Firebolt raises $127M to fuel cloud data warehouse efforts

Cloud data warehouse vendor Firebolt today said it raised $127 million in a Series B round, bringing total funding for the company up to $164 million.

Founded in 2019 and headquartered in Tel Aviv, Israel, Firebolt raised its $37 million Series A in December 2020. The company's core technology is a cloud data warehouse service that can use cloud data lake storage as a repository for information that is then used for business intelligence and data analytics.

The market for cloud data warehouses has been a hot one in recent years, especially in the wake of the initial public offering of Snowflake in 2020. Firebolt is looking to gain share in the growing space with an approach that works for developers and uses a proprietary format for data that the company calls the "polymorphic file format."

In this Q&A, Firebolt CEO and co-founder Eldad Farkash outlines technology the vendor has built and the challenges organizations face in optimizing data in the cloud.

Why are you now raising a Series B for Firebolt just six months after you last raised money?

Eldad Farkash: Over the last six months, we went from being in our stealth phase and working with design partners to being full blown in the market and taking customers into production. So, the last six months have been crazy for us.

Eldad Farkash

We have expanded to many places. So, basically, wherever we find the expertise we need, we build an office. This is why today we are distributed over Tel Aviv, Cluj [Romania], Munich, Dublin, Zurich [Switzerland] and San Francisco.

So, we're two and a half years old and we want to spend the money on building a product that serves our market, which is mostly around data engineers wanting to take data to production and serve data to analytics users.

Why did you help start Firebolt and build a new cloud data warehouse in the first place?

Farkash: I've been building databases my whole career. I co-founded Sisense and was CTO there for 15 years. Sisense, originally, was an HPC [high-performance computing] company and we pivoted very early on that to become a BI tool.

I left Sisense and I wanted to focus on what matters, which is the layer that computes the query. The vision I had was very simple: It was about commoditizing efficiency.

I left Sisense and I wanted to focus on what matters, which is the layer that computes the query. The vision I had was very simple: It was about commoditizing efficiency.

Firebolt is focused on data engineering and data engineers that want to solve real challenges on top of the data lake and they want to serve the data to users.

So, whether you're building an app, a website or a game, whether you're building any type of online experience that utilizes data to actually drive your product -- not just drive insights from your product -- that's what Firebolt is designed for. That is very different from classic data warehouses, which have a very DBA [database administrator] type mindset.

What is the core technology that helps to enable the Firebolt cloud data warehouse approaches?

Farkash: It all starts with our own unique file format, proprietary file format. Unlike Apache Parquet, which focuses on interoperability for many systems, we wanted to design a file format that's actually focused on speed and efficiency.

With Firebolt, when you run a query, we run it against the index that we create. The query will return data ranges for where data is located, instead of downloading files. Those ranges get stored in a unique format in cache. We call it the polymorphic file format because of the way the file is being stored along different storage tiers from Amazon S3 [Simple Storage Service] through the cache, up to the RAM [system memory].

Firebolt is elastic, meaning that you have one database on a version of the data sitting on S3, then you have compute scaling up and down serving different needs, but all with the same data lake.

What are the key challenges for cloud data warehouse users that you have seen?

Farkash: Today, for customer success, companies need to use granular data for products. That is different from back when internal BI and analytics was perfect for the data warehouse, and then you had the database team doing the complex stuff for products.

Now the full-stack engineer who's writing JavaScript or Python will need to run a query that runs over billions of records and aggregates the data on read. So, I think the biggest change is that most of the problems that, in the past, we could solve on write are moving and becoming problems we need to solve on read.

With a data warehouse, you have many ways to really see the data. But the biggest challenge is actually serving the data to your audience, versus just uploading into the report and then doing a daily batch update.

Editor's note: This interview has been edited for clarity and conciseness.