KAYTUS announced the launch of its high-density air-cooled AI compute containerized cabinets, the EVOLEEN-I5000s. Purpose-built for AI applications, the EVOLEEN-I5000s delivers 6x higher computing density compared to traditional air-cooled cabinets. Supporting up to 50 kW of heat dissipation and achieving 25% energy savings, this innovative solution is poised to significantly enhance the efficiency of next-generation air-cooled AI data centers.

Modern AI datacenters are characterized by high-density deployments to unleash immense compute power. As demand for AI applications has surged, efficient thermal management that balances heat dissipation and high density has become essential. This growing need has driven the upgrade of existing data centers for AI adaptation, while newly built air-cooled datacenters must also be equipped to meet these stringent requirements. Furthermore, many countries are imposing regulations for environmentally friendly operations, making the construction of low-carbon-footprint datacenters a top priority for the industry.

In response to evolving industry trends, KAYTUS has launched the EVOLEEN-I5000s, a next-generation containerized computing infrastructure specifically designed for AI applications. These high-density, air-cooled AI compute cabinets offer exceptional computing power density and energy efficiency within a modular architecture. Compatible with mainstream CPUs and heterogeneous accelerators, it seamlessly integrates computing, storage, and networking functions into a single unit. Each cabinet supports 20-50kW of power, meeting up to 80% of computing and networking demands. Moreover, it is equipped with power distribution, UPS, cooling, monitoring, and fire protection systems, enabling efficient and streamlined data center construction. Engineered for modularity, efficiency, optimization, and intelligent management, the EVOLEEN-I5000s supports a broad spectrum of applications, including autonomous driving, biopharmaceuticals, GenAI, and smart manufacturing.

High Density, High Energy Efficiency, Rapid Delivery, and Intelligent Operation - All in One Solution.

High-Density Deployment and Superior Computing Power. The EVOLEEN-I5000s supports up to 40 accelerator cards and 10 CPUs in a single 47U cabinet, delivering a total computing performance over six times the performance of traditional air-cooled cabinets. This high-density design significantly reduces space and power consumption, optimizing data center efficiency. By minimizing module placement and construction space by 40% and 60% respectively, the EVOLEEN-I5000s maximizes overall operational efficiency for modern AI-driven data centers.

Efficient Heat Dissipation. The EVOLEEN-I5000s cabinets optimize cooling efficiency with a sealed cold aisle and dual-group air-conditioning control system. Simulation based on server parameters enhances heat exchange performance, allowing for better placement of cabinets and row air conditioners while preventing localized overheating. In the event of an air-conditioning failure, the dual-group control strategy maintains system reliability and energy efficiency, ensuring continuous and effective heat dissipation.

Energy Savings. The EVOLEEN-I5000s integrated system achieves over 25% energy savings compared to traditional data centers. Its fully integrated megawatt-level power chain significantly enhances power supply efficiency, reducing energy loss by more than 30%. Additionally, by utilizing high-temperature inlet water, the cooling system's energy efficiency is boosted by over 10%, resulting in a highly optimized and sustainable solution for data center operations.

Rapid Deployment. The EVOLEEN-I5000s features a modular design and de-engineered installation approach, reducing data center construction time by up to 70%. With simplified requirements for space, cooling capacity, and power allocation, data centers can be built and fully operational within just one week.

Security and Intelligent Operations. The integrated AI management platform provides real-time environmental monitoring and 2D/3D visualization, simplifying operations and maintenance. Key cooling components can be controlled remotely, and the AI-based monitoring system enhances data center stability and security by 30%, ensuring reliable and efficient operations.

Proven Solution Successfully Implemented in Large-Scale Data Center Upgrades for LLM Applications.

KAYTUS has successfully deployed the EVOLEEN-I5000s to upgrade a large-scale data center for LLM (Large Language Model) training. An energy client needed to enhance its existing data center infrastructure to support the training of an LLM with 100 billion parameters, which demanded significant computing power. The existing data center, originally designed to support 6kW per cabinet and accommodate up to 18 2U servers, faced power and space limitations when adapting to AI workloads-requiring two cabinets for a single AI server. Additionally, traditional air-cooling solutions also struggled to manage heat dissipation in cabinets exceeding 20kW, while local regulations imposed strict requirements for energy efficiency, mandating a Power Usage Effectiveness (PUE) below 1.25 for new data centers.

In response to the client's challenges, KAYTUS introduced its 42kW high-density air-cooled AI compute cabinet solution, delivering exceptional computing performance and achieving over six times the computing density of traditional cabinets. The solution maintained a PUE below 1.25 to ensure maximum energy efficiency. It successfully met the high-performance requirements of AI applications while optimizing space and energy usage in the data center, reducing module deployment by 40% and construction space by 60%. Moreover, by enabling intra-module IB networking, the solution reduced IB cable costs by 50%, providing the customer with a high-performance, energy-efficient, and intelligent infrastructure solution tailored for advanced AI workloads. The EVOLEEN-I5000s effectively addressed all these challenges with its high-density design, efficient power management, and superior cooling capabilities, making it the ideal solution for AI infrastructure upgrades.