With Harry Stebbings, Jason Lemkin, Rory O’Driscoll, and special guest Mike Cannon-Brookes (CEO and co-founder Atlassian)
We’re back on 20VC, with Harry, Rory from Scale, SaaStr’s Jason Lemkin, and a very special guest — Mike Cannon-Brookes, CEOand co-founder of Atlassian. On Anthropic’s $149B revenue projection and whether CIO budgets can actually support it, why “software is dead” is a stupid statement, the input-constrained vs. output-constrained framework that explains everything about AI’s impact on B2B, Harvey’s $11B raise and the TAM question nobody can answer, the Super Bowl ad wars and what they tell us about the top of the cycle, and why the best founders are working harder than ever — and the ones who aren’t should step aside.
We also did a SaaStr deep dive just on Mike’s points here.
Top Takeaways
1. Anthropic’s $149B Projection Has a Revenue Stacking Problem Nobody’s Talking About
Anthropic projects $149B in ARR by 2029. OpenAI projects $180B. That’s ~$350B between two companies in a $700B global software market — and that’s before you account for Microsoft’s $200B.
But Mike pointed out what almost nobody discusses: the revenue stacking makes these numbers misleading. When Atlassian spends money on Anthropic, they actually pay AWS, and AWS pays Anthropic. When Cursor does $1B in revenue, a big chunk of that is the same billion as Anthropic’s revenue flowing through the stack. “The individual revenue number is not necessarily counting the whole stack,” Mike said.
Even with stacking, the numbers are staggering. As Rory framed it: you’re basically saying Anthropic becomes another Microsoft and OpenAI becomes another Microsoft — that’s two new Microsofts that are “pretty voracious mouths to feed out of the same pile of cash.” Either you believe in massive TAM expansion, or the math gets brutal for everybody else.
The trillion-dollar consulting and services market may be the relief valve. AI could eat into the systems integration spend around SAP, Oracle, and Netsuite — that’s a massive slug of budget that could get reallocated to more efficient software. But even the consulting market isn’t simple: implementation consulting for AI itself is booming (Accenture, etc.), while rote integration work gets automated. Swings and roundabouts.
2. “Software Is Dead” Is a Stupid Statement — But the Composition Problem Is Real
Mike was unequivocal: the idea that software as a category is dead is “ludicrous.” Businesses have always bought pre-built technology solutions. They didn’t write everything in assembly before, and they won’t build everything from scratch with LLMs.
He pulled up Atlassian’s competitive docs from 2005, 2010, and 2015. A huge chunk of those companies are gone — merged, acquired, dead. That’s just how capitalism works. AI accelerates the cycle but doesn’t change the fundamental pattern.
The deeper problem, as Rory identified, is compositional. For 15 years, median public SaaS growth held around 30%. But that wasn’t because every company grew at 30% — it was because when one slowed below 10%, PE took it private, and a new 60% grower IPO’d to replace it at the top.
That cycle broke. No new high-growth IPOs in years. PE hoovered up the mid-tier. Big tech got better at M&A. Mike’s SaaS CEO group from 2020 had 10+ public CEOs. Now it’s him, Eric from Zoom, Aaron from Box, and Toby from Shopify. Almost everyone else got bought.
What’s left in the public index is a weird survivor set — too big for PE, too small for big tech, no new entrants. The average looks terrible, but it’s a composition problem, not a death spiral.
3. Product & Engineering Is the Island of Stability — Everything Else Is at Existential Risk
Jason’s thesis, validated by Mike’s data: every category outside of engineering and product is at existential risk of shrinking seats.
Workday said it publicly — seat headwinds because Fortune 500 companies aren’t hiring like they used to. Pave data shows no category has been more decimated in hiring than customer support. Meanwhile, nobody is cutting engineering teams. Replit is at $300M in revenue with 300 people and 11 in go-to-market.
Mike provided the best framework for understanding this divide: input-constrained vs. output-constrained functions.
Input-constrained functions (customer support, legal, HR) have work bounded by external demand. 100 customers ask 100 questions. Make the team more efficient and you need fewer people. Output-constrained functions (engineering, product) have a roadmap that’s never finished. Make engineers more productive and they just build more — the headcount doesn’t shrink, the output explodes.
Atlassian is the proof case. Their cloud revenue is growing 26% (accelerating), RPO is up 44% (accelerating three quarters in a row), and they just launched a customer service product. The service collection business alone could go public. Nobody’s talking about shrinking the engineering tools business.
But Jason pushed further: the Jevons Paradox could flip even input-constrained categories. Legal work that wasn’t worth taking to your lawyer — the 4-day sales contract review, the too-minor-to-bill question — could explode in volume if AI makes it instant and cheap. Same with customer support: SaaStr ran an experiment with Agentforce on leads not worth a human’s time and got 5-6x more inputs immediately.
4. Harvey at $11B Is Either Brilliant or the Most Consensus Trade in Venture History
Harvey raised $200M at $11B. The company is at $190M ARR with 100K active users, growing 300% year-over-year, supposedly heading toward $600M by year end.
The bull case: $200B per year is spent on non-partner lawyers in the US. Harvey at $2K per active user today is tool-replacement revenue. If that becomes $50-100K per lawyer by eating actual associate work, the TAM is almost unlimited. The growth rate speaks for itself.
The bear case: $2K per lawyer across 400-500K addressable big law lawyers is a $1-1.2B marketplace at current pricing. That’s a great business, not an $11B one. And at 50x run rate, you need to believe in 300-400% compounding for three or four years.
Mike’s Atlassian anecdote was telling. When Excel invested $60M in 2010, they modeled Jira growing 30% then declining to 15% by year three. Instead it did 50%, 60%, 70%. They expected a 2-3x. They made 100x. But Excel paid 10-12x forward revenue for a profitable company. Harvey is at 50x run rate. The math for outperformance is just harder.
Rory’s framing tied it together: this is the “you can’t get fired for buying IBM” trade. Of newly minted unicorns in Q1 of last year, 40% already had one or more up rounds by Q4. Capital piles into consensus winners at any price because the downside is a 1x with a liquidation preference and a nice logo on your wall. It’s classic top-of-cycle behavior — and it’s also rational given how hard it is to pick winners at the seed stage.
5. Customer Support Is a Great Category That’s Almost Impossible to Invest In
Harry asked the question a lot of VCs are asking: with 14 companies raising over $100M in AI customer support in the last two years, plus incumbents like ServiceNow, Atlassian, Salesforce, and Zendesk — is this actually a good place to deploy capital?
Mike’s answer was nuanced: Atlassian’s service collection business is massive and growing faster than the overall company. IT service management, HR service management, financial service management, employee service management — these are all distinct submarkets, not one mega-market. And it’s the category where the most AI agents are being deployed across Atlassian’s customer base.
The real insight: the most interesting AI in service isn’t the cost-saving play. It’s the action-taking play. When an agent doesn’t just answer “yes, you can expense that lunch” but actually files the expense claim, resets the password, submits the parental leave application — that’s a fundamentally new TAM that didn’t exist before.
But Jason and Rory agreed on the investment challenge: the competitive dynamics are brutal. Sierra at the high end, Decagon executing well, Mike building a service empire, Salesforce with ServiceCloud, Zendesk’s $300M+ AI support revenue. You never want to find yourself third in a subsegment. The risk-adjusted, opportunity-cost-adjusted answer for a VC is probably to pay up for the consensus winner rather than back a challenger.
6. The Super Bowl Ads Were Classic Top-of-Cycle — But There’s a Deeper Point
Anthropic ran ads mocking OpenAI for having ads. OpenAI ran a “you can build things” spot. Sam and OpenAI’s CMO fired back on Twitter. Most of the 325 million people watching went “what the hell is this? Give me the Doritos commercial.”
Jason made the sharpest observation: of all the things Anthropic could have advertised — how Claude revolutionized coding, product creation, the engineer’s best friend — they chose to advertise the one area where they’re weakest: consumer. “It’s advertising for its worst product.”
Rory’s read: this was two companies exchanging signals to each other and each other’s employees using $5 million of Super Bowl money. Anthropic reinforcing “we’re the enterprise good guys” to the 10,000 engineers they want to hire. Plus their head of safety quit that week.
Mike saw it through a capital allocation lens: when you’re burning $1B a week and people keep funding you, the ads are cheap. There’s no marginal cost calculation happening. That will normalize — it always does — but right now the AI companies are maximalist because they can be, while public SaaS companies face “pesky public market investors” asking about profitability.
Harry pushed back — Wix has bought six Super Bowl ads over the years and their CMO swears by the ROI. Mike agreed: for some businesses like Wix, it’s a totally sensible investment. For others, it’s the CEO wanting a Super Bowl ad because they have $200M burning a hole in their pocket. Both happen every Super Bowl.
7. Public Companies Have an Unfair Constraint — But It Makes Them Better
The deeper question Mike surfaced: private AI companies operate with no marginal costing, uncapped SBC, and no profitability pressure. Public SaaS companies have to deliver quarterly results, manage stock-based compensation scrutiny, and grow simultaneously. Same game, different rules.
Mike’s answer was counterintuitive: “We’re a better company because we’re a public company. We’re better at forecasting. We’re better at planning. We’re better at executing.”
The trap is when financial discipline replaces strategy instead of complementing it. You have to do both: deliver results and invest massively in AI. If you only focus on financial execution, you won’t have a business in 5 years. If you only focus on long-term bets, you won’t have investors in 5 quarters.
Atlassian spends a “huge amount” on fundamental new R&D around AI. You can’t see it in any public company’s accounts — it’s impossible to tell from the outside where R&D dollars are being allocated. But Mike would guarantee most public companies are moving more and more toward AI. The ones that aren’t? “You’re probably in trouble.”
8. The CEO Gut-Check: Would You Choose This Job Again Today?
Mike’s co-founder Scott Farquhar retired 18 months ago after 23 years. Mike starts at 5 AM every day now. He’s working harder than ever.
He’s been doing “SaaS therapy” with a lot of CEOs and founders. His advice:
First, accept reality. Stop pontificating about whether AI changes things. It already has. “Part of creation is destruction. You got to go build value for customers.”
Second, ask honestly: would you choose this job again today? Not every minute — but over a 90-day average. If no, that’s fine. Hand it off thoughtfully. That takes more courage than hanging on.
Third, maintain balance. Spend time with your kids. Go for a hike. “If you spend 100% of your time at work, you make worse decisions.” Mike’s wife tells him flat out: “Stop. You’re going to go insane and then you’re going to become irrational and then you’re going to make bad decisions. We’re going for a hike.”
Jason added color: he talked to a CEO who quit at year-end at hundreds of millions of revenue. The goodbye message was “I was told AI wasn’t important in our space.” Not everyone has it in them to push through the next wave. And that’s okay — but then get out of the way.
Software Isn’t Dead, Atlassian is Accelerating — But Most Traditional B2B Companies May Not Make It
The “software is dead” narrative is lazy, but the underlying reality is more nuanced than either side admits.
The facts: Atlassian is accelerating. Harvey is tripling. AI support companies are exploding. Engineering tooling is booming. TAM is expanding in ways nobody’s spreadsheet predicted.
Also facts: most public SaaS companies are still decelerating. Seat-based businesses outside of engineering are getting compressed. The revenue stacking between model providers, cloud platforms, and application layer makes everyone’s projections suspect. And consensus venture trades at 50x run rate require heroic assumptions to generate returns.
The framework that matters: input-constrained vs. output-constrained. If AI makes an input-constrained function more efficient, you need fewer humans. If it makes an output-constrained function more productive, you just build more. The Jevons Paradox could flip some categories — but don’t count on it for your business model.
For founders: you have to be good. Not “pivot to AI” good. Actually, fundamentally, deliver-more-value-than-the-alternatives good. If you’ve spent the last year with 50,000 engineers and your growth is in the teens heading toward 10% — the clock is ticking. You had Claude 3.5 seven months ago. How long do you get? 12 months? 18? Show us the money.
And if you don’t enjoy it anymore? Congratulations on what you built. How’s your number two?
Quotable Moments
Mike Cannon-Brookes
“The idea that software as a category is dead is ludicrous to me. It’s very efficient for businesses to buy pre-canned solutions of technology. They didn’t write everything in assembly before and they probably still won’t.”
“We’re a better company because we’re a public company. We’re better at forecasting. We’re better at planning. We’re better at executing.”
“If you’re not enjoying it, no harm, no foul. Go do something you do enjoy. It’s tough out there. It’s going to be hard. Welcome to the technology industry.”
Jason Lemkin
“Every category that I know of outside of engineering and product is at existential risk of shrinking seats.”
“We are in a renaissance of software. We are building so much software. It is unbelievable.”
“Our agent closed a $100K deal the other day on a Saturday night. Doesn’t complain. These agents, they don’t want family lives.”
Rory O’Driscoll
“If Anthropic goes from $4.5 billion to $150 billion in a world where Microsoft does $200 billion, you’re basically saying Anthropic is another Microsoft. OpenAI is another Microsoft. That’s two more Microsofts which are pretty voracious mouths to feed out of the same pile of cash.”
“Every time we get one of these new technology disruptions, everyone gets very frothy and yells and screams a lot. A handful of winners are made. A lot of money is lost. And then everyone acts like it’s never happened before.”
“It’s the venture capital equivalent of you can’t get fired for buying IBM. You can’t get fired for sticking money in with a 1x preference on the consensus winner.”
Harry Stebbings
“Customer support is one I just cannot get my head around. When you look at the sheer number of players funded to huge extremes — there are now 14 companies in the last two years that raised over $100 million.”
“I think they were brilliant uses of capital. Both Super Bowl ads.”
“Should I just call it quits and sell my Monday and my Duo and just accept defeat? Like, they ain’t coming back? Or like — they coming back?”
This post is part of the ongoing 20VC x SaaStr collaboration. With special guest Mike Cannon-Brookes, co-founder of Atlassian ($50B+ market cap, $5B+ revenue, 350,000+ customers, 10,000 people in R&D). Atlassian just posted 26% cloud revenue growth (accelerating), 44% RPO growth (accelerating three consecutive quarters), and expanding gross margins.
And want to learn how to WIN in the AI Era in B2B? Join 10,000 of us at SaaStr AI 2026 May 12-14. We’ll give you the playbooks to win in 2026.
