The global market for power semiconductors in AI data centers is set for explosive growth, with projections showing expansion from 2.7 billion USD in 2025 to 19.2 billion USD by 2028, driven by a rapid shift to high-voltage power architectures.
Shift to 800-volt systems drives surge
The growth is closely tied to the adoption of 800-volt high-voltage direct-current systems, designed to support the rising energy demands of next-generation AI infrastructure. These systems are becoming essential as traditional 400-volt alternating-current setups struggle with efficiency losses and heat generation.
Current systems operate at just 85 to 88 percent efficiency, with multiple conversion stages each losing small amounts of energy. In practical terms, a 100 kilowatt server rack can waste around 15 kilowatts as heat, increasing cooling requirements and operational costs.
By contrast, doubling voltage to 800 volts cuts electrical current in half and reduces copper losses significantly. This redesign also increases semiconductor value density from 175 USD to 260 USD per watt by introducing more high-voltage conversion components.
Rapid expansion of AI data center capacity
Global AI data center capacity is expected to grow sharply, with an estimated 81 gigawatts added by 2028. This includes 63 gigawatts of new facilities and 18 gigawatts from replacement projects.
AI chips alone are projected to account for about 54 gigawatts of demand, with additional load coming from networking equipment and cooling systems. Future rack designs, such as 600 kilowatt configurations expected to roll out around 2027 to 2028, will further intensify power requirements.
Semiconductor demand shifts toward advanced materials
The transition to high-voltage systems is reshaping demand for semiconductor materials. Silicon carbide content per watt is expected to double, while gallium nitride could see a sharp rise in usage in mid-level converters. Silicon will remain dominant in voltage-regulator modules, maintaining the largest share of value.
Manufacturers including Infineon, Texas Instruments, STMicroelectronics, Navitas, Analog Devices, and Wolfspeed are positioning themselves across different parts of the power conversion chain, from grid-level interfaces to chip-level regulation.
Recent performance reflects this trend. Texas Instruments reported a 90 percent increase in data center-related revenue in early 2026 and introduced an 800-volt power solution aligned with Nvidia’s upcoming reference designs. Infineon, identified as a leading contender in the segment, has increased capital spending and expects AI-related revenue to reach 2.5 billion euros by 2027.
Infrastructure bottlenecks could slow deployment
Despite strong momentum, infrastructure constraints remain a key risk. Power grid upgrades in the United States typically take three to five years, while data centers are often completed within two years, creating a mismatch that could delay deployment.
Regulators are moving to address the issue. In mid-2026, U.S. authorities introduced measures to accelerate grid connection timelines, targeting approvals within 90 days. However, these changes may require operators to share upgrade costs and reduce energy usage during peak demand periods.
How effectively these policies are implemented, along with utilities’ ability to manage growing connection queues, will play a critical role in determining how quickly next-generation AI data centers come online.
For deeper context on AI’s impact on digital assets, explore our guide on Web3, AI and crypto convergence.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
