Nov 20, 2025 · 1h 26m · 20vc
Peter Thiel and Softbank Sell NVIDIA - Why? & Why VC Will Hit $1TRN and The Opening of Retail · 20VC with Harry Stebbings
Rory O'Driscoll · 33m spoken Tomasz Tunguz · 18m spoken Jason Lemkin · 17m spoken Harry Stebbings · 7m 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
This episode of 20VC features venture capitalists Rory O'Driscoll O'Driscoll, Jason Lemkin Lemkin, and Tom Tunguz discussing the rise of agentic AI coding, high late-stage valuations, macro risks like NVIDIA customer concentration, and the structural shift of venture capital toward secondary markets and retail participation.
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 10.3% of the talking time here. How this is scored →
Harry as informed peer 4.0 Guest teaching 5.2 Guest disagreement 3.3 Harry pushing back 3.4
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 4:55 Rory rejects Jason's $1 trillion developer TAM math
Rory aggressively shuts down Jason's projection that coding agents will capture a trillion-dollar market, calling out the impracticality of assuming huge percentages of GDP allocated to developer tools.
Hardest push from Harry ▶ 1:00:25 Harry challenges Rory on GCAI valuation sensitivityHarry directly confronts Rory over leading GCAI's round at a $550M post valuation, refusing to accept soft explanations and pressing on price sensitivity and portfolio competition.
Biggest teaching moment ▶ 41:00 Tomasz educates on credit default swap spikes and macro riskTomasz clearly explains the mechanics of Oracle credit default swaps to Harry, demonstrating how credit markets are pricing in hidden risks from massive AI infrastructure debt.
Harry holds his own ▶ 33:40 Harry asserts his expertise on late-stage platform leverageHarry pushes back against Rory's view of late-stage difficulty, citing his own media platform leverage to argue that deploying $25M late-stage checks for quick step-ups is far simpler than seed-stage company building.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| The Agentic Coding Revolution & Market Moats | 3 | 4 | 4 | 3 | Harry prompts a debate on Cursor's $29.3B valuation, where guests dive deep into developer TAM expansion, token efficiency, and software margin dynamics. Rory and Jason clash over trillion-dollar TAM projections, while Harry raises competitive threats from OpenAI and Anthropic. | |
| Late-Stage VC Mania & The Step-Up Game | 4 | 5 | 2 | 4 | Harry questions whether early-stage seed craftsmanship is foolish when multi-stage VCs easily capture rapid valuation step-ups like Ramp's jump to $32B. Guests reframe late-stage investing as risk-adjusted capital trading rather than traditional venture building. | |
| Macro Warning Signs: NVIDIA, SoftBank, & Credit Default Swaps | 3 | 6 | 3 | 2 | Harry asks whether stock sales by Thiel and SoftBank signal a market peak in AI. Tomasz educates the group on underlying credit market warning signs, including the tripling of Oracle credit default swaps and rising subprime auto loan defaults. | |
| Early-Stage Dynamics & The Future of Venture Capital | 4 | 5 | 2 | 3 | Harry highlights extreme early-stage YC batch exuberance and asks if total VC capital will reach half a trillion dollars by 2030. Guests reframe the expansion as heavily concentrated in a few mega AI deals while the broader market remains constrained. | |
| Evaluating GCAI and the Myth of VC 'King Making' | 6 | 5 | 4 | 6 | Harry aggressively grills Rory on leading GCAI's round at a $550M post-money valuation despite existing portfolio overlaps in legal tech. Rory defends the investment based on rapid organic growth, near-profitability, and customer pull, rejecting the idea of VC kingmaking. | |
| The Access Premium and the Hidden Costs of Going Public | 3 | 6 | 4 | 3 | Tomasz and Rory explain why private mega-cap startups delay IPOs due to massive public filing friction and the emergence of an access premium in private rounds. Harry suggests going public is reserved for accessing retail capital, which guests nuance. | |
| The Tsunami of Retail Capital and its Ethical Pitfalls | 5 | 6 | 5 | 4 | Harry probes the influx of retail capital into VC funds through 401ks and ETFs, sparking jokes about high management fees. Rory forcefully interrupts the banter to deliver a sobering warning about the moral burden of losing retail investor capital. | |
| The Secondary Market Explosion and the Private Equity-fication of VC | 3 | 6 | 4 | 2 | Jason notes a muted end-of-year IPO market, while Tomasz argues that an expanding private secondary market is permanently replacing traditional IPOs for mid-tier unicorns. The group debates whether secondary liquidity will reshape venture capital permanently. | |
| Quick Fire Round: High-Valuation AI Bets and Founder Incentives | 5 | 4 | 2 | 4 | Harry leads a rapid-fire Kalshi round forcing guests to pick between high-valuation AI bets and predict OpenAI's IPO timeline. The discussion pivots to founder motivation and the unusual corporate dynamics of CEOs who are not motivated by dilution. |
Statements from this episode (39)
Rory O'Driscoll: Winning matters more than price when TAM is huge
“Entry price counts when TAM is unclear. Winning is the only thing that counts when TAM is huge.”
Jason Lemkin: Venture capital requires massive TAM expansion to function
“If you're not seeing massive TAM expansion, there's just no point in even playing as VCs.”
Tunguz: AI coding progress is no longer on a steep improvement path
“Coding is no longer on this extremely steep improvement path.”
Tomasz Tunguz: Next market correction will be fast and brutal
“I think we're at a point where if there's some wobble, the magnitude of the correction will be fast and brutal.”
Tunguz: Agentic coding product-market fit ranks second only to search
“Look, I think product market fit for agent decoding is probably the best of any use case aside from search.”
Tunguz: AI coding assistant Cursor operates with just 30 total employees
“You don't have a lot of ESOP dilution because total employee count is 30.”
Lemkin: AI coding tools will reach 100% developer adoption
“So it's moved to like, we're gonna approach a hundred percent penetration per developer of some sort of price per year, 5006 thousand.”
Lemkin: Built 12 Replit apps used 700,000 times without knowing how to code
“Tomas, I've shipped 12 apps since June on Replit. 12 apps used 700,000 times since June, and I can ship, I built product, but I don't code, right?”
Tunguz: Cursor will retain 75% of its users in five years
“And so as long as they're able to continue to grow, I bet they hold on, I don't know, five years from now, 75% of their audience.”
Lemkin: Replit and Lovable achieve over 50% gross margins using cheap models
“Replit and Lovable, frankly, are using cheap models. Most people don't know or care, and they're well marked up. The margins are north of 50, the gross margins, okay?”
Tunguz: Theory Ventures distilled a 20B model to 1/50th size with 97% accuracy
“And then we've taken a twenty billion parameter model and said Claude Code teach this little model how to call tools. And we can get to 97, I mean, this is a venture capital firm. Yes, we have a great head of AI, but we're not a research lab, and we can get to…”
Lemkin: Replit V3 Far Outperforms Every Other AI Coding Tool
“The latest version of Replit V three blows everything out of the water. It's not just night and day. It's what's more than night and day. It's Pluto and Mercury.”
O'Driscoll: Software Market Shares Lock In After Three to Four Years
“After about three or four years, you grab what share you can, and then in most other markets, then there's a long ten-year, twenty-year period, where even though the market doubles, trebles, 10 X's, the rough market share at the start is the rough market share…”
Lemkin: Replicated Months of AI Agent Prompting in Agentforce in One Day
“So we rolled out agent force for Salesforce. We're probably one of the few organizations of our size to have rolled out agent force. The interesting part is we took the prompt from another AI agent that we trained for months and we gave it to agent force. We i…”
Lemkin: SaaStr now runs 12 AI agents, outnumbering its human staff
“We essentially have 12 AI agents running now at Sastra more than humans. Okay. And we have five sort of SDR BDRs running.”
Lemkin: A price war is coming for GTM AI agents
“There, I think there will be a price war coming, but right now there isn't.”
O'Driscoll: Number of integrations is the top predictor of SaaS customer retention
“I remember looking at churn in SaaS companies, and the number one predictor of retention was the number of integrations.”
Harry Stebbings: Ramp's valuation reached $32 billion from $13 billion in one year
“Ramp was thirteen billion at the start of this year. Now it's 32 with the latest round announced yesterday.”
O'Driscoll: 15% of Q1 newly minted unicorns saw valuation step-ups by Q3
“There was something like 20, 24 minted unicorns in Q one. By Q three, 15% of them already had a step up.”
O'Driscoll: Late-stage VC return profile is entirely dependent on market direction
“The late stage business is either the best business in the world or the worst business in the world, and there's nothing you can do to determine which it is.”
O'Driscoll: Private market trading strategies fail when market dynamics turn downward
“The key sense you're missing is you can't execute a trading strategy if they're private, because when things go wrong, the liquidity won't be, when things go right, you can trade on your way up, but it would be, it'll be a lot harder to get out of one of these…”
Tunguz: Subprime auto loan delinquencies have hit the highest rate in recorded history
“Subprime borrowers in the past 60 days hit the highest delinquency rate on auto loans in recorded history.”
Tunguz: Most hyperscaler GPU capacity is sold out for the next two years
“I mean, most of the hyperscalers GPU capacity is sold out for the next two years.”
Tunguz: Nvidia revenue concentration is 10x worse than Lucent during dot-com bubble
“So Nvidia, two customers for Nvidia represent more than 40% of revenues. Four percent represent more than 50% of revenues. I went back and looked at the dot-com era, the networking companies. NVIDIA is 10 times more concentrated in terms of revenue than Lucent…”
Lemkin: US AI Agent Population Will Soon Exceed Humans
“There'll be more agents in this country than humans soon enough.”
Lemkin: 100% chance the US venture market hits $500B by 2030
“A hundred percent chance. So maybe more than a hundred.”
Lemkin: Half of the $184B invested in VC in 2025 went to four companies
“They're estimated in a hundred eighty four billion in venture capital invested this year by their definition. Okay. The peak was 20, 21 of a hundred eighty three billion. So one more billion this year. Okay. But four of those half of the hundred eighty four bi…”
Lemkin: 900 late-stage unicorn startups generating nine-figure revenues will never go public
“And then we've got 900 unicorns that are never going to IPO. Poor guys. We have nine, we all have one or two in our portfolio that are at nine figures in revenue that are still growing and will never IPO. And there is no PE buyer for them.”
Rory O'Driscoll: VC backing is not dispositive for corporate software buyers
“Step back in the wider us, I don't buy this idea that because X company got money from You know, YVC, that the average corporate buyer cares all that much. They want to solve that problem. So I'm not a believer in king making big dispositive when you have grea…”
Tom Tunguz: Microsoft went public with $50M revenue and $2M fees
“Like I went back and looked at Microsoft, you know, you needed like fifty million in trailing revenue and six quarters of profitability to go public, right? And the cost to take a company public was a couple million bucks.”
Rory O'Driscoll: IPO transaction costs consume 6% to 7% of funds raised
“Well, it's seven percent. It's six to seven percent of the raise and the raises are now 203 hundred million.”
Rory O'Driscoll: Stripe will only IPO when private market capital dries up
“The only time Stripe will go public, we've said this on a call before, is when the capital available in private markets is towards ransom.”
O'Driscoll: VC reaction cycles take up to seven years
“In the end, the only thing that matters is returns. The only question is how long does in the end take? We're in an industry which has very long reaction cycles. You know, you put in the money, you don't get single for five years, you don't figure it out for s…”
Stebbings: Coatue has raised $3B in retail venture funds
“We've got Code Two with three billion dollars now in retail funds, and we're seeing GC be very aggressive in opening up retail funds.”
Tunguz: Secondary transactions now account for 10% to 12% of VC liquidity
“So private equity, you look at private equity, total fraction of dollars in secondaries is a fraction of the asset class is about 25%. Historically venture has been about two to three. Now we're kind of like 10 to 12.”
Tunguz: Goldman Paid Record Asset Manager Multiple for Industry Ventures
“Goldman bought Industry Ventures, right? Leading secondary fund. Paid, paid the highest multiple, I think, ever for an asset manager.”
O'Driscoll: Private Capital Fee Drag Is Significantly Inefficient vs Public Markets
“The private capital has two and 20 fee drag, and the public has almost, you know, 60 bips fee drag. So from a societal perspective, there's no doubt in my mind that assets being managed privately have a far higher aggregate costs between cost to the issuer and…”
Tunguz: PE Took 12% of Public Software Companies Private in 2022
“I think in 22, I calculated PE had taken private 12% of all publicly traded software companies in a year.”
Lemkin: OpenAI's IPO will be delayed until mid-2027 as Altman secures alternative financing
“I think that's a good idea. I think Sam will come up with so much alternative financing. It'll slip into mid 27, but I think that's the straw man today would be my guess. Like that's the plan, but there'll be so much, so many other sources of maybe the governm…”