The AI Power Boom, LITEON embraces opportunities for rapid development
LITEON Technology is actively capitalizing on the AI boom by restructuring its data center infrastructure around megawatt-scale power, rack and liquid cooling. The company has developed high-density 800 VDC power architectures and advanced cooling systems to support massive, next-generation AI accelerators.At COMPUTEX 2026, LITEON hosted the "AI Summit Panel" on the opening day, themed "Powering the AI-driven Future: Scaling AI Across Architecture, Systems, and Infrastructure." Moderated by Colley Hwang, Chairman of DIGITIMES, the Panel brought together industry leaders from NVIDIA, Infineon and GIGABYTE to examine how AI is reshaping the technology landscape. The discussion highlighted how rapid AI adoption is driving structural changes brought across the ecosystem, particularly in data center infrastructure and power demand. As AI scales, power availability and efficiency are becoming central considerations, with AI factories emerging as a defining paradigm in the next phase of industry development.Hwang opened by noting that AI has become a key driver of global productivity growth and is expected to generate trillions of dollars in economic value. As a result, infrastructure and deployment models are being fundamentally reshaped, with data centers rapidly evolving into AI factories. In this transition, the availability and reliability of energy have emerged as critical enablers of large-scale deployment, elevating power and cooling from supporting functions to core elements of system architecture.More than just chips, AI revolution requires huge energyNVIDIA Senior Director, HPC and AI Hyperscale infrastructure Solutions, Dion Harris spoke first. He referenced NVIDIA CEO Jensen Huang's speech in the NVIDIA GTC Taipei held in the previous day, addressing the "five-layer cake" analogy to illustrate AI from being just a software tool into a vast, vertically integrated industrial system. At the very bottom of the cake is "energy". Before the AI can showcase real-time intelligence, the systems fundamentally rely on efficient and stable power supply."Building AI factories goes beyond compute-you have to power them, cool them, and run them reliably to truly monetize AI," said Dion Harris at NVIDIA. "Infrastructure and energy play a critical role in determining token cost and performance per watt. As Jensen Huang's‘five-layer cake’ illustrates, without strong foundational layers, higher-level innovation cannot scale. This is why ecosystem collaboration is essential."Harris further mentioned NVIDIA DSX platform as a blueprint for AI factory development, defining how next-generation infrastructure is designed, built and operated. Under this framework, AI factories are evolving to support rapidly growing token demand driven by today's AI applications. He emphasized that achieving optimal token cost and performance per watt ultimately relies on strong ecosystem collaboration, noting that partners like LITEON play a critical role in delivering the power, cooling, and operational capabilities required to support frontier and open-source AI workloads at scale.Dr. Sergio Rossi, Vice President of Application Marketing at AI and Power division of Infineon Technologies, noted that AI adoption accelerates, the key competitive bottleneck is shifting from compute to energy, especially in regions such as Europe where electricity availability has become increasingly constrained. That's the reasons why LITEON, NVIDIA, GIGABYTE and Infineon engage in collaboration to have several steps forward to make sure the design is using the right technology, architecture and solutions for securing AI development to use energy in the most efficient manner."Just improving energy efficiency by a few percentage points can translate into saving power at the scale of an entire city, highlighting how critical efficiency has become in AI data center development," said Dr. Serio Rossi.Rossi further highlighted that the AI ecosystem is facing three key complexities, including acceleration, with development cycles shortening from 30 months to 6 months; increasing architectural complexity driven by the adoption of 800 VDC in next-generation data centers; and rapid innovation in power technologies such as SiCs and GaNs to enhance energy efficiency. To address these challenges, he emphasized the importance of close ecosystem collaboration, where infrastructure stakeholders work together to anticipate future demand and align on technology roadmaps with partners such as Infineon at an early stage. This enables faster development cycles and accelerates innovation in power and infrastructure solutions for evolving AI requirements.Leo Wang, EMEA regional Product & Marketing Lead at Giga Computing, talked about how the company collaborates with industry leaders to advance AI performance through hardware innovation and liquid-cooled server solutions optimized for AI workloads. He noted that as AI infrastructure evolves, the focus is extending beyond server performance to broader data center challenges, particularly in power and thermal management.Wang added that maximizing the value generated from available energy is becoming a key industry priority, making close ecosystem collaboration essential. Leveraging Taiwan's strong supply chain and close engagement with global partners, this collaboration is helping to establish scalable deployment models and best practices for global AI adoption."We are now dealing with increasingly heterogeneous machines, from training systems to different AI applications, so power and space both need to be carefully managed, as all of these resources are limited," said Leo Wang at Giga Computing, "Everything comes down to how we build a proper data center and a proper environment to serve the server in a consistent way. It is no longer just about building servers, but about optimizing at the data center level to ensure power is allocated and utilized in an effective and efficient way."From Power to Infrastructure: Enabling the Next Phase of AI ScalingIn response to the energy and infrastructure challenges highlighted in the panel, LITEON showcased its integrated capabilities across power, mechanical, and thermal systems to support next-generation AI infrastructure at COMPUTEX 2026. LITEON debuts 800 VDC liquid-cooled power rack, alongside a comprehensive portfolio of solutions, including the 110kW power shelf for NVIDIA Vera Rubin NVL72 platforms and the 280kW in-rack CDU. Together, these offerings address the growing demand for high-density and energy-efficient AI infrastructure, demonstrating LITEON's ability to integrate power delivery, thermal management, and system design to support scalable AI deployment, and reinforcing its role as a key enabler in the evolution toward megawatt-scale AI data centers.
