Airbyte, the well-funded open source data integration startup, has always made it easy for data teams to set up their ELT (extract, load and transform) pipelines, but until now that meant hosting and managing the service themselves, with all the associated complications. Today, the company announced the official launch of Airbyte Cloud, a hosted service that inherits all the features of the open source version, adding hosting and management, as well as a number of additional support options and business features such as access management for teams (although single sign-on support is not yet available). is always listed as “coming soon”).
Currently, more than 6,000 companies use Airbyte in one form or another. That’s up from just 250 at the end of January. Over the course of the year, the company also secured a seed round and Series A round, totaling just over $31 million in funding. The fact that there were only two months between the seed and Serie A rounds is a pretty good indication of how warm this space is.
And speaking of money, Airbyte also decided to confuse its pricing model a bit with Airbyte Cloud. Instead of volume-based pricing, which used to be kind of the norm for these types of services, the team decided to charge for the computation time these tasks take.
Ideally, this takes away at least some of the friction that often comes with this kind of workload. Traditionally, the Airbyte team argues, companies use multiple systems like Fivetran to connect to the most common API resources and in-house developed scripts that the data engineering teams build for their one-time use cases — and then a database replication system on top of that.
“We really wanted to standardize and solve the problem of data integration,” Airbyte COO and co-founder John Lafleur told me when I asked about the pricing model. “The only way to really do that is through an infrastructure-type pricing model like Snowflake: buy credits, consume these credits based on the calculation hours. At that point, because your database replication throughput is very high, it makes this possible. That is why we went for compute.”
As Airbyte CEO and co-founder Michel Tricot added, this also means companies can now think about pricing for virtually all their data services in the same compute-centric way.
Currently, Airbyte has about 130 connectors for services ranging from consumer products such as Instagram to BI systems such as Google’s Looker and virtually every major database system. As the team noted, many customers also use Airbyte’s open source code to build their own custom connectors. In addition, there are now vendors building connectors for their own customers, and the team is looking at how to incentivize its community to preserve connectors over time through some form of revenue share.