Databricks crossed $350M run rate in Q3, up from $200M one year ago

The data analytics company scaled rapidly to put itself on an obvious IPO path

The Exchange regularly covers companies as they approach and crest the $100 million revenue mark. Our goal in tracking startups growing at scale is to scout future IPO candidates and better understand the late-stage financing market.

Today we’re digging into a company that is a little bit bigger than that. Namely Databricks, a data analytics company that was most recently valued at around $6.2 billion in its October, 2019 Series F when it raised $400 million.


The Exchange explores startups, markets and money. Read it every morning on Extra Crunch, or get The Exchange newsletter every Saturday.


The former startup reached a run rate of around $350 million at the end of Q3 2020, up from $200 million in revenue in Q3 2019, putting it on a rapid growth pace for a former startup of its size.

To better dig into the company’s performance, I got on the phone with its CEO, Ali Ghodsi, hoping to better understand how Databricks has managed to grow as much as it has in recent years. Ghodsi took over as CEO in 2016 after serving as the company’s VP of engineering. He’s also a co-founder.

Databricks is an obvious IPO candidate, but it’s also a company with broad private-market options, given its revenue expansion and attractive economics. Today, let’s talk about Databricks’ growth history, how it changed its sales process and what’s ahead for the unicorn more than six times over.

What does Databricks do?

What does Databricks actually do? Normally I’d be content to wave my hands at data analytics and call it a day. Chatting with Ghodsi, however, clarified the matter, so let me help.

Let’s say that a company has a lot of data on its machinery and wants to know when different pieces are going to fail. Or, perhaps a company wants to find patterns in some economic data. How do they find that information?

Ghodsi reckons you need three things: First, data engineering, or getting customer data “massaged into the right forms so that you can actually start using it.” Second, data science, which Ghodsi describes as “the machine learning algorithms, the predictive algorithms that you need to have.” And third, on top, companies “more and more” also want data warehousing and some “basic analytics,” he added.

Databricks brings all three together in its platform, which was the original vision of the company, Ghodsi said. (The Databricks founding story is too long for today, but it involves a combination of Berkeley, open-source software and hippies.)

When the company started back in 2013, growth was initially modest. The firm racked up around $1 million in revenue by 2015, with gross margins that Ghodsi described as “really bad” during our interview. In 2016, however, he took over as CEO and, with a few changes, managed to steer the company into a faster growth cadence.

In 2016, Databricks grew to what Ghodsi described as “double-digit ARR,” adding that it “wasn’t long” until the company reached triple-digit revenues. Given that the company reached the $200 million run rate mark in Q3 2019, we can draw up a funny sort of growth history:

  • 2013: Founded.
  • 2015: Revenues of ~$1 million.
  • 2016: Revenues of $10 million or more.
  • 2017 and 2018: The mystery years.
  • 2019: Reached $200 million run rate at the end of Q3.
  • 2020: Reached $350 million run rate at the end of Q3.

For additional context, Databricks raised its Series A in 2013 ($14 million), its Series B in 2014 ($33 million) and Series C in December of 2016 ($60 million). The company, therefore, raised its Series A and B on the back of slim revenues.

How the hell did the company raise a Series B with minimal revenues? I, too, was curious.

At the time Databricks had a more open-source focus. So, the company attracted capital based on usage of its software, the popularity of its conference and other nonfinancial metrics. Ghodsi joked that during the company’s early fundraising periods, investors would tell the company that “this is the last time we’re giving you money based on your hopes,” and that the next time they wanted to see “hard numbers.”

But around 2016 things shook up at Databricks, leading it to the growth path that we’ve outlined. So what changed? A few things.

Changing focus

According to Ghodsi, the company made three changes that moved it from an interesting open-source startup into the IPO-ready firm we see today: A shift in focus from open-source to proprietary code, a new focus on enterprise sales from a former bottoms-up go-to-market approach, and, third, a focus on what large companies needed from it.

The changes had material impacts. Per the CEO, Databricks went from being “largely open source” to, in his view, “80% proprietary.” The new sales effort, namely adding an enterprise go-to-market motion, helped the company bump its average deal size from $12,000 for a year to $50,000 and later into the six-figures. The third likely helped the second, as enterprises often have messy data and need help.

So a shift to owned code, big customers and tailoring its capabilities to manage larger accounts changed Databricks’ fortunes. Today, with a run rate of $350 million and what the company described to TechCrunch in an email as “80%+ subscription gross margins,” the company has grown into its 2019-era private valuation and could raise more funds at a higher valuation from either private or public investors.

So why doesn’t Databricks go public? Instead of asking Ghodsi when the company would debut (that question is always a waste of time), I asked why Databricks would stay private when public markets are so welcoming to growth-focused software companies?

The CEO noted that his company had said that it would be IPO-ready next year, a statement I protested a little because the company is large and hired a CFO a year ago. Ghodsi then said some standard CEO stuff and noted that an IPO is not a destination, but is, instead, “the start of the race.”

After a few more CEO-isms, he said something rather interesting (quote lightly edited for readability):

There are pros and cons to being public and not being public. There are things we can do, for instance, around pricing that we have done which [will] really help us in the long run. But if you were public, I’m not sure I would have made the same decisions because [ … ] you can get revenue cannibalization. We made a lot of decisions in the last couple of years where we cannibalized ourselves and cannibalized our own revenue and cannibalized our own offerings, but [those decisions set] us up for growth in the long term. You know, it’s harder to do those things when you’re public. So that goes into the equation.

This is perhaps the only good reason I’ve heard yet about why a company is holding off on an IPO. Mostly the reasons are bunk. This one is reasonable.

Regardless, that’s a bit on Databricks, its growth and how it managed it. I suppose this piece fits into our broader $100 million ARR series, which feels nice. More to come when Databricks wraps 2020 and we demand a new run rate figure.