Jason Lemkin | 20VC Summary

Participants: Harry Stebbings, Jason Lemkin, Rory O'Driscoll

1. Key Themes

The Agent Layer Will Determine AI Market Share — Not Human Preference

The shift from human-selected tools to agent-selected tools is the defining transition of the next 18 months. This changes competitive dynamics entirely — brand loyalty and UX advantages that drove Claude's ascent become irrelevant when autonomous agents optimize for performance metrics.

"More and more, the agent is going to choose what models and just what vendors we use... The agents are going to make the decision on everything." — Jason Lemkin 00:07:08

"The public markets think vibe coding and Claude are their threat. No, the threat is what the agents pick." — Jason Lemkin 00:11:56

"If OpenAI wins the agent wars, then you have lock-in. The OpenAI will probably pick OpenAI as the API." — Jason Lemkin 00:12:48

What's striking: Jason revealed he actually monitors which LLMs his internal AI agents prefer — and uses that as an investment signal. His agents prefer OpenAI, which shifted his own conviction back to "Team Sam."

The Compute Forecasting Problem Is Catastrophically Underappreciated

The capital risk embedded in foundation model companies is not just about magnitude — it's about having to bet 4-5x current run rate revenue in CapEx, two years in advance, on speculative demand curves. This is unlike any prior tech business model.

"If you're going to add $90 billion in revenue capacity, someone between you and your partners has to find plus or minus $300 billion to buy chips, dig holes in the ground, build data centers... You're doing $10 billion in run rate and you're effectively saying between you and your partners to be able to meet demand two years from now, you've got to invest $300 billion." — Rory O'Driscoll 00:23:11

"If you get it wrong and you end up doing $200 billion in run rate revenue, you're going to only have half the compute you need. And if you only get to $50 billion in revenue two years from now, you're going to be left with $150 billion of stranded capacity. In software land, it was so easy." — Rory O'Driscoll 00:23:34

"Compute and a shitty model also equals no revenue. See Grok for details." — Rory O'Driscoll 00:26:08

The PE Software Buyout Model Is Structurally Broken

The Medallia collapse is not a one-off — it signals an entire vintage of 2021 leveraged software buyouts facing existential pressure. The combination of AI disruption, vendor consolidation, and debt service costs means even modestly leveraged deals can implode when terminal value evaporates.

"They didn't way over-lever it, they just way overpaid for it. You can service 2 billion plus of debt on a 1 billion low growth company with a pre-AI story that has to transform to AI. You simply can't." — Rory O'Driscoll 00:44:22

"Companies at risk: Coupa, New Relic, Anaplan, Zendesk, Avalara, Smartsheet... They all look like they may not be able to fully repay their debt." — Jason Lemkin 00:45:54

"There goes one of our exit routes. That's the biggest impact." — Rory O'Driscoll 00:50:03

2. Contrarian Perspectives

Anthropic's Compute Shortage Was Actually a Success Problem, Not a Failure

Conventional narrative frames Anthropic as struggling. The reality: they over-succeeded on model quality and were simply hit by their own demand growth faster than any plausible business plan could have anticipated.

"Anthropic did it the exact opposite way — they got the model perfect. In fact, they may have over-succeeded. And as a result, they were light on compute... No one had a business plan last year when they went from one to nine [billion] that said they're going to go to 30 by the end of Q1. So they were hit by their own success." — Rory O'Driscoll 00:21:49

Google Is Actually the Biggest Winner of the Entire AI War

While the narrative focuses on OpenAI vs. Anthropic, Google wins regardless of which LLM wins — they own compute infrastructure, have massive existing capacity to rotate, have Gemini, hold a large stake in Anthropic, and generate cash flows to sustain all of it.

"Google wins whether you use Gemini or whether you use Anthropic now. Two, Google has infinite capacity... They have the surplus that they can route between their customers and themselves and others. They win-win here. They have Gemini, they have Amazon, and they have the capacity, and they have the ability to rotate it when they want. And they have the cash flow." — Jason Lemkin 00:30:02

"Risk-adjusted, I would do Google... it has multiple ways to win. It can win if AI adopts fast. It can win if AI adopts slow." — Rory O'Driscoll 00:32:53

Long-Term Enterprise Contracts Are a Form of Deferred Churn, Not Safety

The conventional wisdom treats multi-year enterprise deals (Workday 8-year contracts, ServiceNow, etc.) as durable revenue moats. The contrarian view: they simply defer the inevitable and mask terminal value destruction from investors and operators.

"Churn that is deferred still exists. It is where the rent to CEO and the mediocre hide... If you believe that public stock prices are the sum of terminal values of cash flows and profits, then deferring churn, masking churn doesn't matter." — Jason Lemkin 00:13:15

"If all you are is a system of record for humans, you are a bounded cash, even if you're not a negative NPV... The only way to get the high price that you need to make these stocks compelling is to have agent-based activity on your platform." — Rory O'Driscoll 00:14:50

The American Sports Team Investment Thesis Has a Hidden Structural Advantage That's Actually Anti-Capitalist

US sports teams have outperformed as investments not because of superior economics, but because the NFL, NBA, and MLB are government-protected monopolies — the only businesses with explicit antitrust exemptions, meaning no penalties for poor performance and no competition.

"The US, in fact, has been even more successful at creating sports money printing machines than even in Europe. The NFL and all the American things have been constructed, partly because they're the only three businesses that have an exemption from antitrust. So they're all constructed as nasty little oligopolies where there's no penalty for failure — which is the definition of socialism." — Rory O'Driscoll 00:12:48 01:12:55

Stripe Is the Best-Positioned Company for the Agent Economy — Rated #1 by the Agents Themselves

Jason ran an agentic API grader evaluating 120 APIs using Claude, OpenAI, and Gemini simultaneously. Stripe received the only A+ rating. This is a non-obvious, data-driven signal about which infrastructure companies the AI agent layer will preferentially adopt.

"I built an agentic API grader where we went, I just had Claude, OpenAI, and Gemini together take the top 120 APIs and grade which ones they thought were the best... Stripe got the highest grade, got the only A+, which is a reason to go along in Stripe. I did not think Stripe would come out on top." — Jason Lemkin 00:10:01

3. Companies Identified

Stripe

Description: Global payments infrastructure company
Why mentioned: Received the only A+ in Jason Lemkin's agentic API grader — rated by Claude, OpenAI, and Gemini simultaneously across 120 APIs. Strong signal for agent-era infrastructure dominance.
Quote: "Stripe got the highest grade, got the only A+, which is a reason to go along in Stripe. I did not think Stripe would come out on top." — Jason Lemkin 00:10:01

Twilio

Description: Cloud communications platform
Why mentioned: Explicitly cited as a company that has outperformed because agents have use for it — contrasted against Atlassian/Monday which agents don't need
Quote: "The ones that actually have outperformed, right? The Twilios, the Cloudflares, and others — the agents still have use for it." — Jason Lemkin 00:12:21

Cloudflare

Description: Network security and CDN infrastructure
Why mentioned: Same rationale as Twilio — positioned as infrastructure the agent layer will actively consume
Quote: "The ones that actually have outperformed, right? The Twilios, the Cloudflares, and others — the agents still have use for it." — Jason Lemkin 00:12:21

Rippling

Description: HR/finance/IT platform, ~$1B ARR growing ~70%
Why mentioned: Called out as a great business that will be priced rationally at fair value — not at AI pixie dust premium, but still a strong compounder
Quote: "I think Rippling is a great story... it's going to trade on a sensible, adjusted PE multiple based on growth, based on cash flows, entirely rationally." — Rory O'Driscoll 00:19:41

Unwrap.ai

Description: AI-native customer sentiment analysis company (Scale portfolio company)
Why mentioned: Cited by Rory O'Driscoll as an example of a next-generation AI-first product that is displacing legacy players like Medallia in the customer feedback space
Quote: "We have an investment in Unwrap. It's a small company that has customer analysis of customer sentiment. There's a whole bunch of way better AI first products in this space." — Rory O'Driscoll 00:45:15

Canva

Description: Design platform, expected IPO candidate
Why mentioned: Mixed signal — will have a successful IPO and strong consumer/prosumer base, but agents won't use it, placing it in the "middle bucket" of terminal value
Quote: "An AI agent is not going to go in and move assets around. It's just going to create the assets. So an agent doesn't need Canva." — Jason Lemkin 00:16:06

4. People Identified

Gary Tan (Y Combinator President)

Description: President of YC, the world's most prominent seed accelerator
Why mentioned: Published guidelines on ARR definition standards for YC companies, framed by Rory as both genuinely useful and strategically shrewd market-making — protecting YC's brand integrity
Quote: "Gary's like, no, man, be truthful and precise about your revenue... I think it was simultaneously really good and really shrewd... Y Combinator has a dominant market share in the seed market, 25%. It erodes the value of their product if a whole bunch of people start thinking the numbers are bullshit." — Rory O'Driscoll 00:06:06 01:06:06

Brad Gerstner (Altimeter Capital)

Description: Founder/CEO of Altimeter Capital, crossover tech investor
Why mentioned: Cited as an exemplar of disciplined public market investing — specifically the practice of having a predetermined exit price on every position
Quote: "I remember talking to Brad from Altimeter. You can tell there's a guy very dialed into every position has an exit price. And it's a discipline that you need as a public investor." — Rory O'Driscoll 01:21:13

Ryan Smith (Qualtrics/Utah Jazz)

Description: Founder of Qualtrics, owner of Utah Jazz
Why mentioned: Cited as a compelling real-world case study for sports team investment returns — used ~$1B from Qualtrics sale to acquire Jazz, which has since 4x'd
Quote: "When Ryan Smith sold Qualtrics, I think he made about a billion dollars after 20 years or so. And I believe that billion, most of it went back into the Jazz and it has quadrupled." — Jason Lemkin 01:11:26

5. Operating Insights

Agent Preference as a Vendor Selection Signal

Jason Lemkin has operationalized a novel procurement methodology: he monitors which tools and APIs his AI agents naturally gravitate toward, and uses that as a real-time signal for vendor selection and investment conviction. This is actionable today for any company deploying multi-agent workflows.

"One of the many reasons I've come back to Team Sam and Team OpenAI is not because I care, because my agents like OpenAI. They love it. So I got to follow my agents. Just like you got to back your team of humans in the old days, like 2024. Today, I have to back my team of agents." — Jason Lemkin 00:08:56

Three-Bucket Framework for Evaluating Software Company Terminal Value

Rory provides a clean, actionable framework for quickly sorting any software business: (1) melting iceberg — AI erodes their core, terminal value negative, especially dangerous with leverage; (2) system of record — humans keep using it but minimal agent activity, positive but bounded terminal value; (3) agent-amplified — agents actively use and compound the platform, high multiple justified. This framework is immediately applicable to portfolio reviews or due diligence.

"If you put those three buckets: melting iceberg, you're in trouble, you have a low stock price, and if you've leveraged, you're dead. Then the middle category is system of record... And then the happy outcome is the agents are using you when you're getting increasing returns from AI leveraging your technology." — Rory O'Driscoll 00:17:42

When a Top-Tier Investor Doesn't Lean In for Super Pro Rata, Exit Immediately

Jason flagged a non-obvious but highly reliable signal for portfolio management: if the lead investor with capital to deploy doesn't fight for super pro rata in a growth round, it's a strong negative signal on internal conviction — regardless of what they say publicly.

"When you do the growth round, when you do the billion dollar run, and you don't see the big fund lean in for the super pro rata, I just think it's a terrible sign in today's world... if they've got the billions to deploy, they're going to put it into your winners. And if they don't stick you in the side of your chest with an elbow to get super pro rata, it's a bad sign." — Jason Lemkin 01:19:42

6. Overlooked Insights

Vendor Consolidation Is a Massive, Underreported Parallel AI Story That's Killing Mid-Tier SaaS Right Now

Buried in the Medallia discussion was a data point that deserves its own conversation: Gartner data shows 30-50% of dollars flowing into AI budgets are coming from software vendor consolidation — not new spend. This means the headwinds for legacy SaaS aren't just about AI displacement in the future, they're happening now, today, through budget reallocation. The companies most at risk aren't the ones with bad products — they're the ones that are "nice to have" versus "must have" in the eyes of a CIO looking to fund AI initiatives.

"The amount of vendor consolidation that's occurring at the same time as AI growth is under-discussed. Whether you look at Gartner's numbers, 30% to 50% of AI dollars are coming from consolidation. Medallia is a top target. Do we really need that half million dollar a year dated survey product? Did we really learn that much from it, guys? No. So it gets cut before you cut your workday or Salesforce." — Jason Lemkin 00:46:23

The Founder Key-Handoff Trend: A Coming Wave of Informal M&A That Creates Real Opportunity

Mentioned almost in passing, Jason flagged an emerging micro-trend that could become the defining liquidity mechanism for mid-market SaaS over the next 2-3 years: founders at $20-100M ARR with no IPO path, no PE interest, and no strategic acquirer simply handing their companies to a stronger operator/founder peer in exchange for equity. This is informal, shadow M&A — invisible to traditional deal flow channels — and creates a significant opportunity for operators or holding company structures willing to consolidate subscale but profitable SaaS assets.

"I'm giving the keys to Rory. And I'm going to give a third of my company... I don't see any exit. And the founder gets out... I think we're going to see this happen all the time. Is founders giving the keys to their friends. Not completely quitting — I mean, it's just I'm at 40 million, 50 million, 20 million, 100 million. I'm not going to get there, guys." — Jason Lemkin 00:55:54

"There's going to be some guy who's a mid-career operator who's willing to take the pain, going to say, I got this. I'll take these five software companies all broadly speaking in the systems management space. We'll put them together. I'll run them like a hard ass. We'll get to 20% growth, 30% EBITDA... This is a chance for me to make 50 million bucks as the CEO." — Rory O'Driscoll 00:57:19