A year after taking Andreessen Horowitz’s money, Databricks finally has something to show. At the Spark Summit, the company unveils its first actual product, a cloud-based platform built around Apache Spark, and shortly after closes a $33 million Series B led by New Enterprise Associates. The funding is the headline, but the product is the real news because it answers a question that has hung over the company since its founding: what, exactly, do you sell when the core technology is free?
The Problem With Selling Free Software
Apache Spark is open source. Anyone can download it and run it for free. That is wonderful for adoption and terrible for building a business, and it is a problem that has tripped up many open-source companies before. If your product is free, you have to sell something around it that people will actually pay for. Red Hat famously sold support and certification around free Linux. Databricks chooses a different path: sell convenience.
Spark is a distributed system, meaning it runs across many computers at once, and getting many computers to cooperate reliably is genuinely hard. You have to provision the servers, install and configure the software on each one, tune the memory settings, handle the networking, monitor for failures, and patch everything when updates come out. A skilled data engineer can do it, but it can take days to stand up a cluster and an ongoing effort to keep it healthy. For a lot of companies, the cost is not the software, which is free, but the team of expensive specialists needed to babysit it.
What They Actually Built
Databricks’ answer is a managed cloud platform that does all of that for you. Instead of spending three days standing up a Spark cluster, you log into a web page, click a few buttons, and have a working cluster in minutes. The platform handles provisioning, configuration, scaling, and failure recovery automatically. When you are done, the cluster can shut itself down so you stop paying for idle machines.
The other piece, and this turns out to matter enormously, is the notebook interface. If you have never seen one, a notebook is an interactive document where you write a chunk of code, run it, see the result immediately below it, and then write the next chunk. It combines code, results, charts, and written notes on a single scrollable page. Think of it like a lab notebook for data: instead of writing an entire program, running it, and hoping it works, you experiment step by step and watch what happens at each stage. Databricks builds collaborative notebooks into its platform, so a team of data scientists can work in the same document, the way people now take for granted in Google Docs. That collaborative, browser-based, no-setup experience for big-data work is genuinely novel right now.
The platform launches in limited availability on Amazon Web Services. That choice, building on AWS rather than shipping software that companies install in their own data centers, is a deliberate bet on where computing is heading, and it defines the company’s strategy.
The 33 Million Dollar Vote of Confidence
The Series B, 33 million dollars, led by New Enterprise Associates, comes on the strength of Spark’s rapidly growing popularity in the developer community, even before the commercial product has proven itself. This is a pattern worth noting, and one that looks set to repeat: the open-source project drives awareness and adoption, and the company monetizes a slice of that adoption through its managed platform. The free software is effectively the world’s largest and most credible marketing funnel, and it costs Databricks almost nothing because the community maintains much of it.
Spark is on its way to becoming one of the most active open-source projects in the entire big-data ecosystem, with contributors from hundreds of companies. Every one of those contributors and users is a potential Databricks customer, and every improvement they make to Spark makes the Databricks platform more valuable without Databricks having to pay for the engineering.
What It Means for the Market
The launch puts Databricks on a collision course with the Hadoop establishment, specifically Cloudera and Hortonworks, the two well-funded companies built on commercializing the older Hadoop ecosystem. The crucial strategic difference is the delivery model. Cloudera and Hortonworks largely sell software that enterprises install and run in their own data centers, the on-premises model that has defined enterprise software for decades. Databricks sells a cloud service. It is not at all obvious yet which approach will win. Plenty of large enterprises are deeply skeptical of putting sensitive data in someone else’s cloud, and on-premises vendors have the relationships, references, and revenue.
But the cloud-native bet carries a structural advantage that could compound over time. A cloud service can be updated continuously, scaled instantly, and adopted without a procurement cycle or a data center buildout. If cloud adoption accelerates across the industry, on-premises Hadoop vendors could find themselves defending a shrinking model while Databricks rides the growing one. The benefit to customers is immediate and concrete: dramatically less operational overhead. You no longer need a dedicated team to keep your data infrastructure running, which means smaller companies that could never have afforded one can suddenly do serious, large-scale data work. That democratization, taking a capability that previously required a big specialized team and making it available to anyone with a credit card and a browser, is the throughline of the entire Databricks story, and it starts with this product.
Whether it actually unseats the incumbents is still unknown. But the model that aims to do exactly that is now live on AWS and backed by $33 million.