | Harry asks whether fintech unicorns Ramp and Brex should be valued like traditional financial institutions or high-growth tech platforms. Rory points out their financial service margins versus software growth rates. | |
| The AI Trickle-Down Effect and Infrastructure Risks | 2 | 3 | 1 | 1 | Jason uses a deep-sea ocean metaphor for AI capital filtering down the tech stack. Jeff compares current AI infrastructure spend to past revenue churn seen at Twilio during the mobile boom. | |
| Sierra's $10B Valuation and the Reality of Capital Allocation | 6 | 3 | 2 | 4 | Harry analyzes Greenoaks' investment in Sierra, calculating fund concentration risk and check sizing relative to fund size. Rory outlines category-winner logic justifying steep ARR multiples. | |
| Late-Stage AI Investing and the Shift to "Venture Cap" | 5 | 4 | 3 | 4 | Harry questions whether every major VC fund requires a foundation model logo investment. Rory defends late-stage growth investing while Jeff jokes about spending LPs' money to stay relevant. | |
| OpenAI's $10B Secondary and the Silicon Valley Talent Battle | 3 | 3 | 2 | 2 | The group explores OpenAI's $10B secondary liquidity event, with Jason explaining how massive AI compensation packages hamper traditional B2B engineering recruitment. | |
| Anthropic's Copyright Fines and the Future of AI Training Data | 3 | 4 | 4 | 2 | Rory outlines the legal logic behind Anthropic's copyright fines, while Jason aggressively criticizes Anthropic for sourcing training data from pirate websites. | |
| ASML's Shock Investment in Mistral and Corporate Cash Strategies | 3 | 4 | 2 | 3 | Harry questions why semiconductor giant ASML invested in AI lab Mistral. Jason breaks down corporate balance sheet cash strategies and EPS impairment considerations. | |
| Corporate M&A Dynamics and the Rise of Sovereign AI Champions | 5 | 5 | 3 | 6 | Harry challenges the idea of sovereignty driving tech success, asking for historical precedents. Rory responds by citing defense sector realities and national champions. | |
| Atlassian's $610M Arc Browser Acquisition and Twilio's M&A Playbook | 4 | 5 | 2 | 3 | Harry highlights Atlassian's $610M acquisition of The Browser Company. Jeff expresses skepticism regarding workplace browsers and warns that AI will decimate SaaS seat bases. | |
| The SaaS Innovator's Dilemma: Seat Cannibalization vs. AI Automation | 4 | 4 | 2 | 2 | Jeff explains how Twilio avoided the innovator's dilemma as an infrastructure provider, contrasting it with seat-based SaaS vendors attempting to capture labor budgets. | |
| AI Strategy and Public Market Survival for Dropbox and Box | 4 | 3 | 1 | 2 | Harry prompts Jeff to assess public tech founders navigating AI transitions. Jeff praises Aaron Levie and Drew Houston as resilient public market survivors. | |
| Twilio's Product Constraint and the API "Jail" | 4 | 6 | 1 | 3 | Harry asks Jeff what structural constraints he faced at Twilio. Jeff details the limits of a three-parameter messaging API compared to broader dashboard proxy surfaces like Cloudflare. | |
| Jeff Lawson's Unified Theory of Developer APIs | 3 | 7 | 1 | 1 | Jeff delivers his Unified Theory of Developer APIs, categorizing successful developer companies into BizDev-as-a-service, CapEx-as-a-service, and Algorithm-as-a-service. | |
| Quick-Fire Round: Figma Valuations and Canva IPO Timelines | 3 | 2 | 3 | 3 | The panel conducts a quick-fire round on Figma stock predictions and Canva's IPO timing, with Jason strongly rejecting Harry's prediction of a Q4 Canva listing. | |
| The Rise of Venture Capital Fraud (IRL Case Study) | 4 | 4 | 5 | 3 | Harry brings up the arrest of IRL's CEO. Jason argues forcefully that imprisonment is necessary to deter founder fraud, while Jeff notes VC negligence enables bad actors. |
Statements from this episode (44)
Lawson: Buyers Want AI Replacing 75% of Staff, Not Co-Pilots
“Oh, we're going to make your human beings 10% more efficient because of the AI co-pilot. When reality is they're gonna want a product that is like, no, no, no, I don't need 75% of these people anymore. Give me that product.”
Lemkin: VCs Are No Longer Attempting Due Diligence
“Venture rounds are all getting done on Saturdays. Forget about no diligence being done two years ago. Now diligence isn't even being attempted.”
O'Driscoll: Executive compensation structure reveals a board's true priorities
“Compensation is how boards reveal their real priorities. Nothing else matters as much.”
Lawson: Complex compensation packages increase feelings of unfairness
“One of my principles of compensation was always the more levers and knobs and things you put into a comp package, the more opportunity it is for someone to just think it's unfair.”
Lemkin: Founders Control the Board in Almost All Portfolio Companies
“Almost all my portfolio companies, the founders control the board, not just from a cap table perspective, but from a relationship perspective, they control it.”
O'Driscoll: Musk's Tesla Deal Will Be a High-Water Mark for CEO Pay
“Elon's case. I don't think that's the norm. I think that that will be the high water mark Not the norm, but I definitely think like all high watermarks, it will push up other people's demands and aspirations.”
O'Driscoll: 75% of Tesla's Valuation Is Elon Musk's Brand Premium
“If you are valued as a car company, you'd be a valued at around 25% of where you are now. Maybe this is a car company where 25% of our current market cap, and then Elon's special sauce, which is the other 75.”
Lawson: Financially rational founders should work at hyperscalers, not start companies
“If you just want to make money, probability adjusted, you should just go get a job at a hyperscaler. You will probability, probability adjusted, you'll make more money. So you don't start companies to make money. You start companies because you love what you'r…”
O'Driscoll: AI reverse acqui-hires are regulatory shams leaving dead target companies
“We actually, in both of those cases, that's a pure pretense to get the government off of us, right? Everyone knows both of these companies are toast because in the case of Scale AI, in theory, Meta owns 50% of it, and in theory, their business is selling to ev…”
O'Driscoll: Fintechs combine financial margins with software growth rates
“They have the margin profile and core dynamics of a financial services company, but they have the growth rate of a software company.”
O'Driscoll: Ramp and Brex will be valued like Amex once growth slows
“Once the growth rate slows, they will be valued. Just like if they're growing the same as Amex, they will be valued the same as Amex. The growth is what's saving them.”
Lemkin: B2B companies failing to gain AI revenue deserve an F
“If you're a B to B company and you're seeing nothing, you're not seeing any boost from AI. You didn't get open AI or anthropic as a customer. You're not seeing any benefits like you get an F. You get an F. There's so much money flowing through this system, rig…”
O'Driscoll: Every venture investment comes down to three sequential questions
“My mental model is always, we say this internally, is there a category that can support a big winner? Are these guys going to be one of the winners in the category? And are you getting paid for the risk? You could argue that's the kind of sequence of questions…”
O'Driscoll: Customer support is the number one application-level AI use case
“Other than coding, which is the InfraPlay, at the app level, I think customer support, customer success is the number one use case for AI, because it's just so obvious. You have lots of people answering phone calls, answering emails. You can do it all with AI.…”
Lawson: Founders commit 100% of capital, making VC portfolio concentration easy
“When I start a company, this is my, I mean, until now, all my prior ventures, that was a hundred percent of my capital allocation for myself, for my life, For my time, for my bank account, for everything. And so the whole idea that an investor would you know, …”
Lemkin: Kleiner Perkins' $100M Anthropic check is just a logo deal
“You got, you can't walk into the partners and not have anthropic or open AI on the website. It's not enough. Yeah. It's a low, I think it's a logo deal.”
O'Driscoll: 80% of venture capital dollars now go to late-stage deals
“Most of the money in quote unquote venture capital is with venture cap. It venture capital is 20% old school venture capital and 80% plus or minus late stage. What would have been fidelity growth, public investing. Right. This is where most of the dollars are …”
O'Driscoll: Investors will overextrapolate growth and overpay for late-stage AI deals
“So in the end, what will happen is people will overextrapolate and a bunch of these will be overpriced. And then people will go, oh yeah, that's why you don't overpay.”
Lawson: Domino's Pizza Outperformed Google Stock Over 10 Years
“Better return than Google over that time period.”
Lawson: No Legacy Non-Tech Company Succeeded Opening Silicon Valley Offices
“Well, and I struggle to think of a single, like, kind of, let's say, legacy company who said, hey, we got to get in on the software thing, and opened up their Silicon Valley office and actually succeeded in that.”
O'Driscoll: Data Labeling Firms Will Sell Legally Compliant Book Datasets for AI
“I assume that there will be a much more efficient way than that of doing it. You're exactly right. I'm sure, for example, that there will be a corpus available of a purchased copy of every, you know, I'm sure that Scale some one of these label guys that says w…”
Lemkin: Anthropic Sourced AI Book Training Datasets Directly from Pirate Sites
“They downloaded this from pirate websites. Okay. This wasn't cutting a little bit of a corner. Okay. This wasn't claiming something that wasn't quite open source was, This is literally, guys, we got to get this rocket ship going. I need a trillion books. I'm g…”
Lemkin: OpenAI Stole YouTube Videos to Train AI Models
“This isn't even stealing YouTube videos like OpenAI did.”
Lemkin: Corporate venture investments allow companies to deploy cash without reducing EPS
“When big companies with a lot of cash make corporate investments, it's weird because if you're generating, at least in the US, if you're generating massive amounts of cash, it's orphaned on your balance sheet. You can't just go hire a thousand engineers. It hu…”
Lemkin: CVC units prioritize avoiding equity write-downs over maximizing overall returns
“Our job is to make money at Salesforce Ventures, but it's more important we don't lose money. Because if we lose money, we have, we may have to take an EPS hit or an impairment charge. But as long as our investment doesn't go down, it's pretty much okay.”
Lawson: Corporate venture investing gives Salesforce direct core business advantages
“In that, in the Salesforce world, the answer I would say definitely is yes. Obviously it cemented the role in the center of an ecosystem. They ended up making acquisitions. They have more information to make their product decisions on, like all sorts of benefi…”
O'Driscoll: Semiconductor equipment is a levered version of semiconductor cyclicality
“The most Cyclical business on the planet almost is a semiconductor capital equipment cycle because it's kind of a levered version of the semiconductor cycle.”
O'Driscoll: European AI investment is driven by sovereignty, not market efficiency
“The Europeans make a whole load of defense equipment that they have no business making from an economies of scale perspective. They simply do it because they're like, they don't want to rely on the Americans. And this is the AI version of that.”
Lawson: Atlassian's work-browser thesis for Arc acquisition does not make sense
“You know, Mike's always been a real forward thinker but I would say the thesis didn't really resonate with me in terms of we need a different browser for work. I could imagine some upsides was, is enough upsides to actually change behaviors. I don't think so, …”
Lawson: AI will decimate SaaS seat bases and human workflow roles
“The thing I would say though about any company in SAS today, and Atlassian is a Prime example of this is they are primed for, ah, disruption right now because AI is going to decimate their seat base for their products. It'll decimate the roles people are playi…”
Lawson: In corporate M&A, missing a deal is worse than a failed deal
“Well, I'll tell you the other thing that, that I think is conventional wisdom, especially at companies that do a lot of M&A like Salesforce, which is you don't worry about the deals that didn't work out. But the thing you regret is the ones you, that you shoul…”
O'Driscoll: Public SaaS growing over 30% should aggressively acquire AI startups
“If you're doing 30% plus, you can be aggressive and you should be aggressive and buy shit as a public SaaS company. What you're saying is if you're doing 10% You can't be aggressive because you've got to fix the growth story... Even though you probably should …”
O'Driscoll: Sierra's $10B valuation requires capturing labor budgets alongside software
“The only way the math works for the Sierra at ten billion is if you eat, you don't just get India, maybe not up front, but over time, you don't just get the service cloud revenue, you get the service cloud revenue plus some slug of the labor.”
Lawson: Atlassian and Mike Cannon-Brookes will stay aggressive on acquisitions
“So I think Mike will be one of the more aggressive on the acquisitions front because they've always been and so I think it's in the DNA of Atlassian, so I do think we are looking at one of those, even if I am not fully on board with the most recent.”
O'Driscoll: Low-growth public tech firms need M&A most but struggle to buy
“When you're at the Dropbox stage in terms of growth, it's just when you need acquisitions the most, But you find it hardest to do as a public company, because you're still in the low-growth penalty box, and that must be a frustrating place.”
Lawson: Cloudflare's DNS position gives it an ideal product expansion surface area
“I always admired the product surface area that Cloudflare had. Because they sit at this super strategic intersection of the world and then your website. And then you can ask the, and it's a dashboard. So once you're inserted into the DNS, and you're proxying a…”
Lawson: Most of Stripe's non-core product portfolio does not drive meaningful revenue
“I have heard whispers that a lot of their product portfolio is not really contributing to the business. It really is the core business, which is, you know, pretty common. I'd say probably the same of, you know, largely the same of Twilio too, right? It's a, it…”
Jeff Lawson: Only three API categories achieve breakaway revenue
“I analyzed every developer thing that was out there, and I decided there were three categories of developer companies, and only, sorry, there were many kinds, only three that actually got breakaway revenue. And there were a lot of folks that were stocked with …”
Jeff Lawson: Developers recreate paid API software internally at scale
“The problem is that developers, A, take your really cool thing that you're trying to charge them for as a challenge. Can I go make that myself? It's like a challenge. Like you're challenging me. You're saying I'm a developer who can't build the thing you built…”
Jeff Lawson: AI inference is not complex enough to sustain premium margins
“I believe inference itself is not such a hard algorithmic solve that, you know, you need to pay someone else to do it for you, but clearly training a model is”
Lemkin: Jailing Fraudulent Founders Is the Best Startup Governance Control
“I think the best control today would be if more founders that committed fraud went to jail. I just think if every month someone went to jail that completely lied, I know Roy disagrees, completely lied in a round, sent financials where they aggregated all their…”
O'Driscoll: Startup fraud peaks alongside market greed rather than generational moral decay
“I think dishonesty ebbs and flows with greed, and I think you tend to see peak dishonesty at a time of peak greed, so you see more of it now, but I don't think human beings have changed.”
O'Driscoll: Scale VP Verifies Bank Balances Directly Before Wiring Funds
“About 15 years back, we instituted the very obvious check the cash balance before you wire test. Because another firm had wired money on the basis of a certain balance sheet. And it sounds really obvious, but they will lie to. So every time we do that, now we …”