Discover more from benn.substack
The series E(verything is fine)
Keeping the dream alive. And how VCs can help.
In the fundraising fever dream of 2021, data startups got overvalued in three, sometimes compounding, ways.
The first was the most obvious. In the wake of the pandemic and the policy responses to it, every company—public, private, tech, blue chip, bleeding-edge AI companies, hundred-year-old conglomerates that manufacture tape—was overvalued. From 2010 through the end of 2019, the price-to-earnings ratio of the S&P 500, which measures how highly companies are valued relative to how much money they’re making, bounced between 13 and 24. In 2020, it peaked at 38, almost double its prior averages.
Moreover, if 2020 and 2021 were outliers for the market, tech companies were outliers within the outlier. Public SaaS companies historically trade at valuations five to ten times higher than their annual revenues; in 2021, the average multiple tripled to over thirty.1 The market’s animal spirits quickly made their way down Sand Hill Road: By late 2021, median startup valuations were two to three times higher than they’d been in 2020. If you walked into a venture capitalist’s office with a startup making $1 million and growing at 100 percent a year, the baseline valuation was about 50x its revenue, or $50 million dollars.
But the inflation didn’t stop there. Some companies got an additional bonus—a multiplier on their multiple—for being, in effect, corporate influencers. Companies with large Slack user groups or big Twitter followings got a 5x community multiple on their revenue multiple; popular open-source projects with lots of Github stars got a 3x Hacker News buzz multiple.2 Though startups like Facebook and Instagram popularized selling adoption metrics as a leading indicators of revenue growth, VCs have typically paid for product traction—daily active users, installs, time in the app. In 2021, investors were willing to pay premiums for looser, more atmospheric, more gameable, and less monetizable signs of progress.
Finally, data companies got yet another boost—a third compounding multiplier—by positioning themselves as the pioneers of some new industry category. At a conference in late 2021, Martin Casado, an investor at Andreessen Horowitz, predicted that the data industry would be a trillion dollar market, and that, “Workday, Salesforce, Adobe—they’re going to be reimplemented as apps on top of the data layer.” This was a common sentiment then: Data technology was the next open frontier of computing, full of green fields and blue oceans, and it was time for a land grab. Investors rewarded companies that could claim to be leaders in some up-and-corner of Matt Turck’s market diagram—in cataloging, observability, reverse ETL, orchestration, code-free data transformation, all-in-one aggregators, contracts, basically every “fourth” category—with another bonus.
Nobody played this game better than Airbyte. In 2021, Airbyte raised $150 million at a $1.5 billion valuation on less than $1 million in revenue, booking, famously, a 1,500x revenue multiple. They published the deck they used to raise that round—and it was the perfect bait for investors’ irrational exuberance.
First, the timing: They said their “initial plan was to raise the Series-B at the end of 2022” but decided to “test the waters in November [of 2021].” Truly great call. Second, the category multiplier: On the first slide of their pitch deck, they described a market on the cusp. “Data infrastructure is huge and growing, but still immature.” Reverse ETL, they said, was only two years old, current solutions only cover a small fraction of the market’s needs, and Airbye “will soon expand to all data movement use cases.” Third, the community multiplier: Before saying a word about revenue, they showed charts of active accounts, installs, how much data they were processing, and—on what they called “possibly one of the most important slides”—Slack community members, Github stars, and open source contributors.
Multiply it all together—the default 50x 2021 multiple, a 5x category multiplier, a 5x community multiplier—and Airbyte’s $1 million in revenue is worth $1.25 billion. Throw in some FOMO and the round number bump, and it starts to almost make sense.
Or, it did. In 2021, when interest rates were zero, the Nasdaq was at an all-time high, and Peloton was worth more than Ford, these effects compounded up. Today, when interest rates are over five percent, benchmark revenue multiples are down sixty to seventy percent, and Peloton is worth less than WD-40, the numbers compound down. Open-core business models aren’t engines for building passionate user communities, but liabilities that are hard to monetize. New categories aren’t billion-dollar opportunities, but unproven markets that no buyer has any budget for.3 Github stars aren’t cool anymore;4 even Matt Turck is over Matt Turck’s market diagram.
The compression math is brutal. The baseline multiple that VCs would now apply to Airbyte, for example, is probably down from 50x to 15x. The community and category multiples are, generously, 2x instead of 5x. Add it up, and the 1,500x company that’s worth $1.5 billion becomes the 60x company that’s worth $60 million.5
The bad news is that it would be a disaster for Airbyte to be worth $60 million.6 The good news is that Airbyte isn’t worth $60 million; it’s worth $1.5 billion. Until Airbyte and its many contemporaries who also raised at the top of the market in 2021 raise money again, the fair market value of their shares are unknown.7 Uninformed idiots can write wildly speculative blog posts about them, but that’s it. Someone has to open the box for us to know if the cat—the startup, the opportunity, the magic, the dream of changing the world—is still alive.
That’s tough, though. For 2023 Airbyte to be worth the same as 2021 Airbyte, 2023 Airbyte needs to be making about $25 million a year. Getting to $25 million from $1 million would require growing by 300 percent for three straight years. Though Airbyte has surely grown, they probably haven’t grown at 300 percent a year for three straight years since 2021 because that’s very hard to do in good markets, because this is not a good market, and because 2021 was two years ago. Unless a company is profitable or close to it,8 they’ll have to open the box eventually.
Another option is to open the box, but not tell anyone what’s inside? Anecdotally, I’ve seen a handful of companies that had previously raised large rounds with publicly disclosed valuations raise new, vaguely-named “venture rounds” that have no letter (e.g., Series A, Series B) and no price. These are probably down rounds, but ones that companies can plausibly deny as such. We didn’t need the money; we wanted to double-down on our groundbreaking AI investments to keep ahead of the competition.9 It wasn’t about the price; it was getting strategic investors on the cap table. The leaked reports of the cat’s death are greatly exaggerated.
The best option, probably, is to raise money opportunistically. Private companies aren't valued by the aggregated opinion of the market, but by a single highest bidder. If a startup can convince one fund that, sure, the magic community and category multipliers are dead, but the magic AI multiplier is higher than ever, it can potentially raise money at its prior valuation. The check might be smaller; the investor might be worse; there could be a lot of structure on the term sheet. But it sends a signal to the market—to customers, to competitors, and to people who write tabloid gossip—that, despite the bad market, the company is the same as it ever was: A once-in-a-lifetime opportunity.
Databricks recently closed a round exactly like this. In the fall of 2021, Databricks raised $1.6 billion at a $38 billion valuation. Almost exactly two years later, in September of this year, they closed a $500 million round at a $43 billion valuation.
Did Databricks need the money? Did they want T. Rowe Price, Nvidia, and Capital One on the cap table? Did they raise it because now’s the moment invest in generative AI, and they wanted to amp it up? I have no idea. It could all be corporate financial maneuvering. But it’d be hard to believe the valuation didn’t have something to do with it. By closing the round, Databricks can put a loud end to the endless speculation: The 2023 market has spoken, the discounts aren’t necessary, and Databricks is as valuable as ever.
I doubt they’ll be the last to do it. Today, everyone assumes everything is overvalued and every cat is dead. Surely, there are a number of other companies quietly canvasing the market—”we’re not raising now, but would be excited to bring on the right strategic partner”— looking for that one investor that can give them the multiple they need to prove otherwise. In the coming months, I suspect there will be more rounds that are as much about messaging as they are money: Though the room may be on fire, the dream is still possible, the cat is alive, and everything is fine.
Good VC, Bad VC, Useful VC
Every four years, I entertain the possibility of working for a political campaign. There’s some structural appeal in the job to me. Major campaigns are like extremely fast-growing startups, raising hundreds of millions of dollars and hiring thousands of people in a matter of months. But unlike startups, campaigns are built to expire. You aren’t chasing an uncertain IPO or acquisition some unknown distance over the horizon; instead, you know, on the Wednesday after the Tuesday after the first Monday in November, it’s over. You spend all of your money; you find out if you won or lost; you go home.
So, every four years, I email a handful of people who are working on campaigns and ask if there are ways I, an egomaniacal tech person, can help. And every four years, they tell me the same thing: No. Working on a campaign isn’t a drive-by job; it requires real expertise; if you show up once a week for a few hours, you’ll just get in the way.10 There are lots of part-time jobs on campaigns—knocking on doors, making phone calls, the real work—and if you want to help, do that. But we have no use for people who are looking to parachute in with context-free opinions about something they partially understand so that they can feel good about themselves.
Speaking of venture capitalists, I recently came across an old Christoph Janz post about what makes them good and bad. In it, he warns VCs about volunteering on the campaign:
A good VC is aware that there is a huge information gap between founders and VCs with respect to the founder’s business. He understands that the founder has thousands of hours of experience in his industry and with his customers and intimately knows the people on his team, whereas the VC’s knowledge of the startup is often much more superficial. He understands that many if not most of the ideas he will come up with are things that the founder has already considered and knows that while he can provide great input, advice and a different perspective, he should neither try to micro-manage nor try to make decisions for the founders….
A bad VC overestimates his insights, tries to micro-manage, tries to exercise control and becomes a maintenance burden for the founders.
Sure, yes, however: What sort of “advice” is actually useful?
Most commonly, that advice is an opinion—“here’s what I think makes sense; here’s how I’d reason about the problem; here’s my feedback on what you’re saying.” Generously, it’s an effort by VCs to solve the puzzle together; cynically, it’s their attempt to prove how smart they are.
The problem with opinions, aside from often being wrong, is that they’re easy to ignore. It’s easy for startups to say, “we have thousands of hours of experience in our industry and with our customers and intimately know the people on our team, whereas the VC’s knowledge of our startup is much more superficial.” Then they nod and say thank you and move on. And more often than not, that’s probably the right decision.
But there is a kind of advice that I wish more VCs would give: Tell startups what you think is going to happen next. VCs have seen a thousand companies make a thousand decisions and know how they’ve all played out. Don’t try to synthesize that into a clever plan; just play the tape.
“I know you think you can do both of those things at once. Maybe you’re different, but every other time I’ve seen a company try that, they got distracted, first like this, then then, and then these bad things started to happen.”
“I know you think you’re going to build something that’s not BI. Maybe you’re different, but every other time that’s been attempted it’s become BI.”
Tell the stories. Do it with details. Don’t tell us what you think; tell us what you’ve seen. As I’ve said before, most startups are “killed by the thing that everyone knew about, everyone saw, everyone felt, and nobody fully faced.” Good VCs are humble advisors, bad VCs are overconfident micromanagers, and useful VCs tell the blunt stories that make sure startups fully face what’s probably coming next.
Why not use P/E ratios here? One, because I couldn’t find them after three Google searches and I gave up. And two, I’m not sure aggregate P/E ratios make as much sense for SaaS companies because a lot of their earnings are negative? But maybe they do? I don’t know. I work for startups; we’re supposed to make fun of MBAs and their bean counting. For revolutionaries like us, accounting figures are rough values, and could be slightly off; there is also obviously a chance of typos etc.
These exact numbers are a joke; I doubt that VCs literally did this sort of math. But I’m sure these signals were part of their mental calculus. “Yes, this startup only has three paying customers, but their launch got 400 upvotes on Product Hunt, so we’re gonna YOLO an observer-only term sheet at them for $20 million on a $100 million post.”
“In some categories, pretty much all companies are/would be up for sale.” The point here is subtly devastating. For some data startups, they aren’t just overvalued; they solve a problem that the market doesn’t put any meaningful value on solving. Such is the risk of investing in green fields; they might be undiscovered fertile lands, or they might be deserts.
If that seems low, it is—for, uh, Facebook, when it was going absolutely vertical in 2005. That year, less than two years after its founding, Facebook had 5 million users, was the seventh most popular website on the internet, and was making $9 million in revenue, which was up 2,300 percent from 2004. For that explosive growth, Facebook was rewarded with a valuation of $550 million—or 61x their revenue.
In a prior post, I said that down rounds are mostly a fictional disaster, and I still think that’s true. However, going from $1.5 billion to $60 million is an extreme case; it would obviously be very bad if a company turned $200 million in cash into a warped preference stack full of some risky preferred shares worth a total of $60 million and a bunch of common shares worth $0. But less extreme down rounds—going from $1.5 billion to $1 billion, say—aren’t as economically bad as they’re made out to be.
The problem with down rounds, even the more modest ones, is their psychological effect. VCs want to invest in exciting businesses; down rounds signal that the business is no longer exciting. From Matt Levine this week:
WeWork is probably the funniest, but there are lots of companies like this, companies that were funded by venture capital firms in the boom based on ambitious growth targets and plans to change the world, and that failed to achieve those ambitions, but that are, you know, fine, viable, possibly even profitable one day.
In some ways that is the worst way to fail? For a startup founded by a visionary entrepreneur and funded by venture capitalists? Like:...
The awkward, what-do-we-do-now outcome is that you don’t change the world, the company chugs along, it makes enough money to pay salaries and maybe open an office in Dallas and perhaps write a quarterly dividend check to its VCs. But the VCs are like “I do not want a quarterly dividend check, a quarterly dividend check isn’t cool, you know what is cool, a trillion dollars.” And the founder is like “man I am paying myself $400,000 a year and working on the problem that excited me five years ago but I am not a billionaire and why am I in Dallas.”
A down round by itself doesn’t mean a company can’t change the world. But it does mean that people no longer believe it can, and in Silicon Valley, that can be enough to ruin a company. Or at least force it to sell itself to a PE firm in Dallas.
Yes, there are secondary markets, but these aren’t a great barometer. For most startups, shares aren’t very liquid, transaction volume is low, and people buy and sell for messy reasons (they want to put a down payment on a house and need the cash; an institutional investor needs to retire a fund; etc.).
Though honestly, being profitable isn’t a panacea. A profitable company making tens of millions of dollars a year in revenue—which is on the high end of where a lot of data startups are today—is how you end up in Dallas, working on a problem that excited you five years ago.