20VC x SaaStr was back this week, this time live at SaaStr AI LDN!! A deep dive into Databricks’ $134B valuation, the death of per-seat pricing, and why 16% median growth is the new normal in public SaaS.
Top 10 Takeaways
- The TAM Trap is real. The majority of public SaaS companies are stuck—not because founders are idiots, but because we created so many companies that we saturated every adjacent market. When you needed to expand, there was already a venture-backed competitor there.
- Databricks at $134B (32x revenue) might actually be cheap. At 55% growth and reaccelerating at $4.1B in revenue, there’s literally no public comp. Palantir is the only public company growing 30%+, and it trades at 80x. When growth compounds and reaccelerates at scale, traditional valuation models break down.
- Seeds are for suckers (maybe). The risk-adjusted, time-adjusted certainty of putting money into Databricks at $134B or Anthropic at $180B may actually be better than a Series B with 7-10 year duration uncertainty.
- Per-seat pricing is an existential threat. Workday said it publicly. Microsoft says they’re past “peak employee” permanently. HubSpot is 2.8x more efficient than 2021. If everyone’s getting more ARR per employee, you’re running out of seats to sell.
- SaaS has become Japan. Great economy, but if everyone only has 0.9 kids, there’s only so many seats to go around. The average public SaaS company is growing 16%. That’s never happened before.
- Security is the revenge of the enterprise. Gainsight has been locked out of Salesforce for 2+ weeks. Drift is dead—permanently kicked off. When incumbents can use security as an excuse to cut off ecosystem apps and sell their own agents, that’s an existential threat to startups.
- The fastest AI companies have the lowest burn multiples. Even with high inference costs, revenue grows faster than costs. You can’t brute-force $100M ARR in 10 months with humans—you can’t hire fast enough, and they won’t all be great.
- Overpayment only works when TAM is huge. In finite TAMs, you’ve got to bid more tightly. The PagerDuty and Eventbrite acquisitions (2x and 1.5x revenue respectively) show what happens when you hit the wall.
- Model providers probably won’t compete in most verticals. OpenAI’s “code red” is essentially an admission they need to focus on their core mission. Coding is obvious competition. Wealth management AI is not.
- You don’t get 5 years anymore. Google launched their Lovable/Replit competitor in less than 10 months. Datadog just launched a PagerDuty competitor. If you blow up, competition comes in months, not years.
The Full Breakdown
OpenAI’s Code Red: From Offense to Defense
The Thrive-OpenAI partnership announcement feels like ancient history already. Within 24 hours, the narrative shifted completely: OpenAI declared a “code red” to focus on their core product.
Google did a code red on OpenAI three years ago. Now OpenAI is doing one back.
What does this mean? Healthcare agents—pushed back. Ads—pushed back. The consumer hardware device? That sound you hear is it slipping out of the room.
The message is clear: win the ChatGPT wars and you’re worth $2 trillion. Why fuss around with little vertical markets that can be worth a couple hundred million bucks?
For founders, this is actually good news. The model providers will compete in coding—that’s obvious. But they’re unlikely to expand into every vertical app. If you’re building AI for wealth management, tax planning, or specialized enterprise workflows, you’re probably not competing with OpenAI directly.
Databricks vs. Snowflake: The $134B Question
Here’s the most interesting valuation debate happening right now:
Snowflake: ~$4B revenue, 28% growth, profitable, valued at $80B (20x revenue)
Databricks: ~$4.1B revenue, 55% growth, unprofitable but reaccelerating, raising at $134B (32x revenue)
The fundamental venture question this poses: How much extra in multiple do you pay for how much extra in growth?
The math gets wild quickly. If that extra growth persists for 3-4 years, the premium is worth every dollar and then some. Compounding is relentless.
But here’s what breaks the model entirely: Databricks is reaccelerating at scale. That’s almost mathematically impossible to value. If you stipulate even going from 50% to 55% to 60% growth at this scale—that’s approaching infinite value because that’s what the math says.
We’ve seen this dynamic before. When Anthropic went through its burst of reacceleration this year, everyone realized their models were wrong. That’s why you saw the step function increase in valuation.
The uncomfortable truth: Databricks might be the second-best public company if it were public today. And at $60M post on a safe for a seed deal? Databricks honestly seems like a better risk-adjusted bet. That’s how broken early-stage economics have become.
The TAM Trap: Why Nobody Figured It Out
The median public SaaS company is growing 16%.
Let that sink in. We’ve never grown this slowly. Ever.
How did the Aaron Levies, the Drew Houstons, the Eric Yuans not figure this out? These are some of the smartest founders in tech history.
The answer isn’t that they’re idiots. The answer is there might not be an answer.
We made so many SaaS companies that we saturated every market. By the time you got to the point where you needed to expand beyond your initial market, there were already other venture-backed SaaS companies in every adjacent space. You just ran out of room.
Zoom is the perfect example. Everyone who needed a Zoom account has one. Everyone who doesn’t have a Zoom account has a Teams account (the poor bastards). The obvious next product was contact center—but they couldn’t get the acquisition done, and there were already incumbents.
The pre-AI playbook was broken. You ran out of time.
Per-Seat Pricing: The Existential Threat
Workday called it publicly: seat reductions are an existential threat.
Jeff Lawson (Twilio) said on this show: “We are unwaveringly going to see the movement away from seats.”
The data is brutal:
- HubSpot is 2.8x more efficient than 2021
- Salesforce is 2x more efficient
- Microsoft says they’re permanently past peak employee
If you’re a seat-based SaaS company, your total addressable market is literally shrinking. Not because companies are buying less software—they’re buying more. But they’re doing it with fewer humans.
Software prices on value delivered. When you couldn’t measure value, all you had was per-seat pricing. Then AWS pioneered usage-based pricing, which was inherently more rational for buyer and seller.
Now we’re moving to AI-based value pricing. But here’s the problem: it’s a lot harder to measure value than seats. You can count butts in seats pretty easily. Every login is a butt. When you’re trying to measure value delivered? That’s tricky.
Security: The Revenge of the Enterprise?
Two stories that should terrify every startup founder:
Drift: Security breach. 700 organizations’ data downloaded. A pirate group is asking for millions per instance in ransom. Drift was kicked off Salesforce. Permanently. Dead.
Gainsight: Locked out of Salesforce for 2+ weeks with no known resolution time.
One time you can blame the vendor. Two times, you might start locking down your platform. Three times? “I’m just going to own all the agents. I’m done with these risks.”
The cynical view: Large enterprise software companies could use security as an excuse to cut off ecosystem apps while conveniently having their own agent product ready to sell.
If you don’t want Glean slurping all your data… if you’re worried about where agents are depositing your data across a hundred different places… the incumbents might look a lot safer.
In board meetings, security in the age of AI has been “close to zero” in the above-the-fold part of the conversation. That might need to change.
The Three Categories of AI Companies (And Why None Need People)
Category 1: Large public companies Only growing 10-15%, so they better be kicking off cash or they’re in trouble. They’re optimizing ARR per employee and focused on free cash flow. Less employment.
Category 2: Model companies Taking huge amounts of capital, but it’s not for humans—it’s for Nvidia. No one’s telling OpenAI to be efficient. Small headcount relative to size. Just grow quickly.
Category 3: AI apps startups Because of foundation models, they’re shipping products and getting such traction that growth is ahead of their ability to hire. They literally can’t spend the money. Gamma ships the product, it’s amazing, people buy it with credit cards. By the time you hire a salesforce, you’re already kicking off cash.
The thing all three categories have in common: none of them need people.
Big companies can’t have people because they need efficiency. Model companies don’t need people because they just need geniuses and GPUs. AI app startups are going so fast they can’t hire people.
That’s… not great if you’re people. And it’s why the tech labor market is tough right now.
Google’s Vibe Coding Clone: Too Late or Too Early?
Google launched a Lovable/Replit competitor this week. No database. No OAuth. Just… a clone.
Sometimes that’s okay for big companies. But here’s what hasn’t changed in the age of AI: big companies only have so many priorities. They can introduce a lot of little tests, but keeping a big initiative going takes enormous energy.
The bigger concern for startups: you don’t get 5 years anymore. Google launched their competitor in less than 10 months. Datadog launched a PagerDuty competitor in the last 24 months—and PagerDuty was founded in 2008.
If you blow up, you should attract competition. But it’s not a free lunch.
The Lovable vs. Supabase Debate
At similar valuations (Supabase at $5B, Lovable at $6B):
The case for Lovable (from the growth investor): If vibe coding isn’t a category, both are screwed. But if it IS a category, Lovable is the front end and captures more value. Lovable gets $20/user and pays $2 to Supabase. If you’re in a highly risky category, the dumb bet is “if I win, I get a little, but if I lose, I lose 100%.” May as well be in for a pound.
The case for Supabase (from the durability perspective): Databases are a hard problem. You can only lose so much data. Five years of everyone using your database builds real lock-in. It ain’t easy to churn and leave your database. This was the year of “growth but nothing else.” But a little defensibility would be nice. Hard problems are reassuring.
Can AI Disrupt the Crappiest Products in the World?
Here’s what’s exciting about AI beyond coding:
Goldman Sachs and Morgan Stanley charge 1% of assets and do essentially nothing except give you loans (which are admittedly valuable—you can borrow against your Nvidia stock instead of selling and paying 50% in taxes).
But wealth managers will “help you with your trusts” (they refer you to someone who doesn’t call you back). They’ll “help with your taxes” (they tell you they’re not allowed to talk about taxes).
One founder took 11 months to set up three trusts. His celebrated Silicon Valley trust lawyer said: “Good news—most of my clients never even finish them.”
If AI can take large markets like wealth management that are utterly broken and actually fix them—do estate planning, taxes, retirement optimization, QSBS rollovers, all of it—you could build a $20-50 billion company.
The pitch isn’t for the ultra-wealthy—they have a million people who flatter them. It’s for the doctor, the dentist, the entrepreneur earning good money but whose affairs are modestly complex. They don’t want to screw up their Roth IRA withdrawal. They want to leave their house to the kids without estate tax problems.
Those are problems that are more complex than nothing but not where you can spend $20K on a lawyer. That’s a huge market.
The Growth vs. Efficiency Debate (Resolved?)
For founders hearing “I want growth, growth, growth” and also “I want efficiency, efficiency, efficiency”—here’s the reality:
In the fastest-growing AI companies, no one gives a rat’s ass about the bottom line.
The math works differently now. The fastest-growing AI companies, even with high inference costs, have the lowest burn multiples because revenue grows faster than costs.
You literally cannot brute-force $100M ARR in 10 months with humans. You can’t hire them fast enough. They won’t all be great. Maybe Larry Ellison or Marc Benioff could do it, but no one else.
If you’re doing $100M in 10 months, it’s because of massive inbound demand and a lot of AI. And if you’re doing that, no one cares how you got there.
But the 2020-2023 DNA of “I need 200% headcount growth to grow 100%” is still ricocheting around the ecosystem. It’s still in the majority of executives people talk to. That mindset is over.
The new expectation: grow 100% next year with 50% headcount growth. That’s healthy today.
Quotable Moments
Jason Lemkin
“Google did a code red on OpenAI three years ago. Now they’re doing a code red back.”
“The majority of public SaaS companies are in a TAM trap.”
“SaaS has become like Japan. It’s a great economy, but if everyone only has 0.9 kids, there’s only so many seats to go around.”
“In the fastest-growing companies that I’ve invested in, no one gives a rat’s ass about the bottom line.”
“When I go to a board meeting and a CMO says ‘I could do that, but I need 50 people’ or a product guy says ‘The reason we’re late is I need another 80 people’—I think it’s time to part ways. Give them a nice package and a good recommendation.”
“You literally cannot brute-force $100M ARR in 10 months with humans. You can’t hire them fast enough.”
“Seeds are for suckers. The risk-adjusted certainty that Databricks has a 3-5x from here—you could make a coherent case that it’s a better deal than a Series B with a 7-10 year duration.”
“I couldn’t imagine going through what’s happening at Gainsight. They’ve been locked out of Salesforce for two weeks. Drift is permanently dead. I think security is going to benefit the incumbents.”
Rory O’Driscoll
“How much extra in multiple do you pay for how much extra in growth? That’s fundamentally the question.”
“If that growth persists for three or four years, the premium is worth every dollar and then some. Compounding keeps going.”
“Reacceleration at scale is really hard. We see it maybe one in three companies for a single year. Only one in ten does it for two years. Very rarely do you see reacceleration when you’re already at 50%.”
“Overpayment only works when the TAM is huge. In finite TAMs, you’ve got to bid more tightly.”
“It’s not that everyone was idiots and couldn’t find the market. We made so many companies that we saturated the markets. By the time you needed to expand, there were other venture-backed SaaS companies in every adjacent market.”
“Snowflake and Databricks aren’t going to peacefully coexist. They’re going to struggle and fight against each other for the next 10 years, just like SAP and Oracle fought for the last 20.”
“There’s no such thing as blue collar venture. You’re not going to make it on value. You’re trying to choose on certainty of a big outcome.”
“All that matters is: can you build a big company here? That rule is so hard that there can’t be any other rules. No rules on stage, no rules on sector. Just big.”
Harry Stebbings
“PagerDuty: 2x a billion-dollar valuation at $500M ARR and 4% growth. Eventbrite acquired for 1.5x revenue with a 50% premium. This is a harsh new reality.”
“We’re playing a relevance game. Ramp raises four rounds in a year. Media matters more than ever. And some funds are saying ‘it’s slow compounding, coming soon.’ That’s a tougher game.”
“Has your view on market size changed? Every ops team on the planet uses PagerDuty and it’s a billion-dollar valuation. A 10% holding would be $100M back. Very sobering.”
“LPs are seduced by incredible follow-on investors, quick up rounds, and numbers. I’d rather be playing that game than the ‘it’s coming’ game.”
“Do you have to be picking slow compounders because you’re sitting in the Valley at Series B against Andreessen and Founders Fund and Sequoia and you can’t beat them?”
The Opportunity Remains Massive
The TAM Trap is real, and most SaaS companies already fell into it. Per-seat pricing is dying. Security might be the ultimate weapon incumbents use to fight back against the startup ecosystem.
But if you’re building AI that actually solves hard problems—not thin wrappers, but genuine disruption of broken markets—the opportunity is massive.
The rules have changed:
- Growth still matters most, but you can’t brute-force it with humans anymore
- Five years of runway is now ten months of competition
- Model providers probably won’t compete with you in specialized verticals
- The companies that can prove value delivered (not just seats occupied) will win
And maybe, just maybe, if you can take the friction out of wealth management, healthcare, compliance, or any of the other massively broken markets in the world—you can build something that compounds for decades.
Just don’t spend all the money until you prove it.
