End-to-end, fully factory-prefabricated modular AI Factory deployment in just a few months, with full-stack intelligent management from GPU to rack, to the entire data center.

HAMBURG, Germany: #AIFactoryManagement--KAYTUS, a leading provider in AI infrastructure and liquid cooling solutions, unveiled two core solutions at ISC 2026, designed for next-generation AI data centers: a gigawatt-scale, fully prefabricated containerized liquid-cooled data center solution and KSManage Ultra, an intelligent operations management platform for AI Factories. Together, the solutions target the escalating efficiency, performance, scalability, and reliability requirements of AI-era infrastructure, enabling faster deployment, optimized thermal management, and unified operations across large-scale AI computing environments.

The KAYTUS gigawatt-scale solution is built around a 3MW base delivery unit and scales flexibly to 1GW, reducing deployment timelines by up to 60% compared with conventional data center construction. KSManage Ultra provides full-stack visibility and proactive operations management across GPU, rack, and data center layers, helping improve resource utilization, fault detection, and operational efficiency.
At ISC 2026, KAYTUS also showcased its broader full-stack AI infrastructure portfolio, spanning compute, storage, networking, liquid cooling, intelligent operations, and performance optimization. Together, these capabilities provide European customers with end-to-end support for deploying scalable, high-density, and energy-efficient next-generation AI infrastructure.

Key Challenges Facing European AI Infrastructure Operators

As demand for AI training and inference, scientific computing, and industrial simulation workloads accelerates, cloud service providers, supercomputing centers, research institutions, and enterprises across Europe are facing increasingly complex and tangible infrastructure challenges:

• Slow deployment: AI compute capacity must be brought online rapidly, yet conventional data center construction cycles can take up to 18 months, delaying infrastructure readiness and time-to-service.
• Scaling flexibility: Building large-scale AI infrastructure in a single phase requires substantial upfront capital investment. Operators need a modular, demand-driven deployment model that allows them to start with lean capacity, scale incrementally as workloads grow, and avoid premature cost lock-in.
• Fragmented management: Modern AI data center architectures introduce substantial operational complexity. With single-rack power densities exceeding 120kW and increasingly heterogeneous hardware environments spanning GPUs, servers, switches, PDUs, and CDUs, unified infrastructure management can no longer depend on manual processes.

KAYTUS directly addresses these challenges through its latest AI data center infrastructure and intelligent management solutions, redefining both how AI facilities are deployed and how they are operated. By combining factory-prefabricated infrastructure, modular scalability, and unified proactive operations, KAYTUS enables faster deployment, flexible capacity expansion, and more efficient management of complex, high-density AI environments.

Fully Prefabricated, Gigawatt-Scale Data Center Solutions

KAYTUS’ gigawatt-scale containerized liquid-cooled data center solution standardizes the full infrastructure stack into three factory-prefabricated container modules: IT Cube, Power Cube, and Cooling Cube. Unlike approaches that prefabricate only IT racks or cooling subsystems, KAYTUS factory-builds the complete data center infrastructure as a fully integrated system. The architecture scales incrementally from a 3MW base unit to 15MW, 100MW, and ultimately 1GW, providing cloud providers, AI service providers, and supercomputing platforms with a reliable, energy-efficient, and rapidly deployable foundation for large-scale AI infrastructure.

• IT Cube serves as the standardized compute module for high-density AI and HPC workloads. In its 3MW base configuration, it integrates 18 liquid-cooled racks at 150kW per rack, together with network, storage, and server management infrastructure. Enclosed hot-aisle containment and a redundant CDU architecture enable efficient thermal control and cross-module cluster scheduling, while real-time monitoring of coolant flow, temperature differential, and pressure across all secondary-side circuits supports cost-efficient deployment, operational reliability, and remote management.
• Power Cube prefabricates the PTU enclosure and diesel generator enclosure as independent factory-built units, adopting a 2+1 redundant power architecture. All power components are factory-integrated and fully validated across defined operating conditions before shipment, supporting continuous 24/7 power delivery for high-density GPU clusters and mission-critical AI workloads.
• Cooling Cube comprises two factory-prefabricated subsystems: a liquid cooling module and an air-cooled chiller module. The design integrates redundant auxiliary equipment, a thermal buffer tank, and a long-duration emergency water reserve. All systems are factory-validated across full failure scenarios, providing short-term cooling continuity during power events and sustained cooling capacity during water supply interruptions. This architecture is designed to meet the thermal resilience and cooling efficiency requirements of high-density rack deployments.

Together, the three Cube modules deliver four core customer outcomes: accelerated time to revenue, turnkey single-vendor delivery, consistent factory-controlled build quality, and flexible incremental scaling. A 3MW delivery unit can be brought online within one month of arrival on site, while providing a clear, modular roadmap for expansion to gigawatt-scale AI infrastructure.

KSManage Ultra: Intelligent Management for AI Factory

KAYTUS also introduced KSManage Ultra, its next-generation intelligent infrastructure management platform purpose-built for AI Factory environments. KSManage Ultra delivers full-stack visibility across GPU, rack, and data center layers by integrating in-band and out-of-band system telemetry, to identify performance bottlenecks and enable coordinated, system-level operations management. The platform helps customers move beyond fragmented, component-level monitoring toward unified operational control of large-scale AI infrastructure.

• Unified management across traditional infrastructure and AI rack systems
  KSManage Ultra breaks down the operational boundaries between IT systems and physical infrastructure, as well as between component-level monitoring and full-rack management. Its multi-tier resource view spans components, nodes, racks, clusters, and data center layers, providing a consistent, end-to-end operational view across the entire AI infrastructure environment.

• In-band and out-of-band integration for proactive operations
  KSManage Ultra consolidates in-band telemetry from operating systems and drivers with out-of-band data from BMCs, firmware, and hardware logs, alongside infrastructure-level metrics. This enables cross-layer correlation across system status, hardware health, network topology, and power and cooling infrastructure. When GPU anomalies, cooling irregularities, or other operational issues are detected, the platform triggers proactive alerts, helping shift operations from reactive troubleshooting to early intervention.

• Health assessment and compute reassignment
  KSManage Ultra provides continuous node- and rack-level health monitoring, enabling automatic anomaly detection, impact analysis, and fault isolation. The platform clearly identifies units requiring replacement as well as available rack resources, allowing operators to rapidly reassign compute capacity, minimize service disruption, and maintain AI Factory business continuity.

• Automated operations at scale
  KSManage Ultra supports bulk node discovery and automated onboarding, reducing single-rack commissioning time from 50 minutes to under 3 minutes. The platform also enables one-click L10/L11 stress testing, accelerates fault root-cause analysis from hours to minutes, and supports automated rack-level initialization and configuration management. These capabilities help eliminate configuration drift, reducing the risk of performance degradation and workload failures across large-scale AI infrastructure.

Darren Cox, General Manager of KAYTUS Europe, said: "KAYTUS brings a differentiated combination of vertically integrated solutions and deep engineering expertise - spanning hardware, cluster infrastructure, management platforms, and code-level services. Underpinned by our 4D engineering expertise - Design, Develop, Deploy, and DC Operations - we can bring a 30MW large-scale data center live in as little than six months. We are committed to the European market for the long term and look forward to working with local ecosystem partners to help customers in cloud, research, manufacturing, and beyond build and scale their AI and HPC infrastructure efficiently and with confidence."

KAYTUS has deployed more than 200MW of liquid-cooled AI data center infrastructure across Europe, the United States, Japan, Southeast Asia, and other global markets. In Europe, KAYTUS delivered a hundred-rack, high-density liquid-cooled AI data center for a European infrastructure unicorn in just four months, reducing deployment time by 80% compared with conventional construction timelines. Through professional on-site tuning and system validation services, KAYTUS also achieved a 20% improvement in overall system performance, enabling a stable, efficient launch of the customer’s AI workloads.

About KAYTUS

KAYTUS is a leading provider of AI infrastructure and liquid cooling solutions, delivering a diverse range of innovative, open, and eco-friendly products for cloud, AI, edge computing, and other emerging applications. With a customer-centric approach, KAYTUS is agile and responsive to user needs through its adaptable business model. Discover more at KAYTUS.com and follow us on LinkedIn and X

Fonte: Business Wire