Key Takeaway
Frustrated with Oracle’s inability to handle vast unstructured data, Benoit and his co-founder left in 2012 to create a new system. Disliking Hadoop for its slowness and complexity, they aimed to design a platform that separated data storage from compute power, allowing rapid server allocation. Their breakthrough in data sharing stemmed from using Google Docs, inspiring them to apply real-time collaboration principles to large datasets. Today, Snowflake can allocate 1,000 servers in under a second. Jeff Hollan, a former Microsoft employee, now leads the company’s AI initiatives, focusing on conversational data interfaces.
Oracle’s systems struggled to adapt to this new reality. Hadoop – an open-source framework designed to process large volumes of unstructured data across clusters of commodity servers – was one option, but it frustrated both founders.
“I disliked Hadoop: it was slow and complicated,” Benoit recalls. The common justification was that complexity was unavoidable due to the data volumes. “We didn’t accept that.”
In August 2012, they departed from Oracle and rented a small apartment in San Mateo.
Benoit purchased a whiteboard, and for months, the two engineers worked in isolation, attempting to design a system that could decouple data storage from computing power, enabling thousands of servers to be provisioned instantly and released just as quickly.
Today, Snowflake can provision 1,000 servers in under one second.
The breakthrough in data sharing stemmed from their own frustrations while preparing for Series A presentations.
They exchanged slides via email until Thierry discovered Google Docs.
“We looked at it, and I said, ‘Wow, this is incredible – sharing documents,’” Benoit shares. They realized that the same concept should extend to datasets. If the cloud could facilitate real-time document collaboration, why not enable petabyte-scale data sharing?
AI agents with built-in governance
Jeff Hollan joined Snowflake three years ago after spending 15 years at Microsoft. Currently, as Director of Product for AI agents and Cortex AI applications, he leads the company’s initiative into conversational data interfaces.
86 Comments