Databricks Expands Into Cybersecurity with Lakewatch

The data and AI company bets its open lakehouse architecture can redefute how enterprises defend themselves, just as it readies for a potential public listing.

Key Takeaways

• Databricks has launched Lakewatch, an agentic SIEM platform built on its lakehouse architecture, promising up to 80% lower total cost of ownership versus legacy security tools by eliminating proprietary data formats and per-byte storage penalties.
• Two undisclosed acquisitions, Antimatter and SiftD.ai, bring deep expertise in AI-native authentication and large-scale detection engineering, providing Lakewatch with a technical foundation built by the architects of the technologies it aims to displace.
• With a revenue run-rate above $5.4 billion and a $134 billion valuation, Databricks is using Lakewatch to broaden its addressable market and deepen platform stickiness ahead of what could be a 2026 IPO, framing cybersecurity as a data problem the lakehouse is uniquely equipped to solve.

A Platform Finds Its Next Frontier

When Ali Ghodsi founded Databricks on the thesis that data and intelligence belong on a single, open platform, cybersecurity was not the obvious destination. It is, however, a logical one. Databricks announced Lakewatch, an agentic security information and event management platform built natively on its lakehouse architecture, accompanied by the disclosure of two previously unannounced acquisitions. The move extends the company’s gravitational pull into one of the enterprise’s most consequential and data-intensive functions, arriving at a moment when the incumbent SIEM market is overdue for structural disruption.

The announcement is not a sideshow. It is a deliberate expansion that reflects both the maturity of Databricks’ core platform and the strategic logic of a company approaching the public markets. For institutional observers, the question is not whether Databricks can build a security product. It is whether the architecture that redefined analytics and AI development can do the same for security operations at scale.

What Legacy SIEM Gets Wrong

To understand Lakewatch, it helps to understand what it is designed to replace. Traditional SIEM platforms were architected in an era when enterprise data volumes were manageable and threats were comparatively linear. That world no longer exists. Modern attacks unfold at machine speed, exploit AI-driven reconnaissance, and leave forensic traces scattered across cloud environments, endpoints, identity systems and application logs that no single legacy tool was designed to ingest simultaneously.

The consequences of architectural mismatch are measurable. Security teams currently discard up to 75 percent of their telemetry because retention costs make comprehensive storage economically untenable. Investigations are fragmented. Correlation is limited by what vendors choose to index. Alert fatigue is endemic. The SIEM has become, for many organisations, a compliance artefact rather than an operational advantage.

Databricks’ response is architectural rather than incremental. Lakewatch stores security data in open formats, specifically Delta Lake and Apache Iceberg, within the customer’s own cloud object storage. This decoupling of storage from compute removes the per-byte penalties that force defenders to filter their data before they understand its value. The company claims total cost of ownership reductions of up to 80 percent, a figure that will attract scrutiny but reflects a structural cost advantage that is genuinely difficult for incumbents to replicate without dismantling their own pricing models.

The Acquisitions That Make It Credible

Announcements of intent are common in enterprise technology. What distinguishes Lakewatch is the depth of the technical foundation Databricks has quietly assembled beneath it.

The first acquisition, Antimatter, was completed in 2025 but disclosed only now. Founded by researchers from UC Berkeley, the company specialised in provably secure authentication and authorisation for AI agents, a narrow but increasingly critical problem as agentic systems gain the capacity to act autonomously across enterprise environments. As AI agents proliferate, the question of what they are permitted to access, and how that permission is enforced, becomes a foundational security challenge. Antimatter’s founder, Andrew Krioukov, now leads the Lakewatch product team, a signal of how seriously Databricks treats the agentic dimension of the platform.

The second acquisition, SiftD.ai, was completed more recently and brings a different kind of credibility. Its founding team includes the creator of Splunk’s Search Processing Language and several lead architects of that company’s search infrastructure. These are people who built the tools that defined the previous generation of security analytics. Their decision to rebuild detection engineering on a lakehouse architecture, and to bring that work inside Databricks, carries more weight than any marketing claim about disruption.

Together, the acquisitions give Lakewatch something rare in enterprise technology launches: genuine domain authority on both the security and infrastructure sides of the problem.

Agentic Defence at Machine Scale

The most forward-looking element of Lakewatch is its embedding of AI agents directly into the security operations workflow. This is not a chatbot layered onto a dashboard. Databricks has deepened its partnership with Anthropic, whose Claude models power a suite of capabilities designed to automate detection, triage, investigation and response. Agent Bricks allows security teams to build and deploy custom agents that operate within the governed lakehouse environment. Genie, Databricks’ conversational AI interface, translates natural-language queries into multi-step threat hunts that would previously have required specialist scripting. Detection-as-Code enables security rules to be managed, versioned and tested like software, bringing engineering discipline to a function that has historically resisted it.

The rationale is grounded in a simple asymmetry. Attackers are deploying AI agents capable of continuous, coordinated action at a speed no human analyst can match. Defenders have been slower to automate, constrained by tools that were not built for the volume or velocity of modern threat data. Lakewatch, in principle, allows defensive agents to operate at the same pace and scale as offensive ones, drawing on the full, unfiltered breadth of enterprise telemetry rather than the filtered fraction that traditional SIEMs can afford to retain.

Anthropic’s decision to run its own security lakehouse on Databricks is worth noting. It functions simultaneously as a validation, a partnership signal and a quiet acknowledgement that the two companies’ futures are becoming increasingly intertwined.

The Pre-IPO Architecture of Ambition

The strategic context cannot be separated from the financial one. In its February 2026 update, Databricks reported a revenue run-rate exceeding $5.4 billion, up more than 65 percent year on year, with free cash flow positive over the preceding twelve months. AI-related products alone crossed a $1.4 billion run-rate. The company completed a $5 billion equity raise at a $134 billion valuation earlier this year, part of more than $7 billion in total new financing. Ghodsi has indicated that a public listing in 2026 remains plausible, subject to market conditions.

In that context, Lakewatch arrives with obvious strategic purpose. It expands the total addressable market, deepens platform stickiness among existing customers and demonstrates that the company’s growth is not dependent on a single product category. Global SIEM spending is growing at a double-digit compound annual rate, driven by cloud migration, expanding regulatory requirements including NIS2 and DORA, and the imperative to handle ever-larger volumes of security data. By positioning the lakehouse as the natural substrate for security operations, Databricks is making a claim on that market without building a purpose-built security company from scratch.

The partner ecosystem reinforces the seriousness of the commitment. Akamai, Arctic Wolf, Okta, Palo Alto Networks, Proofpoint and Zscaler are among the integrations announced at launch. Early design partners include Adobe, Dropbox and the National Australia Bank. The roster suggests that the platform is being evaluated by organisations with demanding requirements and genuine alternatives.

The Thesis, Extended

What Databricks is ultimately arguing with Lakewatch is that security is a data problem before it is a tooling problem. That framing is not novel; analysts have been gesturing toward it for years. What Databricks offers that others have not is a platform with the scale, openness and AI integration to make the argument operational rather than theoretical.

The competition will not concede gracefully. Established SIEM vendors have invested in cloud-native capabilities and AI features of their own, and many have deep relationships with the security operations teams Databricks is now courting. Displacement in enterprise infrastructure is slow, and early adopters remain few. The measure of Lakewatch’s success will not be the launch announcement but the rate at which security operations centres can ingest, retain and act on unfiltered enterprise data at a cost that makes comprehensive visibility viable for the first time.

For a company whose entire commercial thesis has rested on exactly that promise, the expansion is not a departure. It is a continuation, arrived at by the most direct path available.