At Davos, during the World Economic Forum, NVIDIA’s CEO Jensen Huang emphasized a point that’s reshaping industry priorities: Artificial Intelligence is not just about chips and models, but about deploying infrastructure that depends, layer by layer, on fundamental elements like power and electrical systems. He described AI as a “platform” composed of multiple layers that must scale simultaneously, starting with energy and progressing through chips, cloud data centers, models, and applications.
This narrative shift—from “who has the most powerful GPU” to “who can supply and cool the next-generation data centers”—is raising the profile of companies specializing in power electronics and thermal management. Among them, Delta Electronics, based in Taiwan, is increasingly prominent in market discussions for its role in power supplies, high-efficiency electrical distribution, and cooling solutions for data centers.
From Chips to Outlets: The New AI Economy
The advancement of generative AI and large-scale training and inference systems is driving a known industry phenomenon: data centers are now planned not just by square meters or rack counts but by available electrical capacity, grid limitations, and cooling feasibility. As a result, the “value” of infrastructure shifts toward providers capable of solving what was once an almost invisible chapter: delivering more power, reliably and with minimal losses, within the data hall.
This dynamic is evident both in the public discourse of big tech companies and in the technical specifications that are beginning to dominate the debate: more centralized power architectures, designs optimized for higher density, and a greater emphasis on liquid cooling—not only as a “premium” option but as a necessity for intensive AI workloads.
Delta Electronics: The Supplier Aiming to Cover “From Grid to Chip”
Delta positions itself as a global player in power solutions and thermal management, offering a range from power components to data center infrastructure. Its corporate materials highlight the advantage of integration: combining expertise in power electronics with cooling capabilities to develop more efficient solutions for high-density environments.
The clearest indicator of why the market looks to Delta is the growing significance of AI in its business lines. According to Taiwanese media reports, during the first three quarters of 2025, AI-related power supplies accounted for 23% of the company’s total revenue, with PSU units for AI servers representing approximately 50% of its server power supply business during that period.
Simultaneously, the company has aligned its roadmap with trends matching AI’s power demands: advanced power solutions, high-voltage products, and heat dissipation technologies as a commercial lever for 2026.
High Voltage and Efficiency: HVDC and the Shift to 800 V as Industry Trends
A technical debate filtering into mainstream media involves HVDC (high-voltage direct current) for data centers, driven by the need to reduce losses and simplify energy conversion amid increasing consumption. Delta has provided concrete proposals in this area.
Its technical documentation for the OCP (Open Compute Project) ecosystem outlines solutions for ±400 VDC environments, with scalable systems capable of powering IT loads up to 2.4 MW and efficiencies up to 98% in its “in-row” power systems.
Additionally, Delta has announced that it is developing products based on a 800 V DC structure, with mass production plans in the second half of 2026, aligned with the evolution of power architectures for major cloud providers.
The industry’s takeaway is straightforward: if AI is about infrastructure expansion, competitive advantage lies not only in computational performance but also in delivering power efficiently, managing heat effectively, and ensuring reliable 24/7 operation.
Liquid Cooling: From “Luxury” to Necessity for AI
Another major vector reshaping the value chain is cooling. Delta claims to have a portfolio that includes air and liquid cooling solutions focused on enterprise data centers, colocation, hyperscale, and HPC/AI workloads, leveraging its scale as a fan manufacturer and its expertise in energy management.
At OCP 2025, the company showcased coolant distribution units (CDUs) capable of 300 kW in liquid-to-air configurations, as well as in-row liquid-to-liquid systems with up to 2 MW capacity for new-build AI data centers, alongside other components like cold plates and rack-mounted CDUs.
In simple terms, this reflects the market’s trajectory: as densities and energy costs rise, cooling ceases to be a “operational expense” and becomes a design factor that constrains what can be deployed and where.
Market Shift: Why “Power” Is Now Part of the AI Narrative
Huang’s comment about the “largest infrastructure deployment in history” is more than just a headline. It’s a strategic investment guide. If AI drives the construction of more data centers, network upgrades, and energy efficiency initiatives, the focus shifts to companies that supply what the ecosystem needs to keep computing going: power conversion, distribution, backup, cooling, and control.
Within this context, Delta competes in a category that could be described as “AI’s peaks and valleys”: it doesn’t produce GPUs but supplies system parts without which GPUs are useless. And where energy availability and cooling capacity limit deployment, advantage is no longer measured only by “more compute” but by “more compute per watt” and stable operation capabilities.
The challenge is also well-defined: evolving electrical standards, efficiency pressures, global competition, and the need to operate at industrial scale in a supply chain that, in AI, is under strain on multiple fronts.
Frequently Asked Questions
Why is HVDC (800 V) so frequently discussed in AI data centers?
Because adopting high-voltage DC architectures aims to reduce conversion losses and facilitate power delivery in high-density environments where electric efficiency becomes a critical factor.
What is a CDU (Coolant Distribution Unit), and why is it key for AI?
A CDU manages the liquid cooling loop (flow rate, heat exchange, distribution). In AI workloads, it helps evacuate heat more efficiently than air, allowing higher densities and improved thermal stability.
How does electrical availability impact AI growth?
The ability to supply sufficient and stable power can become a bottleneck for new data centers. That’s why electric systems and energy efficiency are increasingly central to infrastructure planning.
Why is Delta Electronics gaining attention with the AI boom?
Because its business involves critical data center components—power supplies, high-voltage systems, and cooling—and it reports a growing share of products linked to AI in its revenue mix.
via: Delta