Databricks disclosed on March 24, 2026 that it had acquired Antimatter, a data security startup founded by UC Berkeley security researchers. The news came in the same announcement in which the data and AI company launched Lakewatch, a security information and event management (SIEM) product, at the RSAC conference.
Terms were not disclosed. TechCrunch reported that the Antimatter deal had closed in 2025 but was not made public until the Lakewatch launch. Antimatter co-founder Andrew Krioukov had been working at Databricks for months and leads the Lakewatch team.
Databricks also announced a second security deal, for SiftD.ai, a small startup founded by former Splunk engineers. It said the two teams and their technology would be used to build out Lakewatch, and confirmed that the startups' employees had joined the company.
What Antimatter built
Antimatter came out of stealth in March 2022 with cryptographic infrastructure for software-as-a-service companies. Its system used secure enclaves, a processor feature that keeps data encrypted while it is being processed, so that SaaS providers could prove to customers that their data met residency, governance and tenancy requirements. The company said its service sat underneath an application and did not require code changes.
It later turned the technology toward AI agents. TechCrunch described its product as a data control plane that let enterprises deploy agents securely while protecting sensitive data, and Krioukov presented it at the RSA Conference Innovation Sandbox contest in 2024. In its announcement, Databricks said Antimatter's founders had laid the foundation for provably secure authentication and authorization for AI agents.
TechCrunch reported that Antimatter had fewer than 50 employees, based on LinkedIn.
“Security teams can no longer rely on manual workflows to outpace AI-driven attacks.”
The founders and the funding
Antimatter was founded in 2020 by Andrew Krioukov, Michael Andersen and Beau Trincia. Krioukov and Andersen both earned PhDs in computer science at UC Berkeley. Krioukov had previously founded Comfy, which Siemens acquired in 2018; Trincia was on Comfy's founding team and spent seven years at the design firm IDEO.
In 2022 the company raised a $12 million Series A led by NEA, with General Catalyst and UNION Labs. TechCrunch reported at the time that founders of Snowflake, Okta, Dropbox, VMware, Segment and Databricks also took part. "We're solving one of the most difficult problems we experienced as SaaS founders," Krioukov said when the round was announced.
Why Databricks is buying
Lakewatch is Databricks' first product built specifically for security teams. It collects security, IT and business data in one governed environment on the Databricks platform, using open formats, so that companies do not have to copy logs into a separate breach detection tool. Logs can stay in storage such as Amazon S3, and Databricks said it does not charge per-byte licence fees. The company claimed Lakewatch can cut total cost of ownership by up to 80%, and said high ingestion costs lead organisations to throw away as much as 75% of their data.
The product uses custom security agents built with Databricks' Agent Bricks and its Genie assistant to triage alerts and plan investigations, and several features run on Anthropic's Claude models. Detections are managed as code, and governance runs through Unity Catalog. Lakewatch launched in private preview, with Adobe and Dropbox named as customers.
"Security teams can no longer rely on manual workflows to outpace AI-driven attacks," said Ali Ghodsi, Databricks' co-founder and CEO. On stage at RSAC he described the approach as a way to "fight agents with agents".
The second deal: SiftD.ai
SiftD.ai was founded by Steve Zhang, who was chief scientist at Splunk until 2021 and created Splunk's Search Processing Language, together with lead architects of Splunk's search stack. It launched an interactive notebook for humans and AI agents to work on security investigations together in November 2025.
According to TechCrunch, the SiftD deal came together within a few weeks and closed the day before the announcement. PitchBook had no record of the company raising money, and TechCrunch described the deal as appearing to be an acquihire. Asked whether more acquisitions would follow, a Databricks spokesperson said: "We're always looking to what's next — our goal is to stay ahead of the market."
