Databricks CEO Ali Ghodsi speaks at the Spark Summit in June 2017.
Just months after announcing a $60 million fundraise, the creators of popular open-source big project Apache Spark and their startup have dwarfed that amount with a massive new fundraise — and added a new buzzword to the mix.
The new funding for Databricks clocks in at $140 million, and the opportunity that CEO Ali Ghodsi is targeting with it is one on the lips of marketing teams across Silicon Valley: artificial intelligence. "What we're seeing is just one percent of companies are successful with AI," Ghodsi says. For the 99%, companies not named Google, Facebook or Microsoft with large dedicated teams of AI experts, "It's a big gap between the haves and have-nots."
Silicon Valley's venture capital firms are believers in Ghodsi's gospel. Databricks' first investor Andreessen Horowitz led the new round, along with a previous lead investor, New Enterprise Associates and a slew of new investors including Battery Ventures, Future Fund Investment Co., A.Capital Partners, Geodesic Capital and Green Bay Ventures. Databricks didn't disclose its valuation after the new raise, but was already valued at more than $500 million by its previous round and raised at a higher price tag this time. Databricks sought a valuation in the ballpark of $900 million with the new round, according to a source with knowledge of the process.
Founded only in 2013 by seven members of the Spark team at the UC Berkeley AMPLab, Databricks grew alongside an explosion in popularity for their open-source project, a data processor used within the Hadoop big data ecosystem. Databricks has more than 500 customers today, with some paying in the "many millions" per year, says Ghodsi. But unlike many software companies built by architects of popular open source projects, Databricks doesn't make its money predominantly offering support and services to make it easier to use Spark. "We were shocked to find that the main problem people were facing didn't have to do with Spark," Ghodsi says. "It was a people problem."
Databricks calls its solution a "unified analytics platform," and what it functions as is a shared workspace for different types of employees to work together on big data projects including in AI. Engineers get a toolbox of solutions out of the (metaphorical) box such as for data access and data cleaning, along with a range of processing tools including but not limited to Spark; they then can chat with a company's business analysts and data scientists within the Databricks software. The company hosts it all in the cloud, meaning a customer doesn't need to maintain a version of Databricks on its own servers.
With the new funding, Databricks plans to expand into specific verticals such as financial services, government, healthcare and media. It'll also make a push outside the United States, says Ghodsi, who took over as CEO from cofounder Ion Stoica in 2016. While the company isn't profitable, Ghodsi says Databricks has "best-in-class" revenue growth and took on such a large sum strictly due to market opportunity.
"They're definitely one of the needle movers in our portfolio, and we have billions in our funds," says investor Pete Sonsini at NEA. "You take your XYZ open-source company, and they build a services model around training and support that takes a pile of money. These guys said, that's a bad way to build a business."
Instead, Databricks finds itself very much in the middle of the AI conversation, much as it was big data a couple years ago. This one, Ghodsi argues, isn't a fad. "When I say 99% of companies aren't successful with AI yet, I actually mean it," he says. "We see huge pent-up demand."