Neo4j buys GraphAware to offer “no kill switch” intel stack vs Palantir

Neo4j is making a very specific bet in the intelligence software market: that governments are getting serious about “no-kill-switch” architectures, and customers want options that do not depend on a vendor you cannot fully control. CEO Emil Eifrem points to that shift as a core motivation for Neo4j’s acquisition of GraphAware, an intelligence analysis software platform built on Neo4j. And the timing has a clear regulatory shadow: on the same day as the European Commission’s European Technological Sovereignty Package, Neo4j is trying to land an argument that matters to European procurement, policy, and risk teams.

GraphAware is explicitly positioning itself as an alternative to Palantir, the controversial US spy-tech company. Its thesis is not just “we compete,” but “we can run where and how you need without someone else pulling the plug.” GraphAware CEO Michal Bachman tells The Register that the platform is open by design. It is built on the Neo4j graph database, uses the Cypher query language, which is very close to the open language GQL, and the intelligence platform is designed so it does not modify data in a proprietary way. That is a big deal in regulated environments: if you want to export it, you can export it.

The sovereignty narrative is where this gets hot. The Register ties the deal to growing concerns that European companies and government agencies are dependent on technology subject to US law, and frames the fear as a “kill switch” risk. Those concerns were stoked in June 2025, when Microsoft admitted under oath in a French court that it could not guarantee digital sovereignty if American authorities demanded access to data held on Microsoft servers located on foreign soil. Even though Microsoft is not the target of Neo4j’s product, the logic travels: regulators and buyers increasingly worry about what happens when the vendor is outside your jurisdiction.

One weakness identified in an analysis of European dependence on US technology is Identity and Access Management (IAM), where Microsoft’s Entra ID is described as a near-monopoly with few alternatives. That background matters because it explains why “on-prem” and “air-gapped” are not just deployment preferences. They are control levers. Bachman says the GraphAware and Neo4j stack can be deployed on-premises or in a private cloud, and that it does not “call home.” He goes further: it can be deployed in an air-gapped environment, without internet access, and “Nothing, no one can turn it off.” The message is blunt, and it is aimed at the procurement checklists that increasingly look like security audits plus political risk assessments.

There is also an architectural reason this pitch lands. Neo4j’s database is described as operating on an open-core model. In plain terms, that typically means the company offers a foundation that can be more flexible for enterprise customers, while still building commercial value around additional components. For customers running intelligence workflows, flexibility is not theoretical. Palantir is “mostly run as SaaS in the public cloud,” The Register notes, even though Palantir products can also be run in a private cloud or on-prem. So Neo4j is trying to win the discussion where it often decides: not just model quality or dashboards, but where the software lives, who can reach it, and what happens if you need to sever the network.

This is why GraphAware’s product positioning matters. Bachman describes an intelligence platform built on Neo4j graph semantics, suited for intelligence gathering. Palantir’s comparable offering, Gotham, is designed for defense, intelligence, and law enforcement agencies, helping operators connect the dots between disparate sources. In that competitive arena, the question is often who controls the workflow pipeline end-to-end. Graph databases are particularly attractive for “connect the dots” tasks because relationships are first-class objects. But sovereignty buyers care about something else too: control over deployment, data storage, and operational autonomy.

The regulatory and political pressure around Palantir is not hypothetical either. In the UK, concerns about Palantir’s growing presence in the public sector have been rising. The Register reports that last week lawmakers warned the spy-tech vendor’s increasing presence was an “unacceptable point of weakness” and urged the government to exercise the 2027 break clause in the NHS Federated Data Platform contract, which is based on the Palantir Foundry system. That matters because it signals to other governments that vendors who are seen as hard to fully control can become liabilities, even if the technology works.

Palantir’s leadership has also stoked controversy. The Register cites that last year CEO Alex Karp said the founders planned to build a company that “could power the West to its obvious innate superiority.” Earlier this year, he claimed AI would mean Western economies do not need immigration. Those are not product details, but they feed the broader trust environment around the vendor. In Neo4j’s framing, Eifrem adds another layer: the “cultural transformation” in technology and politics. He points to a shift in trust away from centralized institutions, describing a long arc from the 1970s through the internet era, social media, and decentralized voices. Whether you agree with the cultural diagnosis or not, the operational implication is clear: buyers want alternatives that reduce dependency on centralized gatekeepers, especially when political leverage could translate into technical access.

For Neo4j and its competitors, the strategic stakes are straightforward. If buyers treat “no-kill-switch” as a baseline requirement, then architecture becomes procurement leverage. This deal is not just an acquisition. It is Neo4j trying to productize sovereignty, translate open graph infrastructure into deployable intelligence stacks, and offer an option to customers who are tired of being told their security and autonomy depend on someone else’s jurisdiction and terms.