Startups & Business

Helsing defense AI and sovereign compute for European militaries

Munich-based Helsing reached an $18 billion valuation by developing advanced defense systems like the HX-2 strike drone and CA-1 Europa UCAV. The company addresses the European sovereignty gap by managing technical control over the computing stack and AI gateways to ensure data residency.

Helsing defense AI and sovereign compute for European militaries

Technical scale and drone production

Helsing raised $1.8 billion in July 2026, which brought its total valuation to $18 billion. The Munich-based firm produces HX-2 strike drones that reach speeds of 250 km/h and carry 5 kg of ammunition. These drones use Altra software to combine data feeds from different drones to identify targets. The company also manufactures the CA-1 Europa, a 4-tonne unmanned combat aerial vehicle with a 1,400 to 1,800 km operational range. This vehicle uses the Centaur AI agent for autonomous flight. Helsing also produces the SG-1 Fathom, an underwater drone that operates submerged for 90 days. This underwater system uses the Lura AI to process acoustic data and detect ships. The company was founded in 2021 by Torsten Reil, Gundbert Scherf, and Niklas Kohler. HX-2 drones are manufactured in a factory in southern Germany and have a range of 100 km. The company provides 6,000 HX-2 strike drones to Ukraine. The HF-1 is a cheaper version of these drones that uses a plywood fuselage. Helsing’s software is integrated into the Eurofighter Typhoon EK’s onboard system and the radar system of the Saab JAS 39 Gripen.

Specification HX-2 Strike Drone CA-1 Europa UCAV
Top Speed 250 km/h High subsonic
Range 100 km 1,400 – 1,800 km
Payload 5 kg 500 kg
Weight N/A 4 tonnes

Addressing the sovereignty gap

The distinction between data residency and technical sovereignty defines modern European defense procurement. Data residency means data stays on servers within a specific jurisdiction, such as a Frankfurt datacenter. Technical sovereignty involves controlling the entire computing stack to prevent foreign legal reach. The US CLOUD Act allows US law enforcement to access data from American companies even if the servers reside in Europe, creating a sovereignty gap that residency alone cannot fix for European defense agencies. You should understand that relying on a US-based hyperscaler even in a European region does not provide full control over the legal framework. AI Gateways manage request routing and inference execution to help meet these requirements. These gateways determine which model endpoint receives a request and which region performs inference. They also control whether prompts, responses, or logs are persisted. As AI adoption accelerates, data residency has emerged as an essential requirement. Data residency applies to where inference happens, how requests are routed, and where metadata like logs or prompts are generated. Without explicit controls at the AI Gateway layer, organizations cannot reliably guarantee that sensitive data remains within required geographic boundaries. Data can unintentionally cross borders during inference, retries, or failover. High-risk AI systems must follow strict rules under the EU AI Act starting August 2, 2026. Companies face penalties of up to 7% of global annual turnover for non-compliance.

Regulatory compliance and deployment models

AI deployment models vary based on the level of control required. Fully managed SaaS gateways provide low operational overhead but offer limited control over where control-plane data processes. Regionally deployed VPC-based gateways provide stronger residency guarantees and control over logs. On-premise or air-gapped gateways provide the maximum control over data flow and the strongest sovereignty guarantees. The Cyber Resilience Act requires manufacturers to report actively exploited vulnerabilities and severe incidents from September 11, 2026. The EU AI Act prohibits certain practices, such as social scoring or manipulative AI tools, starting February 2, 2025. General purpose AI models face obligations around classification and systemic risk starting August 2, 2025. High-risk systems intended for use by public authorities must comply by August 2, 2030. The AI Act provides for a 24-month transitional period for many provisions.

Does the current regulatory environment favor hardware manufacturers over software providers?