Jason Lemkin | 20VC Summary

1. Key Themes

The AI Compute Scarcity Economy Is Forcing Strategic Rationalization

The hidden story beneath OpenAI killing Sora isn't failure — it's the emergence of compute as a scarce, rationed resource that forces prioritization. AI companies are now being forced to allocate compute to highest-revenue-per-compute-unit activities, which will reshape which AI products survive.

"You're seeing the economists, the accountants have wandered into the room and they said, we have a scarce resource here. Let's optimize it. Let's devote this compute to the people who can pay the most for it." — Rory O'Driscoll 00:14:01

"Video generation is extraordinarily compute intensive, relatively speaking, and the revenue is almost minuscule. Conversely, cogen, while it is compute intensive, is orders of magnitude less compute intensive, and there's real dollars attached to it." — Rory O'Driscoll 00:13:32

The Agentic AI Era Will Be the Golden Age of Cybersecurity — Not Its Destruction

Despite cybersecurity stocks selling off 6-9% on the Anthropic Mythos leak news, the panel argued the opposite: agentic AI will massively increase threat surface area, driving demand for security. This is one of the most actionable investment thesis points in the episode.

"If you're in the agentic world, this is the golden age of security. The number of security threats and issues is going up orders of magnitude." — Jason Lemkin 00:23:52

"Instead of having people trying to get into your firewall, everyone is now downloading an agent, giving it full root access to their computer and telling it have a go." — Rory O'Driscoll 00:25:48

"No one yet knows the exact approach that we're going to have to take to defend against agents running within the organization. But everyone 100% understands that this is an emerging mega threat because of the velocity adoption times the power of the solution." — Rory O'Driscoll 00:26:16

ARR Is Being Gamed Into Meaninglessness Across the AI Stack

From tranche rounds to token double-counting to $0/month free trials being recognized as $240 ARR, the entire revenue measurement ecosystem in AI is being systematically distorted. This has implications for investors underwriting AI companies at massive multiples.

"If I do a $0 a month product that's discounted as a marketing cost and I churn after 30 days, does that count as $240 of ARR? I think for a lot of startups, it does." — Jason Lemkin 00:33:12

"Cursor is selling it again and recognizing the revenue, right? The same... people keep reselling these tokens again and again and recognizing them as their own ARR. How many times do we get to resell these poor little tokens?" — Jason Lemkin 00:30:57

"Until we all have to get profitable, all this can continue. And then at some point, someone's going to have to say... assuming we want to have a net present value and a cash flow, what's going on here?" — Rory O'Driscoll 00:31:49

2. Contrarian Perspectives

The Cybersecurity Selloff Was Completely Backwards — It Should Have Been a Buy Signal

Most investors panic-sold cybersecurity stocks when Anthropic's Mythos leaked, fearing AI would replace security tools. The contrarian read: agentic AI creates exponentially more attack surface, making security more essential, not less. The selloff created a potential buying opportunity.

"This is one where it's just back-ass backwards. If you're in the agentic world, this is the golden age of security." — Jason Lemkin 00:23:52

"Everyone should be benefiting when you see an explosion in application production and a change in the paradigm. The change in the paradigm is good for everybody except Windows Defender from 1996." — Jason Lemkin 00:25:33

Killing Sora Was a Sign of Strength, Not Weakness — And a Strategic Admission Worth Noting

The conventional narrative was that killing Sora embarrassed OpenAI. The contrarian take: it was exactly the right capital allocation decision, and refusing to kill bad bets is what kills companies. The real concern is the prior strategic error of betting on consumer video at all.

"I think shooting in the head is even more significant. I think it's saying that a big part of the whole strategic direction of the company was flawed." — Jason Lemkin 00:12:20

"If it's a mistake, give him credit for at least saying it's a mistake, move on." — Rory O'Driscoll 00:12:06

The "Singapore-Washing" Play for Chinese AI Companies Is Now Dead

The conventional VC wisdom was to redomicile Chinese AI companies to Singapore to enable US acquisitions. The Manus founders being trapped in China by the Chinese government after the Meta acquisition closes that playbook permanently — and has implications for every fund that was using this structure.

"I don't think another deal like this would happen to you. I think this whole Singapore, Washington thing is over. It's over." — Jason Lemkin 00:54:10

"Every Chinese founder who was thinking about doing this is going, hmm, I don't know if I can do this deal. I do know if I do this deal, I am never going home again." — Rory O'Driscoll 00:54:15

Tranched Rounds Are a Mathematical Deception That Hurts Later Investors More Than Anyone Admits

The market celebrates headline valuations on tranched rounds without acknowledging that late participants are paying a 50% premium to be associated with a "hot deal." The lead investor knows the real blended price; followers don't.

"The lead investor, let's say it's Sequoia... they are admitting it's only worth $600 on average. And they're just doing this fakie transaction. The company is admitting it's only worth $600 on average because they're taking the money at a blend and cost of $600." — Rory O'Driscoll 00:38:45

"I don't believe there's right or wrong in money, there's just money. That's where at the minimum you have to look yourself in the mirror as the other investor and say, wow, that's the price of being cool." — Rory O'Driscoll 00:37:47

California's Billionaire Tax Will Net Negative — Hurting the Very People It Claims to Help

The counterintuitive argument: taxing billionaires aggressively in California doesn't just fail to collect revenue — it reduces the tax base enough that the marginal dollar cut from the budget will be social welfare programs, not teacher salaries.

"Somewhere in Sacramento, someone will zero out a line item on the budget. And let me give you a clue, it won't be payments to the teachers. It won't be payments to the firemen. It'll be marginal services to marginal people that your craft stupidity and desire to make a political point has ended up costing them money." — Rory O'Driscoll 00:01:01

3. Companies Identified

Anthropic

Leading AI lab, creator of Claude. Mentioned for achieving a $6 billion revenue run rate in a single 28-day month (February), surpassing Databricks' lifetime revenue. Also highlighted for the leaked Claude Mythos model — reportedly a 10-trillion parameter model focused on cybersecurity, and for smart enterprise GTM: pre-releasing Mythos to CISOs as a fear-based marketing strategy.

"They are leasing the Mythos model first to CISOs within companies. It's kind of like, oh, it's so scary, we're going to give you this model and give you time to figure out how to use it. Of course, part of that time will involve giving a million bucks to Anthropic. So it's just great marketing." — Rory O'Driscoll 00:27:15

Oura

Consumer health wearable company (smart ring). Mentioned for investment by Scale Venture Partners (via acquisition) and potential IPO, praised for defensible hardware-software consumer product with genuine technical moat.

"Unlike the other two guys, I've run a textile manufacturing company 30 years ago. The technology required to make an Allbirds or a shoe is not the same as the technology required to make a modular electronic device that sits on the human finger and measures blood." — Rory O'Driscoll 00:48:53

Whoop

Fitness and health wearable company. Announced $500M raise at $10B valuation. Mentioned as validation of Jason Lemkin's 2027 prediction for human healthcare data and longevity becoming a major investment theme — potentially arriving even earlier.

"It looks like it might even be 2026. And you know, the great thing about both stories is very defensible... these are fundamentally standalone products with a clear consumer value proposition." — Rory O'Driscoll 00:41:26

Checkout.com

Digital payments processor. Mentioned as sponsor but with substantive performance data: processed $300B+ in total volume in 2025, up 64% YoY, returned to full-year EBITDA profitability, serving 1,000+ enterprise merchants including 63 processing $1B+ annually (eBay, Amex, ASOS, Vinted, Temu).

"In 2025, Checkout.com processed over $300 billion in total volume, up 64% year over year, and returned to full year EBITDA profitability." — Harry Stebbings 00:02:44

CrowdStrike / Palo Alto Networks

Cybersecurity incumbents. Mentioned admiringly for their M&A playbook: proactively acquiring emerging security solutions before the threat category is fully defined. Highlighted as the smart strategic response to agentic AI security threats.

"They know damn fine that when a new threat emerges and a new solution emerges for that threat, when an earlier winner comes out, you better spend your 300 million bucks, your 500 million bucks and just swoop up the winner and add it to your product." — Rory O'Driscoll 00:26:46

4. People Identified

Dario Amodei

CEO and co-founder of Anthropic. Praised implicitly throughout for consistent leadership, clear strategic focus on safety and capability, and organizational stability versus the chaos at OpenAI.

"I have to think a company organized like that is just going to out-execute someone with that level of drama. I almost can't take it." — Jason Lemkin 00:17:03

Steve Jurvetson

Venture capitalist, early investor in SpaceX and Tesla. Praised for making a generational career bet on Elon Musk's companies and now realizing massive liquidity, recently relocating from California to Incline Village, Nevada.

"He made a brilliant call to align with SpaceX, been on the board of Tesla and SpaceX... he's put his money in a compounding machine. And now he's clearly hit the DPI moment." — Rory O'Driscoll 00:58:56

Massa (Masayoshi Son)

CEO of SoftBank. Mentioned for securing a $40B bridge loan to buy more OpenAI stock — framed as extraordinary risk tolerance. Noted he survived the 2002 Nasdaq crash (85% decline) which validates his high-leverage, high-conviction approach.

"You haven't lived till you've seen an 85% decline in an index. And obviously, if that happened or anything like it, you'd just be way underwater. It's a fairly high amount of leverage for an investment fund, to say the least." — Rory O'Driscoll 00:20:03

Matthew Prince

CEO of Cloudflare. Mentioned for a viral moment where he publicly said he didn't remember Ron Conway helping Cloudflare — framed as an honest, non-mean crystallization of the VC value-add debate.

"He literally just meant he couldn't remember getting any help from this beloved VC. And I think it just said so much to me about VCs adding value, but also VCs thinking they add value." — Jason Lemkin 00:04:13

5. Operating Insights

Conform Your Company to Your Customers — Not to VC Preferences on Revenue Structure

VCs historically pushed founders to convert variable/hardware/usage revenue into ARR contracts. The panelists now admit this was often wrong and that durable SaaS ARR is trading below S&P 500. The operating lesson: structure your pricing and revenue model around how customers want to buy, not around what your investors want to report.

"One of the things I hated about venture was when people would say, oh, make it all recurring revenue... you should conform your company around your customers and your model, not your VCs." — Rory O'Driscoll 00:47:36

"It used to matter. We'd be in board meetings where you would torture companies so that they would have more ARR and that they would have less variable revenue. I mean, that seems archaic today." — Jason Lemkin 00:47:20

Headline Valuation Obsession Is a Founder Trap That Backfires at the Next Round

Founders who accept tranched rounds to get a high headline valuation are implicitly agreeing they're worth less than the headline — and setting themselves up for a visible down round if growth doesn't materialize fast enough to justify it.

"If you accept that combined deal, you're implicitly saying, I know I'm only worth $600, but I'd like the optics of a billion. You better be damn sure that your next round you're at one and a half billion, otherwise you'll have the optics of a down round." — Rory O'Driscoll 00:37:19

The $0/Month Free Trial That Converts to Paid Is a Gray-Area ARR Inflation Tactic — Investors Are Noticing

If your PLG motion uses a $0 first month with auto-conversion to paid, you may be counting churned trials as ARR. Sophisticated investors are now flagging this. The operating insight: be precise about what you count as ARR before investors interrogate it — or you'll face credibility damage at scale.

"I'm pretty sure that means they recognize $240 in ARR that first month when you're paying zero... So is that, if I do a $0 a month product that's discounted as a marketing cost, and I churn after 30 days, does that count as $240 of ARR? I think for a lot of startups, it does." — Jason Lemkin 00:33:12

6. Overlooked Insights

Anthropic's CISO-First Go-to-Market for Mythos Is a Masterclass in Fear-Based Enterprise Sales

This was mentioned briefly but deserves much more attention. Anthropic is reportedly pre-releasing its most powerful, unreleased model exclusively to CISOs — framing it as: "here is the threat, now pay us to defend against it." This is a textbook fear-uncertainty-doubt enterprise sales motion, but executed at the frontier model level. It simultaneously creates urgency, captures the budget holder with the most authority, and positions Anthropic as both the threat vector AND the solution — before competitors even know the product exists.

"They are leasing the Mythos model first to CISOs within companies. It's kind of like, oh, it's so scary, we're going to give you this model and give you time to figure out how to use it. Of course, part of that time will involve giving a million bucks to Anthropic. So it's just great marketing... saying to the CISOs, you're going to have to figure this out. This is the new terrifying weapon we've invented. Please give us a million dollars and we'll let you defend yourself with it also." — Rory O'Driscoll 00:27:15

The implication: any AI security startup that can replicate this "we know the threat, we built the defense" positioning — especially one that can get ahead of model releases — has an extraordinary wedge into enterprise security budgets right now.

The Token Double/Triple Counting Problem Will Be the Next Major AI Valuation Scandal

Jason Lemkin briefly raised, and Rory quickly agreed, that the same tokens are being resold and re-recognized as ARR across multiple layers of the AI stack — foundation models, API resellers, and application wrappers all counting the same underlying compute as their own revenue. This wasn't explored deeply, but it's a structural accounting problem that will matter enormously when these companies try to go public or raise at disciplined multiples.

"Cursor is selling it again and recognizing the revenue, right? The same... people keep reselling these tokens again and again and recognizing them as their own ARR." — Jason Lemkin 00:30:57

"Everyone's got amazing revenue growth because it's the same little token going... If we all agree to have essentially 0% gross margins, an infinite number of us can keep reselling tokens to each other, can't we?" — Jason Lemkin 00:31:26

The investor implication: when underwriting any AI company's ARR, you need to stress-test gross margin and net revenue after pass-through costs to the foundation model — and ask whether their revenue would survive being reported net rather than gross.