Exclusive: ClickHouse CTO Alexey Milovidov on AI and analytics

ClickHouse is setting its sights on the Asia-Pacific region as demand for real-time data infrastructure and AI-ready analytics continues to rise - and the company's Co-founder and CTO, Alexey Milovidov, believes the future lies squarely at the intersection of the two.

Speaking with TechDay following the Open House by ClickHouse event in Sydney on 2 October, Milovidov said Australia was a key part of the company's growing footprint.

The roadshow drew a packed room of developers, data engineers and enterprise customers eager to learn how ClickHouse is powering next-generation analytics.

ClickHouse was officially founded in 2021, but the technology itself goes back much further. Milovidov first developed the ClickHouse database in 2009, for a web analytics use case while working at a European internet provider. It was released as open source in 2016 - a move he now describes as the most important decision of his career.

Since then, ClickHouse has become one of the most widely adopted analytical databases globally, especially for high-throughput use cases like observability, financial services, and user-facing analytics - a growing priority in sectors like digital banking, telecoms, and streaming in Australia.

"In 2016, there were no alternative solutions of that quality," Milovidov said. "Some open-source databases didn't support compression or frequent inserts. Some were proprietary and too expensive. I proved in a prototype that I could do better."

ClickHouse is a column-oriented database that uses a unique storage engine, MergeTree, to combine fast insert performance with high compression.

Milovidov credits these early architectural choices with the database's growing edge in real-time analytics today.

"ClickHouse was initially created for web analytics – thousands of people visiting a site and querying traffic in real time. That required low-latency analytics, and that's what we built for."

Now, the rise of AI agents is fuelling a whole new class of data workloads – and ClickHouse is evolving once again.

"When you ask an AI agent a question, it doesn't just run one SQL query. It can run tens or hundreds, including exploration and retries, in a very compressed timeframe," Milovidov explained. "And you get used to that speed. You want it even faster."

This emerging approach, which he calls "agentic analytics", blends real-time query execution with the intelligent flexibility of generative AI.

"Agentic analytics is a combination of real-time analytics and AI agents," he said. "It's that simple."

That blend is already powering some of the world's largest AI-native companies. ClickHouse is used by OpenAI, Anthropic, xAI and Cursor, among others, to handle observability workloads on massive scales.

"The amount of traffic they analyse is petabytes per day," Milovidov said. "They do their own comparisons and POCs, and they find ClickHouse is the best."

But it's not just Silicon Valley's elite leaning in. The company now has thousands of paying customers globally, with its managed cloud offering available on AWS, GCP and Azure.

In June, ClickHouse raised $350 million in Series C funding to fuel that momentum.

According to Milovidov, the priorities are clear: customer-first product development, scalability, and continued expansion.

"We are trying to accommodate customer needs - but it's not easy when thousands of them all want to influence the direction," he said. "We try to unify the requests and prioritise what brings the most value."

One recent example? A major optimisation effort ahead of a high-stakes cricket match between India and Pakistan, where a ClickHouse customer was tracking video stream quality for 180 million viewers.

"All eyes were on dashboards generated from ClickHouse," Milovidov said. "We had to work directly with the customer to ensure it didn't go down."

Despite its cutting-edge performance, Milovidov is realistic about the physical limits of scale.

"If you generate petabytes of data daily and run a SELECT * across a year's worth of data - yes, that can take hours," he said. "Latency depends on the query, the data, and how you prepare your database structures."

To optimise performance at scale, ClickHouse relies on advanced features like projections and secondary indexes, and encourages users to design data models around frequent query patterns. But as the database becomes increasingly AI-aware, it's not just speed that matters - it's also clarity.

"When AI agents use a database, discoverability becomes important," Milovidov said. "You need a large prompt just to explain the data - its nature, how to query it. Even basic stuff helps. Good column names, clear comments - it makes a big difference."

That perspective also feeds into how ClickHouse approaches open source development. While the GitHub repository is open to issues and feature requests, Milovidov admitted the volume of community input far exceeds what the engineering team can handle.

"We cannot implement everything," he said. "But sometimes we can generalise and make one architecture change that solves many problems at once."

With the database continuing to gain traction across industries - from social media analytics to infrastructure monitoring - Milovidov sees the role of ClickHouse growing alongside the evolution of AI itself.

"Today, people are shifting from dashboards to autonomous data platforms," he said.

"We're seeing data analysis becoming more intelligent, more contextual, more automated."

It's a shift he welcomes, even if it challenges the traditional database model.

"When you explore unfamiliar data – like financial data, for example – even a human can struggle," he said. "But with the help of AI agents, and the right kind of analytics engine, we can make it feel natural. That's where we're headed."