With Harry Stebbings, Jason Lemkin, and Rory O’Driscoll
This was the week private M&A broke every historical ceiling we had. Cursor sold to SpaceX/xAI for $60 billion, three years after founding, with a $10 billion break clause. That’s nearly double Wiz ($32B) and almost 4x WhatsApp ($16B) in nominal terms. And it happened in a quarter of the time.
The bigger story: when your stock trades at 100x revenues, you can buy anything trading at 10x all day long. Elon just proved it. Seven other companies are now sitting on market caps above $2 trillion, watching closely. The Overton window for what’s “doable” in enterprise M&A just expanded violently.
Meanwhile, Anthropic turned down $800 billion funding offers and crossed a trillion on secondary. Tim Cook announced his retirement from Apple. Claude launched a full design application that will slowly maim Figma, Canva, and Adobe without replacing any of them outright. Rippling hit $1B ARR growing 78% and accelerating, which should permanently retire the “SaaS is dead” narrative for anyone still spreading it.
And underneath all of it, the real 2027 battle is taking shape: agent fabric. Not orchestration. Not evals. Agent fabric. The layer that manages what 100+ autonomous agents are doing inside your company, in real time, with governance and security a CIO can actually sign off on. Salesforce is making its bet. Whoever wins this wins the next decade of enterprise software.
Top Takeaways
1. Cursor at $60B Is the Biggest Private Venture M&A Deal in History. Full Stop.
The numbers are absurd. $60 billion, three years from founding, with senior engineers already moving over to xAI before the deal closes. It’s structured as an option: close at $60B in six months post-IPO, or pay $10B break fee. Either way, it’s the largest privately-held venture acquisition ever announced, beating Wiz ($32B) by nearly 2x.
The deal makes industrial sense for both sides. Cursor had an exploding business with shitty gross margins because it needed its own model and compute. xAI has reasonably good models, the Colossus data center, and almost no revenue. As Jason put it, it’s a marriage made in heaven. Cursor’s estimated to finish the year at $6B in revenue, making this a ~10x revenue multiple deal, which, by the standards of 2026, is almost cheap.
Why now? Because this is the perfect moment to sell. The founders have only been at it three years. Michael Truell hasn’t aged eight years the way Gary Tan has. You don’t get hit by founder fatigue until year four or five. As Jason noted, this is the time to sell, not after another tour of duty that doesn’t need to happen.
2. The Real Unlock: 100x Stock Lets You Buy 10x Stock All Day Long
Here’s what most people missed. As Rory framed it precisely: if SpaceX IPOs at $2 trillion on roughly $20B in revenue, that’s a 100x multiple. When your stock trades at 100x, you can buy things trading at 10-15x revenue all day long. A $60B check is roughly 3% of SpaceX’s alleged market cap in return for 15-20% of their total combined revenues.
“If your stock is valued at 100 times revenues, you can buy things that are trading at 10 or 15 times revenue all fucking day long,” Rory said. That arbitrage doesn’t last. Which is exactly why you do these deals now.
This was the only game in town for Cursor. The number of buyers who can write a $60B check for a break-even-gross-margin business can be counted on one hand. Most of them couldn’t do it for DOJ reasons. Only Elon, founder-led, with a 100x stock and no one to answer to, could pull this off.
3. Jason’s Prediction: A $100B Private M&A Deal in the Next 12 Months
Rory disagreed. He thinks Cursor/xAI stands as the high watermark of private M&A for a decade. Jason pushed back hard: “There will be a $100 billion deal in the next 12 months.”
The math is simpler than it looks. There are now seven companies with market caps above $2 trillion: Nvidia, Apple, Meta, Amazon, Alphabet, Microsoft, and soon SpaceX. Any of them can write a $100B check at roughly 5% of their market cap to not fall behind in AI. Jason pointed out this is exactly what happens inside these boardrooms. He’s sat in them. Zuck is already thinking: bring me candidates that move the needle. For 5% of market cap or less, write the check.
The analog is Benioff trying to buy LinkedIn for $30B years ago and failing. Today, Benioff would pay $100B for the right AI asset to change the face of Salesforce. He would do it tonight. Every one of these CEOs would. The capital exists. The appetite exists. It’s just about finding the right target.
Rory’s counter is fair: even if SpaceX trades at a trillion (50x revenue), and Cursor gets to $6B, SpaceX is still buying at 10x with stock at 50x. Those arbitrages are real but finite. Everyone else trades sub-10. So to write a $100B check, you’d need to buy something worth $10B a year. That narrows the field to Stripe-sized outcomes. Which, as Jason noted, could literally happen tomorrow.
4. Anthropic Crosses a Trillion on Secondary. Every Dollar Wants Its Home Here.
A European LP who manages money for the largest families in Europe told Jason this week: “All of our families just want one thing, Anthropic. We can’t give them anything else.”
That’s the mood. Anthropic turned down $800 billion funding offers. Secondary markets are pricing them at a trillion. The implicit market belief: they’ve won the enterprise race, surpassed OpenAI, and have the next leg of growth ahead of them.
Rory’s call: they should go public as fast as humanly possible. The biggest private round in history ($122B for OpenAI) is now bigger than the biggest IPO in history ($75B for SpaceX). That’s bizarre. Given the capital requirements of the frontier model race (compute, compute, compute), public markets are the only place with the liquidity to support $200B+ raises over time.
Jason’s counter: Figma IPO’d perfectly and is now down 83% from peak. What if Anthropic IPOs at a trillion, trades up to $2T, then takes a hit when the zeitgeist shifts? Rory’s answer is the right one: if that happens, Anthropic’s competitors who stayed private are existentially screwed. You always want to be the public company when capital markets contract. Always.
Prediction from Rory: Anthropic goes public Q4. The IPO will be fabulously successful at current pricing. Whether it holds six months later is a different question. But you raise a huge slug at peak FOMO, and then you’ve got a decade of capital access your private competitors can’t match.
5. Tim Cook Steps Down: Operationally Perfect Exit
Cook took Apple from roughly $350B to $4 trillion market cap over 15 years. The stock is 20x what it was when he took over (buybacks juice this number). He retired at 65, with an internal successor in John Ternus ready, and the stock moved less than 0.5% on the news.
That’s a master class in operationally excellent exit. No Shantanu Narayen-style surprise. No Netflix-style uncertainty. A clean transition the board had been planning for years.
The uncomfortable subtext: Cook, Narayen, Hastings. These pre-AI CEOs are leaving in waves. Apple probably doesn’t have existential AI problems the way Adobe does. Netflix has media problems more than AI problems. But the pattern is hard to ignore. Running a pre-AI company in 2026 requires a level of intensity most 60-year-old CEOs aren’t signing up for.
Jason’s line on this cuts deep: most humans aren’t up for 996-12-12-8. The AI-native CEOs respond to Slack in 60 seconds. They know what’s launching next month before it launches. The older CEOs say “we’ll catch up in the next release” with complete confidence as they head off to their triathlon training. That confidence gap is the real story.
6. Claude Design Is an Application. Not a Feature. And That Matters.
Most of the X commentary on Claude Design missed the point. Yes, it won’t replace Figma for real designers. Yes, it won’t kill Illustrator. Yes, Canva still owns the print and physical design markets. All true and all beside the point.
What Anthropic shipped is an application. Not a prompt. Not a skill. Not a workflow. A full application with sharing, user hierarchy, asset saving, export to Canva, integration with Claude Code. This is the first time Anthropic has built a proper tier-1 application on top of Claude, and it signals something bigger: the frontier labs are going to build applications that compete with the application layer, not just sell tokens.
The maiming thesis matters here. Claude Design won’t kill Figma next quarter. You won’t see it in Figma’s numbers. But over 4-6-8 quarters, it takes 20-30% of the use cases away. Product teams don’t want to wait 30 days for a designer to turn around a Figma file. Engineering teams are already working in Claude Code. When design, product, and engineering can all live in one tool, the integrated workflow wins. That’s the whole ballgame.
Jason’s take: “No one’s saying it’s pixel perfect. I don’t have time. I don’t have time. I don’t have time.” That’s the competitive reality. If human designers have time to redesign it later and make it better, great. In the meantime, ship.
7. Rippling at $1B Growing 78% Kills the “SaaS Is Dead” Meme Dead
Rippling hit $1B ARR growing 78% year over year. They’re accelerating. They were at less than $500M eleven months ago. That’s not decelerating growth. That’s not mature-category decay. That’s a B2B company posting numbers that would have been extraordinary in any era of software.
This exposes the real issue with the whole “SaaS is dead” narrative. Low-growth SaaS is bad. High-growth SaaS is still amazing. The market doesn’t hate your business model, it hates your growth rate. If you’re doing $1B at 78% and accelerating in a category full of people saying you can’t grow anymore, the market will absolutely reward you.
Rory’s framing: this is a win against the run of play. Everyone’s walking around saying the world is dead. Rippling put up 78% growth and acceleration in a “dead” category. Parker Conrad is a candidate for CEO of the year. Driving that level of acceleration at scale without a massive AI tailwind to ride is almost incomprehensible.
The bigger lesson for B2B founders: DAU, WAU and MAU metrics now matter more than revenue. Is your usage growing faster than your revenue? If yes, you’re winning. If no, you’re hiding stealth churn behind contracts that haven’t come up for renewal yet. Jason is seeing this everywhere. Users stealth churn off Netflix because YouTube has better AI-generated Star Wars content. Users stealth churn off Marketo because the API is broken. Users stealth churn off OpenAI because Claude works better. The contracts lag by months or quarters. The usage doesn’t.
8. The Real 2027 Battle: Agent Fabric
Salesforce’s headless announcement confused everyone. People thought it was new. It wasn’t. Salesforce has been API-first since 2006, when Benioff opened up the platform nobody thought should be opened. As Jason noted, Salesforce’s APIs are the best APIs in enterprise software by a large margin. Better than the new guys. Better than everyone. That’s 20 years of compounding work.
What Benioff is actually pitching, according to Jason, is something much bigger than headless. It’s agent fabric. The layer that sits above all your agents (yours, theirs, third-party) and provides context, guardrails, security, governance, and real-time visibility into what 100+ autonomous agents are doing across your company.
This is not orchestration. Orchestration is a dashboard on top of a few APIs. Agent fabric knows every data operation, every decision, every downstream action, in real time, across hundreds of parallel agents with sub-agents running 24/7. If you’re the CIO and an agent goes rogue, you lose your job. You need a trusted fabric. You cannot get this out of ChatGPT or Base44 or whichever YC startup claims to have orchestration.
The CIO question in 2027 will be: “Who can I trust to manage my agents?” Salesforce is well-positioned. So is Microsoft. So is Datadog. So are Databricks and Snowflake. The ones who move fast win. The ones who don’t lose their install base to new entrants who built agent fabric from day one.
The most tragic outcome, per Jason: you have 280,000 HubSpot customers. HubSpot launches an AEO agent. It’s a dud. You should have bought Peec.ai or one of the other 250 AEO startups. Instead you built something mediocre in-house because it was cheaper. And now your install base is getting picked off by the AEO-native vendors. This is the 2026-2027 tragedy in slow motion.
9. Cerebras Files for IPO: 10-Year Venture Journey Pays Off
$510M revenue in 2025 growing 76%. Signed deals with OpenAI and AWS in the last three months. Moved from niche Middle East contracts to mainstream AI infrastructure. This is what it looks like when a hard semiconductor bet works.
Rory’s take: great venture story, deserves to go public well, and in a market where the leading player (Nvidia) is worth $5 trillion, being the only real standalone alternative chip play is worth a lot. As a call option on 1-2% of a $5 trillion market, the upside math is enormous. 1% of Nvidia is $50 billion.
Andrew Feldman earns real respect for this one. Ten years of work. Credible IPO. Hope they crush it.
10. 91% of AI Unicorns Are in the Bay Area
Harry’s read from London: the best AI minds want to be in Silicon Valley, and rightly so. But at 91%, the marginal advantage of being in the Bay Area equals the marginal advantage of being anywhere else. That’s the equilibrium point.
The practical reality: recruiting and maintaining a team in the Bay is nearly impossible without egregious amounts of capital. Competition for talent is at levels you simply don’t see in London or Europe. Harry would rather be a top-three brand in a less-saturated geography than fighting Benchmark, Founders Fund, and Andreessen for every allocation in every round.
Rory’s framing: during that narrow 2022-2024 window when knowledge about what was actually possible with AI was tribal and only shared in Bay Area hacker houses, geography was everything. That window is closing. In five years, knowledge will be widely distributed. But the current Cambrian explosion is real and location-specific.
Quotable Moments
Jason Lemkin
“It’s like, you got these guys called Cursor walking down the street saying, hey, good news, we got an exploding business, a couple of billion dollars in ARR, but bad news, we have shitty gross margins because we need our own model and we need compute. Elon goes, hold that thought. This is a guy walking around with a whole bunch of compute, a reasonably good model, and literally no revenue. It’s like a marriage made in heaven.”
“There’ll be a $100 billion deal in the next 12 months. Which one it is, I don’t know, but you have to. There’ll be a hundred billion dollar deal.”
“This is what old software looks like. As if Anthropic and OpenAI are coming for old software. This is an application. This is not a prompt. We won’t want to use grandpa’s software anymore.”
Rory O’Driscoll
“If your stock is valued at 100 times revenues, you can buy things that are trading at 10 or 15 times revenue all fucking day long. As long as SpaceX is worth a trillion, then Elon can chuck $6 billion on the table. That’s 3% of SpaceX’s alleged market cap in return for 15 to maybe 20% of their total revenues.”
“This will stand as the high watermark of private M&A for a decade.”
“I think rippling is a win against the run of play. You’re playing the SaaS game where everyone’s walking around saying the world is dead. You put 78% growth and acceleration up in a category where most ill-informed people were saying you can’t do that.”
Harry Stebbings
“One LP said their families just want one thing: Anthropic. And the challenge we have is they don’t want anything else, but they all want Anthropic. Every dollar wants its home in Anthropic right now.”
“The majority of great AI companies are in the Bay Area, but I’d rather be one of the top three brands in Europe with much less supply-side than there fighting against Benchmark and Founders Fund and Andreessen and everyone in between.”
“The prize for winning at this level is you have to reinvent yourself every quarter. The AI-native CEOs are already all over what’s launching a month later. The older CEOs say ‘we’ll catch up in the next release’ as they go off to complete their triathlon.”
This post is part of the ongoing 20VC x SaaStr collaboration with Harry Stebbings and Rory O’Driscoll.
