Jul 9, 2026 · 1h 27m · news
Enterprises Fear Frontier Models | Sam Altman Offers Trump 5% of OpenAI | DeepSeek Builds Own Chips · 20VC with Harry Stebbings
Clips from this episode (6)
Short vertical cuts produced from the tape, captions burned in. Where a cut lands on a statement from the ledger, its card says so.
0:00 / 1:05
0:00 / 0:42
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 the 20VC podcast, host Harry Stebbings and co-hosts Rory O'Driscoll and Jason Lemkin dissect the rapidly evolving AI landscape, analyzing regulatory developments, the economics of compute infrastructure, enterprise adoption strategies, and recent shakeups in the venture capital ecosystem.
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 3.0 Guest teaching 5.3 Guest disagreement 3.3 Harry pushing back 1.7
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 8:34 Rory Rejects Jason's Alignment Premise
Rory aggressively cuts in to reject Jason's claim that giving 5% to the US government creates alignment, citing Microsoft's tense 30% ownership of OpenAI to prove equity ownership does not prevent conflict.
Hardest push from Harry ▶ 51:54 Harry Challenges Chinese Valuation HypothesisHarry directly interrupts Jason to refuse his premise of a Chinese AI valuation bubble, pointing out that DeepSeek at $50B and ByteDance at $500B trade at heavy discounts relative to US equivalents.
Biggest teaching moment ▶ 38:07 Rory Corrects Harry's Framing on Compute DemandRory explicitly tells Harry he is phrasing the question wrong regarding Zuckerberg's execution skills, explaining that success depends on underlying market demand for compute rather than individual leadership.
Harry holds his own ▶ 51:54 Harry Counters with Market Valuation DataHarry demonstrates strong domain knowledge by immediately shutting down Jason's speculation on Chinese AI bubbles with concrete valuation numbers from DeepSeek and ByteDance.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Washington Lifts the Fable 5 Ban | 3 | 4 | 1 | 1 | Harry introduces the lifted Fable 5 ban, prompting the guests to explain the broader shift toward structured pre-approval processes in US AI regulation. Rory compares this shift to European GDPR oversight, while Jason outlines how token-based pricing will render it a niche issue. | |
| Analyzing Sam Altman's 5% OpenAI US Government Stake Proposal | 4 | 5 | 7 | 2 | Rory strongly criticizes Sam Altman's 5% US government stake proposal as grandiosity akin to Atlas Shrugged. When Jason defends it as an alignment mechanism similar to Klaviyo and Shopify, Rory aggressively cuts in, citing Microsoft's troubled 30% stake in OpenAI to prove equity does not create political alignment. | |
| Founder Dilution and the Psychology of Late-Stage Capital | 3 | 6 | 3 | 2 | Jason and Rory discuss how late-stage capital rounds have normalized massive founder dilution, referencing Dario's low equity percentage at Anthropic. Jason pushes back on Harry's praise for Linear's capital efficiency, questioning whether skipping growth rounds limits ultimate exit scale. | |
| Alex Karp on CNBC: Enterprise Skepticism and Data Privacy | 3 | 5 | 2 | 1 | Harry prompts a discussion on Alex Karp's CNBC appearance regarding enterprise skepticism around AI ROI and data privacy. Jason highlights HubSpot's recent backdown on sharing prospect data to show how vendors are pushing boundaries on corporate IP. | |
| Meta Launches Metacompute and the Hyperscale Capex Strategy | 3 | 6 | 3 | 2 | Harry asks if Mark Zuckerberg can execute on Metacompute as well as Elon Musk did with excess compute. Rory corrects Harry's framing, explaining that success depends on underlying market demand rather than CEO execution, while analyzing hyperscaler Capex dynamics. | |
| Nvidia's Aggressive Growth: Financing Its Own Demand | 2 | 6 | 1 | 0 | Harry introduces Nvidia financing its own demand via compute backstops. Rory delivers a detailed breakdown of AS 606 revenue recognition upfront versus long-term contingent liabilities, comparing the structure to late-90s telecom vendor financing. | |
| The Rise of Custom AI Chips: Anthropic, OpenAI, and DeepSeek | 2 | 5 | 5 | 1 | The guests explore why AI model labs like Anthropic, OpenAI, and DeepSeek are developing custom chips. Jason forcefully rejects the premise that Nvidia cannot meet specialized customer needs, calling the custom silicon argument corporate fluff to mask margin pressure. | |
| AI Video & The Chinese AI Ecosystem: Kling vs. Sora | 5 | 5 | 4 | 6 | When Jason speculates that Chinese AI video models like Kling operate in a valuation bubble, Harry immediately pushes back with concrete market metrics on DeepSeek's $50B valuation and ByteDance's $500B valuation to prove Chinese models trade at discounts. | |
| Frontier Models vs. Open Source: Performance & Cost Management | 3 | 6 | 2 | 1 | Harry notes enterprise adoption moving toward open-source models. Jason uses a personal coding example in Replit versus Claude Opus to illustrate why paying high costs for frontier models is far cheaper when solving complex, unbounded problems. | |
| Enterprise AI Adoption: Microsoft's Service Play vs. P&L Impact | 3 | 6 | 6 | 2 | Harry cites MIT's study showing 95% of enterprise AI pilots produce no P&L impact as Microsoft deploys 6,000 engineers. Jason forcefully interjects that the strategy will fail due to a severe shortage of qualified forward-deployed engineering talent, triggering a debate with Rory over IT change management. | |
| Ashton Kutcher's VC Move: Personal Brand vs. Firm Equity | 2 | 5 | 2 | 1 | Harry asks about Ashton Kutcher leaving Sound Ventures to start a new firm with Morgan Beller. Rory explains that celebrity GPs operate under different rules because Kutcher's personal brand carries higher equity than traditional VC firm names. | |
| Employee Liquidity & The 'Single-Shot VC' Strategy | 3 | 5 | 4 | 1 | Jason argues that talented startup employees should only join companies with clear secondary tender offer programs. Rory disputes this framing, noting that employees must act like 'single-shot VCs' and join prior to tender offers to capture true equity upside. |
Statements from this episode (44)
Lemkin: Founders No Longer Care About Making Late-Stage Investors Money
“No one's worried about making their last round high-priced investors money anymore. Literally no one is.”
O'Driscoll: Every Tech Company Either Fails or Becomes Next-Gen IBM
“Every technology company either goes bust or lives long enough to become next generation's IBM.”
O'Driscoll: AI developers now require Washington permission before shipping software
“Six months ago, you could ship software like a free man, and now you have to get permission from Washington before you do it. It's a big change, right?”
O'Driscoll: Tech volunteering for Washington oversight is Atlas Shrugged rewritten
“You start with pre-approval, suddenly you end up with an ownership interest, then you end up with a board member, and what the, are these people doing? It's like rewriting Atlas Shrugged. Where John Galt goes to Washington and says, why don't you regulate me m…”
Lemkin: Sam Altman is one of the most successful investors ever and runs OpenAI with that playbook
“Sam Altman beyond being the CEO of OpenAI is one of the most successful investors of our of effort of all time. And he knows everything about how startups and scale ups are run and he's seen it all. And I think in some ways he runs open AI with that playbook i…”
Lemkin: Selling 5% Equity to Corporate Partners Creates Disproportionate Alignment
“What I've learned from my portfolio and I think you guys will agree is it creates an out an unexpectedly large amount of alignment. You get, you sell five percent of your company to a hundred billion dollar partner. It don't matter to Rory's point. It just don…”
O'Driscoll: Microsoft and OpenAI Are in a Stale Marriage Looking for Divorce
“Microsoft owns 30% of OpenAI. If ownership stake resulted in besties, they'd be besties. They're not besties. They're in a stale marriage looking for a divorce, but can't quite pay, you know, pay the tax, right?”
O'Driscoll: AI Will Not Replace 50% of White-Collar Jobs in 5 Years
“I mean, look, in five years' time, if Dario is correct, and 50% of white-collar jobs have been replaced by AI, which I don't believe for a second, Then you damn right. There's going to be political. Political controversy. And if you think five percent is going…”
Lemkin: Massive Dilution Has Been Institutionalized in the AI Era
“I think in the age of AI massive dilution has been sort of institutionalized.”
Lemkin: Effective seed valuation is four times nominal price due to dilution
“As a seed investor, I used to think my real entry price was twice what it looked because of dilution. Right now I'm thinking it's four times.”
O'Driscoll: Carta data shows dilution per round is decreasing
“The only fact-based comment I'll make is, you know, the data from Carter, which is always excellent, says that dilution per round is going down.”
O'Driscoll: Trillion-dollar prizes require capital, but over-raising kills billion-dollar exits
“Look, there's no doubt if the price is a trillion dollars, which it has been in at least three cases, it looks like, it really doesn't matter what it takes to get there. You just have to get there. And if skimping on it reduces the probability of getting there…”
Palantir Stock Rose 9% Day of Alex Karp CNBC Interview
“And worth pointing out, people can bitch in the moment, but the stock went up nine percent on the day, right?”
Lemkin: OpenAI and Anthropic Lied About Books and YouTube Training Data
“Open AI, an anthropic, Kind of lied about the books, and they definitely lied about training on YouTube”
Lemkin: Slowing Software Vendors Will Cut Privacy Corners for AI
“Every vendor that is seeing massive competition or slowing growth is going to be tempted more and more to cut corners on training, privacy, and HubSpot got caught.”
O'Driscoll: Meta and SpaceX rented compute after failing to build AI assets
“Two companies Have done the same thing, which is buy a load of compute to build proprietary assets, fail to build those assets, and then bring that, and then decide instead to sell that compute to others, and both of them have had a very positive market recept…”
O'Driscoll: Meta's buy-it-then-resell compute logic holds only while demand stays tight
“And that's the risk is that it turns out that, you know, right now the assumption and Zuckerberg said it is, hey, we should invest because if we can't use it, we can always sell it. And that, and this is what you're seeing right now. Everything there is true u…”
Lemkin: Meta invested $900M in CRED to hire WhatsApp head Kunal Shah
“Zuck also just paid essentially nine hundred million dollars to hire a head for WhatsApp, right? By investing nine hundred million dollars into cred, right?”
O'Driscoll: Board members must vote no on decisions with fatal error risk
“One of the big picture jobs you have, and you have very few jobs, But one of them is, if the company is doing something that could have fatal error risk, that is when you at least record a no vote, and you say, I wouldn't do this, right?”
O'Driscoll: AI capex spending will be stopped by demand, not supply
“This isn't, this spending isn't going to stop because the supply side says stop. Facebook aren't going to say stop. Google isn't going to say stop. Microsoft isn't going to say stop. Really, it boils down to the demand side. As long as, and enterprise customer…”
Lemkin: round-trip revenue is suddenly cool again, so everyone will do it everywhere
“I love it. Now that round trip revenue is like totally cool and not something you go to jail for, like, let's, let's do it every single place we can find it, right? Let's just do it. And I'm not saying there's anything literally wrong with it, but go for it, r…”
O'Driscoll: Nvidia Recognizes Hardware Revenue Upfront Under ASC 606 With Put-Back Rights
“They're basically, quote, selling you the chips up front, so they are going to recognize that hardware revenue up front, and then they're giving you, as the buyer, the NeoCloud, a backstop that if you can't use that compute, you get kind of put back rights on …”
O'Driscoll: Slowing Compute Demand Will Force Nvidia To Debook Prior Revenue
“And if that slows down and there's excess capacity, these deals will look horrible because you'll not just not be, if you're Nvidia, you'll not just be not growing quickly. You'll be debooking prior revenue. You'll be kind of taking, you know, you'll be taking…”
O'Driscoll: Nvidia Has Tapped Out Cash-Rich Buyers And Must Subsidize Growth
“We've reached a point where the number of good customers who can pay cash and have a big balance sheet is tapping out. So you got to find more customers to keep the growth going. And to do that, you got to subsidize them.”
Lemkin: Nvidia would build custom chips for OpenAI if requested
“The only thing that makes zero sense to me is the argument that, hey, at OpenAI, we need to build our own our own chips because we have very specialized needs that NVIDIA can't meet. I have a little bit of experience in the semiconductor industry. If you're dr…”
Lemkin: Custom AI chip push is about margin capture, not customization
“This is just responding to believing that the margins are so high in video to survive. We have to recapture that margin. I just think the idea that it's customized for us is just soft language because everyone's kind of dancing around being, being Zuck-esque a…”
Stebbings: AI video startup Kling will IPO on Hong Kong Exchange
“Kling raises 2.8 billion at an eighteen billion dollar valuation. It's the biggest AI video business in the world. It's doing five hundred million in Q-one, ARR-wise. It's clearly gonna go public in the Hong Kong Stock Exchange soon.”
Lemkin: Higgsfield hits $500M revenue and $2M daily credit billings
“Higgs field just announced they're at five hundred million in revenue actually doing Two million a day now just in credit card billings outside of the enterprise.”
O'Driscoll: 30-second AI video costs $1.30 to $2.00 in GPU compute
“I saw an estimate as well, Justin, for a kind of a 32nd video generation. It's about a buck 30 to two dollars in GPU generation costs.”
Stebbings: Top six models on OpenRouter are Chinese
“The top six models as of today on open router are Chinese.”
Lemkin: Claude Opus Solved Algorithm in 20 Minutes That Open Source Couldn't
“After spending about 10 hours in Replit, I couldn't solve this big algo problem I solved in about 20 minutes in Fable and Opus, right?”
O'Driscoll: Open-Source Token Volume Is Misleading as Revenue Stays With Frontier Models
“The open router, you know, all the tokens with L, you know, with open source is a little misleading, because all the tokens can be in one place, but all the dollars can be in the other place.”
Lemkin: AI Customer Support Pricing Is Standardizing at $0.50 Per Resolution
“We're kind of standardizing this industry around like 50 cents per resolution. Okay. In CX, right? That's sort of the cost, right? It's gone down from a dollar”
O'Driscoll: Enterprise Clients Will Pay Double for Higher AI Resolution Rates
“This is a category where everyone can articulate the 30% where it is there. And then to your point, they can go, ooh, the next 10% is going to cost more. That's a high class problem. Maybe what you're saying is you get, you go from 30% resolution to 65% resolu…”
Lemkin: Microsoft's $2.5B, 6,000-person enterprise AI services push will fail
“I think it's going to fail. I'll tell you why. And then one leader, not Salesforce, not a leader, our FDE went on paternity leave for three months. And the new one told us they couldn't fix our bug for three months until the first guy got back.”
Lemkin: Microsoft's enterprise AI deployment strategy will fail
“I think this is going to fail because I don't think there is enough talent To do what we want to do in the enterprise. The idea makes sense to on a spreadsheet. It makes sense. Sad is smarter than me, but I, my today, my experience is it's going to fail with a…”
O'Driscoll: Microsoft will build a massive, lower-margin AI services business
“So summary, I think Microsoft will build a huge services business here if they want to, which also speaks to it won't be nearly as profitable as selling operating systems.”
O'Driscoll: OpenAI reached $12B and Anthropic $4B in GAAP revenue
“The rate at which, you know, OpenAI and Tropic got to four and twelve billion, respectively, other way around, sorry, 12 and four billion, respectively, in GAAP revenue, never before seen, right?”
Lemkin: Legal AI startup Harvey embeds a lawyer in every customer deployment
“Every deployment they do at Harvey has an FD and a lawyer. Every single deployment has a lawyer, right?”
O'Driscoll: VC Partners Stay Together to Preserve Firm Brand Equity
“Why do venture firms hang together and Paper over the story because, you know, the asset is the firm and the name, and even if you hate each other, you want to manage the process well so you can keep the thing going because the firm has a brand and a reference…”
O'Driscoll: Optimal employee equity timing precedes secondary tender offers
“The whole trick for employees, just like it is for VCs, is to join something that isn't doing a tender offer today, get a healthy grant, and join a company that within a year or two, when you've vested 50, 60% of your thing, starts doing tender offers.”
Lemkin: Tech workers should only join startups with 24-month tender visibility
“Why is a, why would you join anything that you don't have high certainty, not just they're going to be a unicorn because it's not good enough, that they're going to have tender offers, right, in the next 24 months.”
Lemkin: OpenAI and Anthropic have eliminated equity vesting cliffs
“Well, now that they're, now that there aren't even vesting schedules at Open Ananthropic, those issues have been solved. Right? A lot of startups don't have vesting schedules for top employees, right? It doesn't mean you vest into all your stock, right? Sorry,…”