ClickHouse takes aim at Databricks and Snowflake

Welcome to Runtime! Today: ClickHouse sends yet another signal that observability tools could help get enterprise agents over the hump, ServiceNow teams up with OpenAI, and the latest funding rounds in enterprise tech.

Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime each week, and if you value independent enterprise tech journalism, click the button below and become a Runtime supporter today.

Data on data

Companies that have managed to deploy enterprise agents in production tend to have one big thing in common: They started that process with a modern approach to data management already in place. The rush of the laggards to cloud data warehouses and data lakes paid off handsomely for Databricks and Snowflake over the last couple of years, but new challengers always emerge during a platform shift.

ClickHouse announced Friday that it has raised $400 million in new funding, which values the company at $15 billion. The four-year-old startup now counts over 3,000 customers for its cloud database service, which manages an open-source analytical-processing database built around the need for speed.

• ClickHouse's technology was incubated at Yandex, where the goal was to create "the fastest OLAP database on earth," according to the company.
• OLAP (online analytical processing) databases have been around for decades and are designed to help corporate leaders make decisions about the current and future health of their business.
• But ClickHouse's founders came up with a way to store data in columns that allowed users to make super fast, almost real-time queries of that data, which turned out to be an interesting option for companies looking to modernize their approach to data or startups building something new.
• "The other thing that I think is a big driving force in the market that we're seeing with earlier-stage companies is the growth of adoption of ClickHouse, another open-source, very low cost but high performance analytical database system," Battery Ventures Danel Dayan, who is not an investor in the company, told Runtime in 2024.

Databases such as ClickHouse have played a big part in the embrace of observability tools, which need fast and cheap access to data as companies look to troubleshoot and repair issues with traditional and AI applications. Snowflake's acquisition of Observe earlier this month showed that the data analysis and observability sectors are coming together at a rapid pace, and ClickHouse put some of that new capital to work immediately by acquiring Langfuse.

• Langfuse had raised $4 million in funding to develop its open-source LLM observability product, which "focuses on ensuring that non-deterministic and increasingly complex AI systems produce outputs that are accurate, safe, and aligned with user intent," ClickHouse said.
• Just as traditional observability tools allow users to determine the root cause of performance issues, LLM observability tools promise to reduce hallucinations and help companies manage AI agents, which require careful scrutiny given their ability to execute tasks autonomously.
• "With Langfuse in-house, [ClickHouse] now has the tooling to convert every LLM call into a structured, queryable record," Greyhound Research's Sanchit Vir Gogia told InfoWorld.

More than a year after enterprise vendors promised that AI agents were ready to transform the ways businesses operate, the tools and techniques that might actually make that happen are slowly falling into place. Dealing with unstructured data — such as internal emails, presentations, and documentations — has proven much harder than anyone was willing to acknowledge back in those heady days, and the entire premise of agentic AI is that a new era of productivity is just waiting to be found in those reams of unstructured data.

• But real-time observability tools could help companies feel much more confident that they can prevent their agents from going off the rails as they process unstructured data.
• "The three main personas of agentic analytics — data professionals operating databases, developers building AI features, and business users consuming insights — all benefit from knowing that their AI interactions are monitored, evaluated, and improving over time," ClickHouse said in a blog post.
• And expect data companies to bring new capabilities into their products as they become a central management hub for agentic AI, according to Theory Ventures' Tomasz Tunguz, who has been tracking the ongoing consolidation of "the modern data stack."
• "In addition, there are many pieces of the AI stack not represented here, such as evaluations & agent orchestration," he wrote Tuesday. "Those are next."

Model citizen

If OpenAI is going to come even close to finding enough money to honor the infrastructure spending commitments it made in 2025, it's going to have to dramatically increase its presence in the enterprise. ServiceNow is already there, and the two companies announced a partnership deal Tuesday to allow ServiceNow customers easier access to OpenAI's models.

The three-year deal "will give customers direct access to [OpenAI's] frontier capabilities," ServiceNow said in a press release. The Wall Street Journal reported that the deal "includes a revenue commitment from ServiceNow to OpenAI," but terms of that deal were not disclosed.

ServiceNow also agreed to build custom services for customers based around OpenAI's models, including voice agents that can detect speech commands and respond accordingly as well as new automation capabilities built around computer-use models. It's not an exclusive deal;  ServiceNow actually invested in Anthropic's last funding round, and the company wants customers to have access to the best models for whatever they need on its platform, president Amit Zavery told Runtime in 2024.

Enterprise funding

Parloa raised $350 million in Series D funding for its customer service agents, valuing the company at $3 billion.

Emergent scored $70 million in Series B funding as it builds out a no-code software development tool based around generative AI.

Listen Labs landed $69 million (nice) in Series B funding for its customer research technology, which allows customers to use AI to interview current and potential customers.

Aikido Security raised $60 million in Series B funding as it builds out its platform for embedding security into the software-development process.

Novee launched with $51.5 million in seed and Series A funding for its penetration-testing platform, which uses AI to probe customer defenses for security flaws.

depthfirst scored $40 million in Series A funding as it builds out a platform of security agents that can detect and respond to security vulnerabilities.

The Runtime roundup

Security researchers found significant vulnerabilities in MCP servers run by Anthropic and Microsoft, according to Dark Reading, underscoring that while MCP has great potential to connect AI agents with the data they need, right now it's still a security minefield.

The EPA closed a loophole that allowed data-center operators to operate gas turbines without Clean Air Act permits, according to CNBC, which could make life a little harder for xAI's Memphis operation but allow residents to breathe easier.

Thanks for reading — see you Thursday!