With Harry Stebbings, Jason Lemkin, and Rory O’Driscoll
Anthropic raises $30 billion at a $380 billion post-money. Thrive closes a $10 billion fund the same week. OpenClaw’s creator joins OpenAI. Workday’s founder comes back. And public B2B software keeps getting dragged into a gravity well it can’t escape.
This week’s episode was dense. Here’s what you need to understand.
Bottom Line Up Front
Jason Lemkin: “Corporate America is going to will AI into existence. They’ve decided. You can debate whether the ROI is real, whether the labor savings are there — but it doesn’t matter, because by the end of this year, the Fortune 500 will have decided they are replacing humans with AI whether it’s right or wrong. It’s doable. And once that train leaves the station, the growth numbers will be jaw-dropping.”
Rory O’Driscoll: “You’ve never seen a company grow 10x in GAAP revenue run-rate year-on-year for three years. Not early Microsoft. Not early Google. It simply hasn’t happened at scale. So yes, you’re leaning into the singularity here. The short-term reality is that narrative and momentum dominate valuation. And right now, the only narrative is AI replaces everything. That’s the tape. You don’t fight it.”
Harry Stebbings: “Wall Street fell in love with AI and had to fall out of love with SaaS to do it. The burden of proof has completely flipped. SaaS companies used to get the presumption of success. Now they get the presumption of failure. It’ll equilibriate eventually — price clears all markets — but we are not there yet. And the canary in the coal mine for everything is whether enterprises actually get the ROI they think they’re paying for.”
Anthropic at $380B: The Only Play in Venture?
Let’s start with the number: $30 billion raised, $380 billion post-money. Originally a $10 billion round that tripled in size as demand overwhelmed supply.
Jason was blunt about what this means for the rest of venture: “This is just, you know, the same week Thrive raises $10 billion. This is what everybody wants to do. This is the only play in venture. When public software stocks are down 20-something percent this year, this is the play. It is that simple.”
And mathematically, it’s hard to argue. The last several Anthropic rounds have worked quickly. The $60 billion round earlier in 2025 and the $160 billion round before that — both turned into 2x or better in short order. At that rate of value creation, multi-stage funds almost have no choice but to participate. You hold out, you’re not showing up in the only thing that’s working.
Rory made the point about the underlying business justifying it: “The sound bite is going to be three years of 10x revenue growth. I went back and plowed through early Microsoft, early Google, early Compaq. It’s unprecedented. You’ve never seen a company grow 10x in GAAP revenue and run rate year-on-year for three years. At this scale, it just hasn’t happened.”
The caveat — and it’s a real one — is that Anthropic is still losing money. Early Microsoft and Google were wildly profitable during their high-growth years. And Dario himself has said publicly: if they misspend on compute for a year, they could go bankrupt. That’s a $380 billion company where the CEO openly discusses insolvency risk. The market is, for now, choosing not to price that in.
What this means for everyone else: If you’re a multi-stage fund and you don’t have Anthropic exposure, you are explaining to your LPs why you missed the only thing that worked. That’s a brutal conversation. Which is why people Jason didn’t even know were on the periphery of this investment shocked him with their participation.
The B2B Software Gravity Well
Here’s the metaphor that landed hardest in this episode: gravity.
Everything in software right now is being pulled toward Jupiter-level gravity. $6B ARR growing 24% and accelerating isn’t enough for Atlassian — you need to be Databricks growing 60%. Even 70% growth (AppLoving) isn’t good enough anymore. Non, down to a $2 billion market cap. The companies that can’t achieve escape velocity don’t just plateau — they get pulled down.
Meanwhile, Anthropic, OpenAI, Databricks — they’ve left the planet. Claude Code going from $1 billion to $2.5 billion in ARR between the start of the year and today? That’s escape velocity. Everything else is subject to the well.
Jason’s portfolio view was sobering: “In my last fund, two companies are massively benefiting from the current world in B2B. One I’d call a push — benefiting but also being hurt. Everything else is struggling because of AI. All the rest of that fund. And it was so similar in 2023. When you see this on the first of every month when you get investor updates, it’s hard to be optimistic in 2026 about anything not already getting a boost.”
The honest question Rory put on the table: where’s the forcing function that breaks this dynamic? What snaps the spell? The answer is probably enterprise ROI — if companies spend $20-30 billion on AI in the next two years and don’t see the labor savings they expected, you’d get a pullback. But here’s the thing: Jason doesn’t think that pullback is coming, because the decision has already been made.
Corporate America Has Already Decided
This was the most important insight in the episode, and it’s worth sitting with.
Jason’s argument isn’t that AI will deliver ROI. His argument is that it doesn’t matter, because corporate America has already decided to bet on it — and they’re going to will it into existence regardless.
“AI does not miraculously replace 10,000 employees in an hour. It doesn’t work that way. But you make a bet — just like we’ve always made in enterprise software. And then you also decide: I just don’t want a bigger company. I want to do layoffs. I want to be leaner. I want to do it. And so I’m going to lean into this bet.”
He pointed to the Klarna CEO as an example — going from 6,000 to 3,000 employees, now talking about getting to 2,000. Trying to be provocative, sure. But Jason thinks that’s how every CEO he talks to is actually thinking. Not because AI is proven, but because it’s doable, and they want to be smaller.
Rory’s framing: “Corporate America has decided they’re going to make this bet. The Zeitgeist is making this bet. It’s unstoppable now. Capital has decided. People are going to try and make this bet — which is very different than saying when they made the bet, they liked the result. But that’s two or three years out.”
The script: we’re looking at one to two years of mega AI budget, on the presumption of success, not proof of ROI. That’s exactly what happened with hyperscalers two years ago — they said we’re doing this and we’re not blinking. Corporate America is now in that same moment.
The implication: anyone selling AI labor replacement to enterprises is going to have an extraordinary 12-24 months. Anyone selling seats to humans who may not be hired? Structurally challenged, regardless of what the current numbers say.
Old School B2B SaaS: A Necessary Distinction
Rory made a definitional point worth keeping: “SaaS is dead” is a useless statement unless you define which SaaS.
There are three separate categories:
The models themselves — Claude, GPT-4, Gemini. These are technically SaaS (software sold on a monthly basis). Nobody says they’re dying.
New school B2B — Harvey, Sierra, the “wrappers” (or standalone companies, depending on your view). Built post-2022, native to AI. These are what everyone wants to invest in.
Old school B2B — everything built pre-2022. This is what’s dying, or at least heavily impaired. The question isn’t if but when and how fast.
For private old-school B2B companies at $8-20M ARR growing 80-100% — Jason is seeing a real and painful realization set in. “They’ve raised their last venture dollars and they’re going to have a good business, but they are moving off the venture train because they can’t align to what is needed today to make that next round work.”
That’s the quiet crisis that isn’t making headlines. It’s not zero growth. It’s good-enough growth that no longer qualifies for venture math, and founders are starting to accept it.
Figma: A Cautionary Tale of Epic Growth But Missing AI Budget
Figma got substantial airtime as the illustration of what can happen even to the best companies.
The indictment: product prototyping for enterprise product teams is Figma’s core use case. It’s where the majority of their large-customer revenue comes from. Replit and Lovable — companies that didn’t even really work well until September/October 2024 — have taken $300-400M of ARR that Figma should own.
“Replit and Lovable shouldn’t have gotten here, guys. We own this space. We let it go.” That’s the honest conversation that should be happening inside Figma, Jason argued.
For the record: Figma is still growing 30-40%, trading at roughly 10x revenue, and is an objectively great company. But Figma Make was a failure by any honest assessment. They missed an adjacent category that was right in front of them — product prototyping with AI — and now two startups own it.
The broader point: if Figma — the best-run, most technically excellent design company in the world — could miss this, “what hope is there for mere mortals?” The window for moving on adjacent categories is weeks, not quarters.
Rory’s framing was more forgiving: Figma’s space was particularly AI-adjacent. Design and code are close neighbors, and that adjacency made them unusually vulnerable. Not every B2B company sits that close to what AI models do natively. Accounting software, for example, still needs debits and credits and a good UI — that’s 80% of the value independent of AI.
The AI SDR That Booked a Deal on Day One
Jason shared a real example that stopped the conversation cold.
SaaStr recently deployed Monaco — a new AI SDR — and it booked a six-figure meeting on its first day. Completely autonomously. Identified the target (a leading hyperscaler with budget for a major SaaStr sponsorship), loaded them up, reached out to the right VP, and booked the meeting. No human required.
None of SaaStr’s previous AI agents could quite do this. Not before February 2026. They could send personalized emails. They could do sophisticated Marketo/HubSpot-style outreach. But autonomously reaching out to a busy VP at a tech company and getting the meeting entirely booked and ready to go? That happened for the first time in February 2025.
“The products weren’t this good 60 days ago. That’s just progress in 60 days. What will it be like at the end of the year? It could completely close it.”
Harry’s pushback was fair: the SDR market is moving fast but not uniformly. Coding tools (Claude Code at $2.5B ARR) are moving at a completely different speed than sales tools. Not all markets are disrupting at the same pace.
But Jason’s broader point: the demand is so overwhelming that Monaco has demos booked through summer despite barely launching. And SaaStr would switch agents again if something better came out next month — because in this market, everyone is promiscuous. Switching costs are near zero, and the products are improving so fast that loyalty is functionally irrational.
The OpenClaw Moment: Guardrails Are Gone
Peter Steinberger of OpenClaw joins OpenAI. The community loses its mind.
The setup: OpenClaw was designed to break guardrails. Let agents run on your C drive, access your desktop, run pseudo-autonomously 24/7. Everything the major labs chose not to do, because they believed it was dangerous.
When OpenClaw (and Mulitbook) first launched, they immediately exposed users’ private keys, passwords, and sensitive data. Anthropic threatened to sue. The reaction made complete sense: this is exactly why we have a safety team.
And then every single developer thought it was the coolest thing they’d ever seen.
The lesson: “There’s no reason Anthropic or OpenAI couldn’t have built this last year. They just didn’t think it was safe. But it’s too late now.” The horse has left the stables on autonomous agents. The demand from the developer community is irreversible. The question isn’t whether truly autonomous agents will exist — they already do. The question is what governance and security infrastructure gets built around them.
Manis cloning it in one night proves the technology itself isn’t differentiated. You can build this in a day. The guardrails are what take real engineering and institutional commitment.
The emerging opportunity: as Jason noted, anyone building agentic security infrastructure — guardrails for the autonomous agent era — is sitting in front of massive demand. Every CISO at every enterprise is about to have this land on their desk, and none of them have a framework for it.
Stripe vs. Adyen: A $90 Billion Valuation Gap Explained
Stripe at $140 billion private. Adyen at roughly $45-50 billion public. What explains the gap?
Rory broke it down: Stripe does ~$5B in revenue, Adyen does ~$2B. So multiply Adyen by 2.5x right away. Adyen is wildly profitable (nearly 50% operating margins), growth has slowed. Stripe’s profitability is less clear, growth trajectory is better.
“It’s the eternal venture question: how much extra revenue multiple do you pay for how many extra points of growth? If Adyen’s growing at 15-20% and you pay 10x, how much extra do you pay for something growing at 25-30%?”
But there’s another factor: communication. Adyen barely talks to the market. They reveal little about their cash position, their AI strategy, how they’ll deploy capital. Stripe, despite being private, tells a brilliant strategic narrative constantly. The valuation gap is partially a communication gap.
Rory’s conclusion: he’d buy Adyen over Stripe at current valuations. The probability it’s mispriced to the upside is lower — you have the data, you can value it. With Stripe at $140B private, you’re buying on incomplete information.
Jason’s counter: the flexibility of staying private is worth a premium. “The agility you have from being private in today’s world — you don’t have to make the trade-off perfectly. Mike Cannon Brooks at Atlassian said it: I have to be more profitable and massively invest in AI at the same time. Stripe doesn’t have to make that trade-off publicly.”
His prediction: end of year, we’ll see Anthropic, OpenAI, and SpaceX all diving for the public markets. Companies that need massive capex can’t stay private forever — they’ve sucked up all available private capital and now need public markets.
Workday’s Founder Returns: The Age of the Founder CEO
Aneel Bhusri is back at Workday. Eight months after stepping back, he’s solo CEO again.
The panel’s interpretation: this isn’t really about go-to-market. It’s about product roadmap. Workday faces an existential product rebuild for the AI era — and only a founder has the institutional memory to make those decisions quickly.
Rory’s crisp framing: “What you don’t need in this situation is generic business skills. What you need is massively specific knowledge, skills, and courage to make changes you know you have to make. That’s where the founder advantage is.”
The broader pattern: boards are dragging founders back now, across the industry. Daniel Dines at UiPath. Aneel at Workday. The resistance to change at these companies is enormous. Regular VPs want to do the 2023 version of their jobs. The talent that actually wants to embrace AI is leaving for hot AI companies where the product almost sells itself.
And that’s the feedback loop Jason sees playing out in his portfolio: “The smart folks are quitting these public companies and going to the hottest AI companies where the 2022 toolkit works perfectly. Because when your customers have already decided they want to buy AI products — you can sell, you can grow like crazy, your stock can go up. And if you’re not in those things, it’s just hard.”
Key Takeaways
- The “presumption of success” has completely transferred from B2B software to AI. B2B companies now operate under the presumption of failure. This isn’t permanent — price clears all markets — but the forcing function that reverses it is ROI proof, which is 2-3 years away.
- Corporate America has decided to replace humans with AI, whether the math works or not. This is the most important structural force in enterprise software right now. It’s not a question of if — it’s already decided. The only question is whether companies will be patient enough to actually execute it.
- The fastest AI adoption is happening in previously underserved categories. Legal and healthcare — markets that had bad SaaS alternatives — are going all-in on AI faster than categories with entrenched SaaS tooling. The new model doesn’t have to beat the old one; it just has to serve where the old one failed.
- Adjacent category windows are closing in weeks, not quarters. Figma had years to build what Replit and Lovable built. They had months. They probably had weeks after the models got good. If you’re a founder watching an adjacent category open up, the clock is already running.
- Autonomous agents are here. The guardrail infrastructure isn’t. Every enterprise is about to face this problem: powerful, cheap agents that can do real damage if deployed without security infrastructure. CISOs don’t have a playbook yet. That’s an enormous opportunity.
- The next 18-24 months of AI spending are locked in. Even if ROI doesn’t materialize, even if skeptics are right — the budgets are already committed. Enterprises are not reversing course mid-cycle. Which means the next two years will be extraordinary for companies selling AI, and increasingly difficult for everyone else.
Most Quotable Moments
Jason Lemkin:
“Corporate America is going to will AI into existence. They want it to be true. And with AI, you can build almost anything you want now. So even if this were a little less revolutionary, I think you can make a decision — and by the end of this year, that train will be so far out of the station that the growth numbers will be jaw-dropping.”
“I’m going to be the guy saying we’re never going to remove our systems of record. But as a cohort, public software stocks are approaching almost 10% growth annualized. If AI just reduces that another 30%, we’re in the dead zone. Not 5% growth where you can still argue. We’re in the area in the deep dark water where nothing grows.”
“The products weren’t this good 60 days ago. It booked us a six-figure deal on day one. Completely on its own. That is the pace of progress. And I really wonder if anything is safe from disruption — including the agent vendors themselves, because we will switch.”
Rory O’Driscoll:
“You’ve never seen a company grow 10x in GAAP revenue run-rate year-on-year for three years. I went back and plowed through early Microsoft, early Google, early Compaq. It’s unprecedented. At this scale, it just hasn’t happened. So you’re leaning into the singularity here.”
“In the short term we’re narrative and momentum creatures. All that’s happened is the narrative and momentum has shifted to AI is going to do everything. The people left trying to prove otherwise are left proving a negative. That’s a losing hand short-term. Smart traders don’t fight the tape.”
“What you don’t need in a product rebuild situation is generic business skills. What you need is massively specific knowledge and courage to make changes you know you have to make. That’s uniquely the founder’s advantage — they have the memory of why the original decisions were made.”
Harry Stebbings:
“Wall Street fell in love with AI and had to fall out of love with SaaS to do it. The burden of proof is the other way now. And there’s going to be a price at which these things stop declining — markets work — but finding that price is a grind, not a gift.”
“Claude is SaaS. Harvey is SaaS. Salesforce is SaaS. When people say SaaS is dead, they really mean old-school SaaS built pre-2022. And the toughest category within that is workflow-horizontal software without critical mass that isn’t public. That’s probably the hardest place to be right now.”
“Before 2022, you didn’t emphasize the AI because customers would get scared and slow down the sales cycle. Now, if you don’t have an AI strategy as a corporate leader, you’re a buffoon. That’s the single biggest behavioral shift in enterprise buying we’ve ever seen.”
Catch the full episode on 20VC and SaaStr. New episode every week.
