ClickHouse’s Recurring Revenue Passes $350 Million as OpenAI, Agent Use Jumps

Excitement for artificial intelligence agents is benefitting businesses whose software manages and monitors the databases that the AI bots use to accomplish their tasks.

ClickHouse, which sells a kind of database that can quickly analyze large volumes of data, such as website visitor activity, recently generated more than $350 million in annual recurring revenue, up 40% from May, according to a person close to the company. ClickHouse’s offerings, which include free open-source software and paid cloud services to run it, have been gaining more traction with companies including OpenAI, whose use of ClickHouse jumped in the past year, the person said.

OpenAI also has been a user of Datadog, a publicly-traded provider of software that monitors cloud applications for performance issues and cybersecurity attacks. Its stock was on a tear this year until it fell 19% in one day earlier this month after Datadog executives said its largest customer had reduced spending. They didn’t name the company, but OpenAI had already been moving some parts of its business from Datadog to ClickHouse over the past year. This includes log management, or the records of software actions, J.P. Morgan analyst Mark Murphy wrote in May.

While Datadog also has a database for analyzing log data, it can be expensive for customers that use it for large-scale projects, say former customers, as Datadog charges them based on usage. ClickHouse investors and some industry analysts claim its database can analyze log data faster and at a lower cost.

Annual recurring revenue refers to the value of subscription contracts expected over the next 12 months.

Growth at ClickHouse, which investors valued at $15 billion in January, shows how a once obscure pocket of enterprise software has risen in prominence over the last year, thanks to the advent of AI-driven agents. Well organized data is crucial for agents to operate effectively, and that demand also has been a boon for other companies including Snowflake and privately held Databricks and Grafana.

ClickHouse’s gross margins range between 50% and 70%, CEO Aaron Katz said in an interview. That compares to 79% for Datadog and 67% for Snowflake in their most recent quarters.

“If we wanted to be cash flow positive next year, we could be,” said Katz. Instead, the company has chosen to invest in its future growth, he said.

Yandex Spinoff

Russian search giant Yandex launched ClickHouse in 2016 as an open-source project. In 2021, Yandex spun it out as a standalone company, which hired Katz, a former executive at data search company Elastic. Cloud provider Nebius, which previously operated as the Dutch holding company of Yandex, owned a 28% stake in ClickHouse as of May 2025, though that’s likely fallen since then as ClickHouse raised more money.

ClickHouse earlier this year branched out from its roots as a database provider by acquiring Langfuse, a startup that sells software for monitoring performance of AI applications and agents. That deal boosted ClickHouse’s entry to the observability market and put it into more direct competition with Datadog and others.

Features that help developers monitor, or observe, digital actions have been in high demand as AI agents like Anthropic’s Claude and OpenAI’s Codex have grown in use and as companies using agents need to track their activity.

Before implementing agents, “you need to know what’s going on, and be omniscient,” said Max Gazor, an investor at Striker Venture Partners. “You need guardrails and controls.” Gazor is an investor in Cribl, which helps companies route their data to the right storage or monitoring platform to control costs.

Datadog still dominates the market for observability. Revenue rose 36% in the June quarter compared to last year, to $1.12 billion. That was around $45 million more that it forecast last quarter, and it raised its annual revenue forecast by around $140 million.

Still, the stock has yet to recover the high before it warned in early August of reduced spending from a major AI customer.

OpenAI, a high-profile user of ClickHouse for a few years, has been using the startup to store and analyze the continual stream of log file data its servers and applications generate, according to a person familiar with the activity. When OpenAI trains its models, engineers also use Clickhouse to trace what went wrong across a run. Once a model is live, they use the database to track real-world data performance, such as whether the model is responding fast enough. And when agents run on OpenAI’s models, they use it to check whether those agents are behaving as intended, said the person.

OpenAI was drawn to ClickHouse’s open-source model, said OpenAI technical staff Akshay Nanavati and Poom Chiarawongse in a video published in June 2025. The open-source structure means that OpenAI staff can read the code and fix a bug rather than wait on a support ticket. Like many open-source software providers, ClickHouse offers a version of its software for free and charges for more advanced versions, such as a cloud-based product that the startup manages for customers.

OpenAI’s usage has risen about 10 times in the last year, to over thirty petabytes a day, or the equivalent of 30 trillion events a day, according to the person familiar with the activity.

Kevin McLaughlin contributed to this report.

Alix Coutures is a reporter covering startups and VC based in New York City. She graduated from the M.A Business concentration at Columbia Journalism School in May 2026. She previously covered tech and telecom in Paris, at the French business publication Challenges. She can be reached at [email protected] or on Signal at Alix_coutures.34