Jun 25, 2026 · 1h 28m · news
DeepSeek Raises at $50B | The Rise of Open Source vs OpenAI & Anthropic | OpenAI Builds Own Chip · 20VC with Harry Stebbings
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
Host Harry Stebbings leads a deep-dive discussion with investors Rory O'Driscoll and Jason Lemkin on the volatile landscape of generative AI, analyzing the shifting talent wars, the macroeconomic crisis of massive CapEx spending, the rise of powerful Chinese open-source models, and the existential threat AI poses to traditional enterprise business models.
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 8.9% of the talking time here. How this is scored →
Harry as informed peer 5.0 Guest teaching 5.8 Guest disagreement 4.1 Harry pushing back 3.6
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:15:35 Jason's 'Omega Watch vs Rich' Work Ethic Rant
Jason aggressively rejects modern work-from-home and work-life balance trends, demanding founders choose between an Omega watch at a comfortable job or working 6.5 days in an office to get rich.
Hardest push from Harry ▶ 53:11 Harry Presses Guests on Menlo's Fund SizeHarry directly refuses the framing that $3B is a large fund for Menlo Ventures, pressing the guests on why they didn't raise much more given their Anthropic returns compared to mega-funds like Thrive and Lightspeed.
Biggest teaching moment ▶ 7:06 Rory Corrects Harry on Medical Invention TimelinesRory dismantles Harry's rumor about Anthropic having a secret scientific breakthrough by educating him on the 10-15 year economic lag of biotech developments compared to instant LLM deployment.
Harry holds his own ▶ 29:54 Harry Cites Goldman Sachs CapEx vs Sequoia ROI GapHarry demonstrates deep financial prep by contrasting Goldman Sachs' $7.6T cumulative AI CapEx projection with Sequoia's $600B revenue gap model.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Google DeepMind Talent War and Research Environments | 3 | 6 | 4 | 2 | Harry introduces the DeepMind talent departures and asks whether Anthropic achieved an undisclosed technical breakthrough. Rory refutes the rumor by explaining that biological research commercialization takes 10 to 15 years, unlike fast-moving LLM products. | |
| The 'Number Three' Closed Source Dilemma and the Rise of Open Source | 3 | 6 | 2 | 3 | Harry asks why holding the number three spot in closed-source LLMs is more dangerous than in cloud infrastructure. Jason and Rory explain model routing, state-subsidized Chinese open source, and oligopoly pricing dynamics. | |
| Geopolitics of AI: Sovereignty, DeepSeek's Fundraise, and China's AI Prowess | 6 | 5 | 2 | 3 | Harry demonstrates strong preparation by detailing DeepSeek's $7.4B fundraising structure and state control, as well as Zipu AI's coding benchmarks. The guests elaborate on geopolitical sovereign taxes and open-source distillation trends. | |
| The Macroeconomics of AI CapEx, Memory Costs, and ROI | 7 | 5 | 2 | 4 | Harry brings data-heavy questions, citing Tim Cook's memory cost warnings, DRAM contract price surges, and Goldman Sachs' $7.6T CapEx projection. Rory and Jason contextualize these figures with labor replacement calculations and enterprise ROI hurdles. | |
| The Rise of Autonomous AI Agents and the Death of Prompt Engineering | 5 | 6 | 5 | 3 | Harry prompts the guests with Kapathy's coding stats and Brandon McCourt's quote predicting agent training will be the top job category. Jason dismisses McCourt's premise, noting how quickly prompt engineering died as workflows automated. | |
| AI Startup Unit Economics: Growth vs. Gross Margins | 5 | 6 | 6 | 4 | Harry quotes a tweet from Nicholas Desain asserting that poor gross margins cause early-stage AI failures. Rory explicitly disagrees with the premise, citing Cursor and major foundation models as evidence that hypergrowth routinely overcomes negative initial margins. | |
| Venture Fund Sizing, Menlo Ventures, and the SPV Model | 6 | 5 | 3 | 6 | Harry challenges the guests on why Menlo Ventures raised a conservative $3B fund following their Anthropic win rather than matching mega-funds like Thrive or Lightspeed. The guests respond by highlighting SPV flexibility and fund return math. | |
| The Prediction Market Boom: Kalshi's IPO Prep and Meta's Threat | 5 | 5 | 4 | 4 | Harry brings up Kalshi reaching a $2B run rate ahead of an IPO and probes into potential regulatory threats under new political administrations. Rory breaks down CFTC prediction market arbitrage, while Jason debate's Meta's competitive potential. | |
| Accenture's Decline and the AI Disruption of Professional Services | 5 | 7 | 2 | 5 | Harry highlights Accenture's 40% year-over-year stock decline, challenging the assumption that IT services automatically benefit from AI. Rory and Jason explain how LLMs replace systems integration work and dismantle head-count billing models. | |
| Work From Home Debate and Hardcore Startup Culture | 4 | 7 | 8 | 3 | Harry introduces Ryan Peterson's quote calling work-from-home white-collar fraud. Jason forcefully agrees and goes on an extended rant rejecting work-life balance in favor of 6.5-day workweeks in the office. | |
| OpenAI's Custom Chip (Jalapeno) and Vertical Integration Risks | 6 | 6 | 7 | 3 | Harry presents OpenAI's Jalapeno custom chip announcement with Broadcom. Rory strongly criticizes OpenAI's decision to vertically integrate into hardware when hyperscalers absorb hundreds of billions in CapEx, leading to a heated argument with Jason. |
Statements from this episode (37)
Lemkin: Modern open-source AI is fake because China subsidizes training costs
“Open source is a bit of a fake because China's paying for all the trading, okay? It's not open source like a generation ago.”
O'Driscoll: Software pricing based on headcount is worse than seat-based models
“There's only one thing worse than a seat-based model, and that's a model that's based on bodies.”
O'Driscoll: OpenAI and Anthropic rely on $300B spent by other companies
“The whole reason the OpenAI and Entropic models work is because other idiots have spent the three hundred billion dollars on their behalf.”
O'Driscoll: Converting medical AI discoveries into commercial value takes 10 to 15 years
“The lag time between I've had amazing invention and revenue in the kind of LL, core LLM space is a year, or two years, or three years. The lag time if you have a medical invention, an idea, and turning that into real meaningful economic value is 10 or 15 years…”
O'Driscoll: AlphaFold won a Nobel Prize without producing a commercial drug
“The protein folding advance has collected its Nobel Prize and as yet has not had meaningful commercial success or a drug in production yet.”
O'Driscoll: Only Nvidia and Google outperformed the S&P in last 18 months
“Over the last 18 months since 2025, only two of the Mag seven have outperformed the S&P. It's NVIDIA and Google”
O'Driscoll: Impossible for a fourth closed-source AI player to catch up
“It's almost impossible for a number four to a same business model, closed source number four to emerge and catch up at this point.”
O'Driscoll: Most competitive open-source AI foundation models are Chinese-based
“All the competitive models to the foundation models, all the open source models, bar not quite all, but most of are Chinese-based.”
Lemkin: DeepSeek's AI model is intentionally crippled inside China
“And DeepSeek is intentionally crippled in China. It's not as good as it is here. It can't search the web and it is trained on different data as near as I can tell.”
O'Driscoll: Anthropic delayed shipping its latest model due to US government concerns
“We're seeing Anthropic unable to ship the most recent model until they satisfy the concerns of the US government.”
O'Driscoll: Chinese open-source AI caps profitability for US closed-source vendors
“Many US companies are starting to build their own model based on the open source alternative, so it's providing some kind of, it's providing a ceiling on profitability for some of the US closed source vendors.”
O'Driscoll: Apple will raise iPhone prices rather than accept lower margins
“Specifically what it's going to mean is Apple is clearly decided wisely not to swallow the loss and lower their margins. They're just going to raise prices. And they'll sell a few less iPhones because it's probably, you know, there's some price elasticity. And…”
O'Driscoll: The AI industry generates $100B revenue while spending $700B on CapEx
“So AI as a whole is getting a hundred billion in revenue and spending seven hundred billion a year. That's not a great business.”
Lemkin: OpenAI and Anthropic yield 40-70% gross margins on enterprise inference
“They have an enterprise base, which is, it has, you know, between 40 and 70% in, you know, gross margin just on inference.”
Lemkin: Enterprise CIOs will demand proven ROI for AI spend by 2027
“I think going into 20, 27, CIOs and others are gonna be like, okay, just show me the F and ROI. We're not gonna, not just gonna ration tokens based on who we like the most in the company and who makes the best PowerPoint pitch. Show me the ROI.”
O'Driscoll: Widespread AI adoption will boost productivity but not industry profitability
“You can have an improvement in productivity and a massive improvement in productivity in industry by virtue of the advent of an enabling technology like AI, and no improvement in profitability for that industry if everyone adopts the technology.”
O'Driscoll: AI efficiency gains will cut white-collar service costs by 20%
“If consulting or any of these companies, any of these white-collar jobs are 20% more efficient with AI, Then everyone should adopt it, and then everyone will be, then the cost of those white-collar jobs, like audit, should be reduced by 20%.”
Lemkin: SaaStr built an AI VP of Finance better than human staff
“We just built an AI VP of finance while we were in China and it's already better than any of the humans on our team.”
Lemkin: AI agents will reduce corporate contractor spend by 50% to 60%
“If, and if it's a variable price, you could see what you pay to these humans fall by 50 to 60% without even intentionally trying to save money.”
Lemkin: Prompt engineering skills that once paid $150,000 are now worthless
“When we started this show, you need to be a fucking prompt it. They would joke this was the highest paying job out of college was prompt engineers were making a 150,000 dollars out of college because they knew how to write a prompt. That skill's worthless toda…”
O'Driscoll: AI startups with poor margins historically raised capital successfully
“The objective reality is that's not what was happening. In other words, companies with tough gross margin profiles and hyper growth have been getting funded and frankly have been able to improve their gross margins and pull it off. And you can, the foundation,…”
Lemkin: Menlo Ventures raised a $3B fund following its Anthropic bet
“Menlo just raised a three billion dollar fund after 50 years on the back of a crazy bed in Anthropik that, that, that went very well, right? When the margins were crazy”
Lemkin: Doubts recovery on recent AI portfolio investments with negative margins
“I'm sure all three of us are sitting on a portfolio company investment we made in the last 12 to 15 months. Where inference was the marketing strategy and the gross margins were negative. And we're sitting here today and we're like companies doing okay, but I'…”
O'Driscoll: Margin-negative AI startups face bankruptcy when growth decelerates
“And now you're just in a shitty gross margin profile business, which will go bankrupt. Because when things slow down, no one says, hey, let's do an acquisition of an adjacent gross margin negative thing, and it's gonna be fun.”
O'Driscoll: Smaller VC funds offer better risk-adjusted returns than mega-funds
“Provided you've got a big enough check size to play in any round as a GP, you actually, quote unquote, have higher risk adjusted return if you have smaller funds because then you have less cross deal aggregation. In other words, two separate one billion dollar…”
O'Driscoll: 70% of Q1 venture dollars went to just four or five deals
“One of the statistics in Q-one, like, 70% of the venture dollars went to four or five deals.”
Lemkin: SPV models require friction-free, instant LP capital deployment via WhatsApp
“For the SPV work, it needs to be one WhatsApp message, and I get the two hundred million to invest. Then I'm all, then it's the best model venture there is, right?”
O'Driscoll: 90% of Kalshi's prediction market activity is actually sports betting
“90% of what they do is sports betting.”
O'Driscoll: Multiple US states are actively suing prediction market Kalshi
“There are currently some states suing Kalshi, saying you really are doing betting, so we should regulate you.”
O'Driscoll: IT consulting and systems integration are prime targets for AI disruption
“The core business they're in that, you know, 99%, 90% of their revenue is other consulting, and there are few markets more primed for disruption by AI than consulting in general, because it's wide-collar work. It's very, it's already been outsourced. That's wh…”
Lemkin: Seat-based software companies are getting crushed by variable pricing models
“Everyone selling seats for the most part is getting crushed. Everyone selling variably one way or the other is winning.”
Lemkin: LLM-driven migration tools are destroying traditional B2B software moats
“It's a moat destroyer when LLMs will lift you from one vendor to the other, and I, you know, I don't even, like, I literally was doing a pitch this week, and the founder was going on and on about their moats, and I immediately didn't want to invest. Like, I ju…”
O'Driscoll: Anything outsourced to offshore BPO will be replaced by AI
“When you think about what work AI will replace, you know, white collar work, you can say BPO is a good proxy for that because anything a company is willing to outsource to India, they're probably willing to outsource to AI, right?”
Lemkin: Will only invest in startups working in-office six days weekly
“I want small high paid teams that work in the office over six days a week. That's I'm not interested in investing in anything else. I'm just not interested. And it's not because I don't have empathy. It's because they're gonna fail.”
O'Driscoll: OpenAI and Anthropic found the top consumer tech market in 20 years
“OpenAI and Anthropic have discovered the single best tech market in the, in terms of consumer demand in the last 20 years, and they should put all their effort into meeting that demand, right?”
O'Driscoll: AI companies building memory foundries will signal an immediate market collapse
“I really believe what you're saying is if there's the next stage is when someone comes in and says, what we need to do is build a foundry and make our own DRAM. At that point, you'll know the cycle is about to go. Really badly down, you know. If you vertically…”
Lemkin: Open-source models will capture middle-tier enterprise AI workflows
“I firmly believe we need these middle level models from the closed source providers for the business to work, okay? We can't, not, not everyone can afford to run every workflow on Opus, right? And Haiku and Mini are too small, and this is where open source is …”