Data Council - The Startup Barometer
AI Takes Over 🤫
Coming to you live from Oakland, the annual data startup reunion (aka Data Council) with a recap.
My first Data Council was back in 2023 in Austin. I went in starry-eyed and excited, not knowing what to expect. What I found was an interesting combo. It’s not salesy and corporate like the big vendor conferences. It is mostly a data startup event rather than a data practitioner event. The talks are interesting and don’t feel too much like a pitch, but it’s startup employees or founders giving them. I attended in 2024 and 2025 because I knew I would meet with folks at other data companies who are partners/friends/prospective employees.
Given the density of startups, Data Council is a good barometer for the data startup market in general. Remember, in addition to being a commonality organizer, Pete Soderling, the Data Council founder, is also an investor.
So what does this barometer say about the data startup world in April 2025?
Hold on to your seats, I’m going to shock you here, AI was the biggest focus of the conference.
My recollection of previous Data Councils was that you had 2 data tracks and 1 ML/AI track. This year it was 2 AI tracks and 1 data track…
Conversations on the ground further bore this out. Stories of VC firms going 100% AI. Traditional data infrastructure companies struggling to find funding despite having built real businesses and revenue but lacking an AI story.
I’m sure this comes as a surprise to no one reading this but it’s yet another data point on the line.
In a panel discussion, Jordan Tigani of Motherduck captured concisely, something I have felt and written about: “AI is changing the data world a lot slower than the rest of the knowledge work world.” If you went back 5 or 10 years, would that be your prediction? It wouldn’t have been mine.
Having said that, on the same panel, Yury Izrailevsky, of Clickhouse also shared a fun llm demo they have built. This demo is a natural language interface for a bunch of common public datasets, e.g. NY Taxi, Github, etc. I’ve played around with it and will admit it works pretty well. I’ve looked up the Github star trends of a few projects and asked what the busiest day ever for taxi rides (it said September 19, 2010. Not sure if that’s correct?)
As much as with clean, well-structured datasets, like those in the ClickHouse demo, you can get AI to work pretty well. The panel, which also included Kishor Gopalakrishna from Startree and was MCed by MD (Mike Driscoll of Rill Data), made multiple mentions of the challenge for AI to understand the messy data structures of a typical large business. Something we simply haven’t solved yet. But, fear not, I can assure you there is plenty of capital and brainpower working on it.
The biggest highlight of Data Council is the people and community. It was great to see some old friends and make some great new connections.
A question for my readers to wrap up. If we look ahead a couple years to Data Council 2027, will we be back to 2 data tracks and 1 AI (AI growth stalls and excitement fades), up to 3 AItracks (AI is booming and all-consuming), or down to 0 (AI is REALLY booming and humans don’t need to show up to startup conferences anymore…)
My best guess is that we’ll be somewhere in the middle zone. We’ll continue to need data infrastructure, it will be important for AI, AI will continue to grow and get more valuable, but we won’t see artificial superintelligence take over the world in that timescale.
Until next year, Data Councilors! ❤️
Warmly,
Paul Dudley
