The move to megawatt-scale AI racks is increasing semiconductor content while bringing data center architectures into onsemi's established strength in high-voltage power
The AI boom is fueling an unprecedented wave of investment in computing infrastructure, but the industry's next challenge is no longer confined to just processors and memory. As data centers race to support increasingly complex AI workloads, the ability to deliver, convert and manage dramatically more power efficiently, in less space, is emerging as a critical factor in determining performance, cost and scalability.
During onsemi's Investor Day, Achyut Shah, Group President of the Power Solutions Group, outlined how AI is reshaping the economics of data centers. He highlighted how changes in the energy infrastructure – from the grid to the data center – are increasing the amount of semiconductor content required to enable the shift to high voltage inside the data center. This change in the power architecture is built on technologies where onsemi has established proven leadership and high-value intellectual property through decades of serving automotive and industrial customers.
The New Economics of AI
Rapidly increasing power demands of AI processors are exposing the limits of traditional data center designs. As rack power advances toward megawatt scale, efficiency, thermal management, reliability and power density become as important as computational performance. Power is the scarce resource that determines how much useful compute customers can generate from available power, space and infrastructure.
“AI is turning power delivery into a strategic competitive advantage because every watt lost before it reaches the processor is a watt customers paid for but cannot use,” said Shah. “The more useful power data center operators can deliver within the same footprint, the more value they can generate from their AI deployment. Power density increasingly defines the economics of compute.”
For data center operators, higher power density also expands revenue and profitability. More efficient power delivery can enable more compute in the same rack, achieve equivalent compute with less infrastructure and real estate, or reduce resource requirements and improve overall facility operational costs. The common economic outcome is better utilization of existing infrastructure and a lower cost per unit of compute.
High Voltage Moves Into the Data Center
Meeting these demands is driving a shift from traditional low-voltage, alternating-current distribution toward high-voltage direct-current architectures. Rather than repeatedly converting power as it moves from grid through a data center, next-generation designs distribute high-voltage power deeper into the facility, reducing conversion steps and improving efficiency and power density.
The transition is already underway. Near-term power shelves and sidecars increase the use of high-voltage semiconductors, while future architectures can independently add grid-side solid-state transformers and bring 800 VDC directly into the compute tray. Each transition creates additional semiconductor content and together they form a more tightly connected, co-designed system from grid-to-core.
The same transition also affects functions that sit alongside power conversion. Energy storage, fault protection, sensing, telemetry and control increasingly move toward higher-voltage, semiconductor-based implementations. Much of that added content is concentrated in the high-voltage conversion, protection and distribution systems where onsemi has established leadership.
As a result, Shah estimates grid-to-core semiconductor content normalized per AI rack could grow nearly eightfold, from approximately $15,000 today to more than $115,000 in future architectures.
Breadth to Participate, Differentiation to Win
These high-voltage architectures require power to be managed efficiently across an enormous range – from tens of thousands of volts at the grid connection to less than one volt at the processor core. Solving that challenge spans power generation, storage, distribution, conversion, protection and voltage regulation. Because efficiency gains can be achieved at each stage, extreme co-design is required throughout the power tree to optimize entire systems rather than individual devices.
Shah pointed to onsemi's breadth of technologies, including silicon, silicon carbide (SiC), gallium nitride (GaN), advanced packaging, analog and integrated power solutions, as key differentiators. Many of these capabilities were developed for demanding automotive and industrial applications where efficiency, reliability, always-on architectures and high-voltage operation are mission-critical. That experience positions onsemi to help customers optimize power delivery across the entire power tree, from the grid to the processor.
In addition to the technology breadth, onsemi is also investing in differentiated technologies such as vertical gallium nitride (vGaN) and the Embedded Power Platform (EPP) to help solve AI power density challenges. These unique technologies help improve rack-level power densities by up to 20%, significantly reducing energy losses and cooling costs, driving improved total cost of ownership for customers.
“Breadth gets you into the conversation, but differentiation is what wins the design,” said Shah. “Customers increasingly need best-in-class high-voltage technologies, the low-voltage control and the system-level integration to work together. Few companies can bring that combination across the full power tree. By improving the total cost of ownership, onsemi’s grid-to-core power solutions expand our opportunity, enable design wins and drive above market revenue and margin growth.”
AI Is Creating a New Industrial Opportunity
While AI data centers are the immediate catalyst for innovation, the technologies being developed to support them are likely to reshape a much broader set of industries. Shah described this effect as the AI halo driving industrial growth. As semiconductor-based solutions increasingly replace traditional electromechanical systems in areas such as power conversion, energy storage and grid management, innovations originally designed for AI data centers are also becoming critical to renewable energy, energy storage systems and next-generation grids.
For onsemi, the halo effect significantly expands the opportunity, creating a substantial runway for growth. By translating its technology portfolio into scalable solutions and expanding its participation across the power tree, onsemi sees an opportunity to drive sustained revenue growth, share gains and margin expansion through the end of the decade.
“AI is defining today’s investment cycle, but the long-term opportunity is broader,” said Shah. “The technologies required to deliver the next generation of compute are also laying the foundation for more efficient factories, electrified energy systems and intelligent infrastructure.”
More Information:
About Embedded Power Platform:
Caution Regarding Forward-Looking Statements
This press release includes “forward-looking statements,” as that term is defined in Section 27A of the U.S. Securities Act of 1933, as amended, and Section 21E of the U.S. Securities Exchange Act of 1934, as amended. All statements, other than statements of historical facts, included or incorporated in this press release could be deemed forward-looking statements, particularly statements about onsemi’s long term growth strategy; forecasts about the pace, scale and direction of technological change and its impact on onsemi’s total addressable market opportunity; statements about onsemi’s ability to capitalize on such an opportunity; and the potential impact of each of the foregoing on onsemi’s operating results and financial condition. Forward-looking statements are often characterized by the use of words such as “believes,” “estimates,” “sees,” “expects,” “projects,” “may,” “will,” “intends,” “plans,” “anticipates,” “targets,” “should,” “would” or similar expressions or by discussions of strategy, plans, expectations, projections or intentions. All forward-looking statements in this document are made based on onsemi’s current expectations, forecasts, estimates and assumptions and involve risks, uncertainties and other factors that could cause results or events to differ materially from those expressed in the forward-looking statements. Certain factors that could affect onsemi’s future results or events are described under Part I, Item 1A “Risk Factors” in the 2025 Annual Report on Form 10-K filed with the Securities and Exchange Commission (“SEC”) on February 9, 2026 (the “2025 Form 10-K”) and from time to time in onsemi’s other SEC reports. Readers are cautioned not to place undue reliance on forward-looking statements. onsemi assumes no obligation to update such information, which speaks only as of the date made, except as may be required by law.
Investing in onsemi’s securities involves a high degree of risk and uncertainty, and you should carefully consider the trends, risks and uncertainties described in this press release, onsemi’s 2025 Form 10-K and other reports filed with or furnished to the SEC before making any investment decision with respect to onsemi’s securities. If any of these trends, risks or uncertainties actually occurs or continues, onsemi’s business, financial condition or operating results could be materially adversely affected, the trading price of onsemi’s securities could decline, and you could lose all or part of your investment. All forward-looking statements attributable to onsemi or persons acting on onsemi’s behalf are expressly qualified in their entirety by this cautionary statement.
