AI infrastructure is increasingly becoming a manufacturing problem as much as a chip or software problem. Compal Electronics (TWSE: 2324) has opened its new Daxi AI Server Manufacturing Center in Taiwan, investing NT$4.03 billion in a facility designed to expand AI server and rack-level production. The company expects the site to begin contributing revenue in the fourth quarter of 2026, as hyperscalers, cloud providers and enterprise customers continue building increasingly dense AI infrastructure.
The AI infrastructure race is moving deeper into the factory.
Compal Electronics has officially opened its Daxi AI Server Manufacturing Center in Taoyuan, Taiwan, adding dedicated capacity for AI server systems and rack-level solutions as demand for high-performance computing and AI data centers continues to climb.
The facility represents more than another manufacturing expansion for the Taiwanese electronics manufacturer. It is part of Compal’s broader attempt to reposition itself from a company historically associated with PCs and consumer electronics toward a larger role in the AI infrastructure supply chain.
Compal invested approximately NT$4.03 billion in the Daxi project, including property acquisition, facility upgrades and manufacturing equipment. The site covers roughly 7,654 square meters of land and has about 21,706 square meters of floor space.
Production will focus on L10 AI server systems and L11 rack-level solutions, with revenue expected to begin ramping in the fourth quarter of 2026. Compal says the facility will roughly double its AI server system production capacity while adding rack manufacturing capabilities.
That distinction is becoming increasingly important.
The AI server market is moving beyond individual GPU servers toward complete rack-scale systems that integrate compute, networking, power and cooling. Modern AI clusters can require tightly coordinated infrastructure because accelerator density drives substantially higher power consumption and thermal loads.
Compal is therefore emphasizing liquid cooling and rack-level system integration at Daxi. The company says its proprietary liquid-cooling technology will support higher-density deployments while improving thermal management and reliability.
The timing reflects the scale of the market opportunity. Gartner forecasts worldwide AI spending of $2.59 trillion in 2026, up 47% from 2025, with AI infrastructure—including AI-optimized servers, networking and semiconductors—accounting for more than 45% of spending. Gartner also expects spending on AI-optimized servers to triple over the next five years.
IDC’s latest server-market data shows just how much of the current infrastructure cycle is already being driven by accelerated computing. Worldwide server spending grew 30.7% year over year in the first quarter of 2026, while unit growth was only 3.3%, reflecting the much higher value of AI-oriented systems. IDC says supply constraints, particularly around memory, are becoming a limiting factor even as demand remains strong.
For manufacturers such as Compal, that creates both an opportunity and a challenge.
The opportunity is higher-value system manufacturing. Traditional server assembly can be relatively standardized, while AI infrastructure increasingly requires customized configurations, advanced thermal management and rack-level integration.
The challenge is that manufacturing capacity alone does not guarantee a position in the AI supply chain. Companies need access to accelerator platforms, networking components, memory, power systems and sophisticated cooling technologies. They also need the ability to qualify systems quickly for customers operating at hyperscale.
Compal is attempting to address that through geographic diversification.
The Daxi facility will complement manufacturing operations in Taiwan, Vietnam and the United States, giving the company a distributed production network. Compal also expects its Texas operation to ramp more fully next year.
That geographic strategy is becoming increasingly common among AI hardware manufacturers. Supply-chain resilience has moved from a secondary procurement consideration to a strategic requirement as hyperscalers and AI infrastructure providers seek production closer to major markets and reduce dependence on a single manufacturing region.
Compal’s comments about its customer base also point to another change in the AI server market.
Chairman Ray Chen said the company has expanded beyond traditional PC ODM customers into NeoCloud providers and plans to deepen its presence with hyperscaler and enterprise customers. The move reflects the widening customer base for AI infrastructure. Demand is no longer coming exclusively from the largest global cloud companies; specialized GPU cloud providers, enterprises and sovereign AI projects are also adding capacity.
That expansion creates room for manufacturers capable of producing different configurations at scale.
Compal’s strategy is also consistent with its broader product positioning. At Computex 2026, the company showcased AI infrastructure spanning dense accelerator platforms, storage, direct-liquid-cooled systems, coolant distribution units and power infrastructure. The emphasis is increasingly on delivering an integrated rack rather than simply assembling an individual server.
Liquid cooling is particularly significant as AI racks become more power-dense.
Gartner estimates that AI-optimized servers will account for 31% of global data-center electricity consumption in 2026, with their power consumption expected to surpass that of conventional servers in 2027. Global data-center electricity consumption is forecast to reach 565 TWh this year, a 26% increase from 2025.
Those numbers explain why cooling has become a strategic component of AI infrastructure rather than a facility-level afterthought.
For NVIDIA, AMD and other accelerator vendors, higher-density systems require OEM and ODM partners capable of integrating increasingly sophisticated thermal and power architectures. For cloud providers, the objective is not simply to buy more GPUs; it is to deploy those GPUs reliably within constrained power and cooling environments.
Compal’s Daxi investment therefore sits at the intersection of several industry trends: accelerating AI compute demand, rack-scale architecture, liquid cooling, supply-chain localization and the shift toward higher-value infrastructure manufacturing.
It also puts Compal into a competitive field that includes Taiwanese ODMs such as Quanta Computer, Wistron and Wiwynn, alongside global server manufacturers such as Supermicro and established infrastructure vendors. Supermicro, for example, recently projected fiscal 2027 revenue of $65 billion to $72 billion as demand for AI-optimized servers and data-center infrastructure continues to grow.
The competitive question will be how quickly each manufacturer can translate AI demand into dependable rack-scale production while managing power, cooling, component availability and regional delivery requirements.
For Compal, Daxi is an important piece of that equation. The company expects its AI server business to gain further momentum as its U.S. manufacturing footprint expands.
The broader message is clear: the next stage of the AI boom will depend not only on better models and faster accelerators, but on the factories capable of turning those components into complete, power-efficient and deployable computing infrastructure.
Market Landscape
The AI server market is shifting from component-centric systems toward rack-scale infrastructure. As GPU and accelerator density rises, manufacturers increasingly need expertise across compute, networking, power delivery and thermal management.
Compal is competing in this environment alongside Taiwan’s major ODMs, including Quanta, Wistron and Wiwynn, while companies such as Supermicro, Dell Technologies and HPE compete in broader AI server and data-center infrastructure markets.
The distinction between an AI server manufacturer and an AI infrastructure integrator is becoming less clear. Customers increasingly want validated systems that can move rapidly from design to deployment, particularly as new accelerator generations arrive.
Geography is another competitive factor. Compal’s Taiwan-Vietnam-U.S. manufacturing network gives it options for regional fulfillment and supply-chain diversification, while the company’s investment in liquid cooling addresses one of the central physical constraints facing AI data centers.
For enterprises and cloud providers, this means AI infrastructure procurement is increasingly about rack-level performance, power efficiency, cooling compatibility, deployment speed and supply-chain resilience, not simply accelerator specifications.
Top Insights
- Compal’s NT$4.03 billion Daxi investment expands AI server capacity, giving hyperscalers, NeoCloud providers and enterprises greater access to rack-scale manufacturing.
- The facility will focus on L10 servers and L11 racks, reflecting the industry’s shift toward integrated AI infrastructure rather than standalone GPU systems.
- Liquid cooling becomes strategically important as rack density rises, helping data-center operators manage the power and thermal demands of modern AI accelerators.
- Compal’s Taiwan, Vietnam and U.S. manufacturing network strengthens supply-chain flexibility, potentially helping global AI customers localize production and improve fulfillment.
- AI infrastructure spending remains a major growth engine, but manufacturers face constraints involving memory, power, cooling, accelerators and increasingly complex rack integration.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
