Today the battlefield is more data-driven – or digital – than ever before. A digital backbone must operate reliably in all environments, from arctic cold to desert heat, whilst supporting a continuous sense-decide-effect cycle. Real-time data from sensors and intelligence sources inform key decisions and enable the rapid deployment of platforms, personnel and autonomous assets.
Yet, despite ambitious goals and investments, the UK’s defence sector is not AI-ready. The House of Commons Defence Committee has highlighted a gap between rhetoric and reality, urging the Ministry of Defence to address this without delay. The main obstacle is not technology or skills, but infrastructure.
Why traditional infrastructure falls short
The integration of AI into defence operations requires a fundamental rethinking of mission critical and digital infrastructures; most existing defence IT systems were designed for conventional computing workloads, not the extraordinary demands of modern AI applications.
According to McKinsey research, average power densities have more than doubled in just two years, rising to 17kW per rack from 8kW, with projections suggesting this could reach 30kW by 2027 as AI workloads increase. Advanced AI training workloads may require 100-200kW per rack or more. Legacy data centres – the secure, scaleable and critical IT infrastructure that sits behind the battlefield – simply can’t deliver this density without a comprehensive redesign.
Traditional facilities also face scaleability limitations. Conventional data centre builds take 18 to 36 months from planning to initial operational capability, which is incompatible with the rapid evolution of AI technologies and the urgent pace of military operations.
For the UK military, infrastructure must not only deliver exceptional performance but satisfy stringent security requirements whilst operating in diverse, often challenging environments. The lack of infrastructure compliant with utmost security standards and frameworks such as the Surreptitious Threat Mitigation Process (STaMP), for example, presents a significant often critical vulnerability.
Equally, sensitive and critical infrastructures should be manufactured, controlled and maintained within the UK whenever possible, since in an unpredictable world, robust sovereign capabilities provide maximum security, flexibility, redundancy and continuity.
Key requirements: security, power and flexible compute
AI infrastructure in military applications also faces unique security challenges that require integration from the start. Some of the primary considerations are physical security, electromagnetic protection, Sensitive Compartmented Information Facilities (SCIF) compliance, and data sovereignty.
Computing environments must be hardened against physical threats, incorporating LPS 1175 SR2 to SR8-rated construction systems and NPSA-certified modular walling to protect against both unauthorised access and environmental hazards.
AI systems handling sensitive defence data require TEMPEST shielding compliant with NSA 94-106 and NATO SDIP-27 standards, protecting against sophisticated electromagnetic eavesdropping and cyber-physical attacks.
Solutions should be certified to meet the requirements of ICD 705 and the NPSA for SCIFs, addressing both physical and technical security standards essential for handling classified intelligence.
Sensitive defence data must remain under complete national control, limiting reliance on cloud-based alternatives and reinforcing the case for on-premises, UK-manufactured infrastructure.
Forward-operating computing environments require comprehensive security but also power resilience. AI workloads create high power demands, especially where stable grid connections are unavailable.
Additionally, several mission-critical capabilities depend directly on the availability of appropriate computing infrastructure. They include autonomous systems, intelligence analysis, and tactical cloud solutions.
Unmanned aerial, ground and maritime systems require sophisticated computing resources for real-time sensor processing, autonomous navigation and collaborative operations across multiple platforms. These capabilities demand high-performance computing close to operations.
And modern operations generate large data volumes that require AI-enabled processing for automated analysis, pattern recognition, anomaly detection, and predictive analytics. Field intelligence loses value quickly if not analysed promptly. This requires modern tactical cloud capabilities that rely on flexible computing that can move with forces, supporting distributed computing, resilient data storage, and secure information sharing across coalitions.
Bridging strategy and execution
The UK’s defence AI ambitions present a strong vision for national security. Achieving this requires infrastructure that delivers secure, high-performance computing whenever and wherever missions require.
Prefabricated modular data centres can deliver the critical link between tactical capabilities and strategic objectives. They enable rapid deployment of AI-ready environments with integrated security and resilience, forming the foundation for an AI-enabled defence sector. Modular data centres have advanced beyond previously offered containerised solutions. Today’s systems are sophisticated, enterprise-grade infrastructures engineered for secure, mission-critical applications.
A commercial-off-the-shelf approach means solutions can be designed, built, delivered and deployed at pace, ensuring a highly adaptable methodology that meets the complex and evolving needs of the digital battlespace. The use of virtual reality in the design process enables customers to explore their environment prior to deployment, fast-tracking specification whilst ensuring confidence in component security criteria.
EfficiencyIT’s ModularDC range is precision-engineered in controlled factory environments, ensuring consistent quality and comprehensive testing. These purpose-built solutions are configured for AI workloads, with integrated high-density power, advanced cooling, and advanced security features optimised for defence. Security is integrated in the core design at source, not added later. Comprehensive security integration, based on BAD (Barriers, Access, Detection) principles, ensures forward-operating SCIFs and tactical data centres can withstand increased physical and cyber threats. This includes reinforced construction, integrated monitoring, multi-factor access control, and coordinated cyber-physical protection.
The advantage for deploying prefabricated modular data centres is significant. Whilst traditional data centres require years to plan and build, the prefabricated modular solutions can be operational in 12-18 weeks. For defence organisations, this speed provides a clear strategic advantage.
By Anthony Soares, Defence Lead, EfficiencyIT
