Crunchy Information
Monday introduced the primary shock from Snowflake Summit 25: the acquisition of Crunchy Information for a reported $250 million. Crunchy Information is a longtime South Carolina firm that has been operating and constructing upon the open supply Postgres database for effectively over a decade, so what prompted the $70-billion knowledge large to purchase it now?
When the cool youngsters in IT had been going gaga for Hadoop and NoSQL databases again within the early 2010s, Paul Laurence, the president and co-founder of Crunchy Information, had a distinct thought. Whereas these different applied sciences had their backers and successes, Laurence was centered on one other rising pattern–one which additionally concerned enthusiasm for the potential for open supply, however with out all of the rule-breaking that Hadoop and NoSQL entailed.
“They had been having a bet on a brand new open-source knowledge administration toolbox, the place they’d these NoSQL databases. That was a part of the place they had been going,” Laurence advised BigDATAwire in a July 2024 interview. “However additionally they noticed this as a chance to shift in the direction of open supply on the relational and SQL aspect. They usually had evaluated the panorama and chosen Postgres as sort of their huge guess.”
Laurence additionally determined to make a guess on Postgres in 2012, when he co-founded Crunchy Information in Charleston, South Carolina along with his father, Bob. Over time, the corporate launched a variety of Postgres-based choices, together with a supported distribution of the database referred to as Crunchy Postgres, a model of Postgres to run on Kubernetes containers, a cloud-hosted model referred to as Crunchy Bridge, and at last an analytics model referred to as Crunchy Bridge for Analytics.
By the point 2024 rolled round, Crunchy Information had constructed a stable enterprise with 100 workers and 500 clients. All of its merchandise are 100% suitable with Postgres, a assure that’s crucial for the corporate’s clients in addition to the gas that drives the continued progress of Postgres. The guess on a boring piece of relational tech created by Michael Stonebraker 35 years in the past had paid off, which is a small miracle within the hype-driven echo chamber of IT.
Crunchy Information President and co-founder Paul Laurence
“Postgres was largely dismissed after we first began. It was actually all about Hadoop and MongoDB,” Laurence advised us. “It was very contrarian on the time…Other people would say ‘You guys are loopy. What are you guys doing beginning a SQL database firm? SQL is useless. That’s not an actual factor anymore.’”
Snowflake Finds a Postgres
You wouldn’t mistake Snowflake for being a giant fan of Hadoop or NoSQL, both. In truth, the corporate’s success relies in no small half on the wave of Hadoop failures within the late 2010s.
“I can’t discover a blissful Hadoop buyer. It’s form of so simple as that,” Bob Muglia, the previous CEO of Snowflake, advised this publication again in March 2017. “It’s very clear to me, technologically, that it’s not the know-how base the world shall be constructed on going ahead.”
Muglia, after all, was proper: Hadoop would go on to implode in early June 2019, thereby paving the best way for the rise of cloud-based knowledge warehouses and lakehouses, together with Snowflake. Simply over a 12 months later, amid the COVID-19 pandemic, Snowflake famously went public because the “largest ever” software program IPO ever.
Since then, Snowflake has been in a neck-and-neck race with Databricks to develop the business’s preeminent cloud knowledge platform. Databricks, after all, is the corporate behind Apache Spark, which additionally contributed to the decline of the Hadoop stack. Whereas Spark remains to be there, it’s simply part of what Databricks does anymore.
Like its rival, Snowflake has additionally been transferring away from the place it began. Whereas Snowflake has adopted newer analytics architectures for its RDMBS, together with MPP-style column retailer and separation of compute and storage, below the covers, Snowflake’s huge guess has been on conventional SQL and relational applied sciences.
Snowflake’s huge awakening below new CEO Sridhar Ramaswamy has been the significance of openness and a shift away from proprietary tech. That’s what drove it final 12 months to embrace Apache Iceberg, the open supply desk format that solved a bunch of the consistency points that arose with Hadoop’s buffet-style knowledge entry patterns.
The help for Iceberg tables as a substitute for Snowflake’s conventional tables additionally opened up the door for Snowflake clients to deliver an array of different question engines to bear on Snowflake knowledge, together with open supply Trino, Presto, Dremio, Spark, and Apache Flink, amongst others.
Now with the acquisition of Crunchy Information, Snowflake has a complete open supply database on provide. Snowflake Postgres, as the brand new providing known as, brings clients all of the open goodness of Postgres and its large ecosystem.
So what modified? Why is Snowflake doing this now?
What Modified at Snowflake
What modified is agentic AI emerged on the scene, and Snowflake sensed that clients wouldn’t be blissful utilizing its proprietary relational Snowflake database to deal with this rising workload.
Whereas Snowflake’s database is highly effective and succesful, the corporate isn’t about to open it up and permit clients to switch the database to deal with these rising agentic AI workloads in the identical method that they might with an open supply database–that they might with Postgres.
Snowflake EVP Christian Kleinerman addressed the explanations behind the acquisition of Crunchy Information throughout a query and reply session with the press on Monday at Snowflake Summit 25.
(nikkimeel/Shutterstock)
“I feel before everything, it confirms our dedication to interoperability after which buyer alternative,” Kleinerman mentioned. “I feel open supply is slightly bit much less vital. However what issues with Postgres isn’t whether or not it’s open supply or not. What issues is there’s a really massive neighborhood of folks that worth Postgres constructed for Postgres, construct on prime of Postgres.
“That’s what drove us,” he continued. “That was the requirement for lots of the clients and companions that we heard, which is we like numerous the know-how you construct, however it could make it method simpler for us to deliver computation, to deliver functions, to even deliver brokers onto the information in case you have one thing that’s Postgres suitable.”
With regards to interoperability and requirements, Postgres is king. Snowflake’s transfer exhibits that clients are prepared to place up with rather less efficiency in the event that they make it again with interoperability and requirements.
Snowflake regarded on the potential to construct a Postgres suitable layer on prime of its current database, however ultimately, it determined to purchase Crunchy Information as an alternative.
“Clients and companions mentioned, please give me Postgres compatibility,” Kleinerman mentioned. “We requested one in every of our absolute prime engineers, world class, prime 10, prime 20 on this planet in database design, can we layer the Postgres semantics and Postgres API on prime of our engine? And the reply was a really, very lengthy doc on why it gained’t work.
“And if clients need compatibility with Snowflake, Snowflake is nice,” he continued. “What we did with Unistore and hybrid tables is nice. But when they need compatibility with Postgres, it’s not simply the syntax, the semantics and all of that. It must be Postgres. The one option to obtain compatibility is with Postgres.”
By the best way, Databricks did the identical math, and determined to purchase Neon, a developer of a serverless Postgres database, for $1 billion.
Regardless of all of the progress in huge knowledge, nothing apparently can beat SQL and relational tech.
Associated Gadgets:
Databricks Nabs Neon to Resolve AI Database Bottleneck
Snowflake Widens Analytics and AI Attain at Summit 25
Crunchy Information Goes All-In With Postgres
