Infineon Technologies is acquiring Bangalore-based C2i Semiconductors to expand its digital power-management capability for AI data centres, adding controller and system-architecture expertise to an existing portfolio of power semiconductor devices.
The transaction is expected to close during the third quarter of calendar 2026. Financial terms have not been disclosed. C2i specialises in software-defined multiphase controllers, smart power stages, and system-level architectures for the rapidly changing electrical loads created by AI servers and high-performance computing systems.
The acquisition moves Infineon further into the control layer of the power-delivery chain. A data centre receives electricity at relatively high voltage, while processors operate at much lower voltages and increasingly high currents. Several conversion and regulation stages are therefore required between the grid connection and the computing silicon.
Those stages have become more difficult as accelerator power increases. Conversion losses that appear modest at component level become significant across a facility containing thousands of processors, while the electrical load can change quickly as groups of accelerators enter or leave computationally intensive workloads.
Multiphase voltage regulation is one method of managing those currents. Rather than pushing the complete load through one switching phase, the regulator distributes it across several phases operating in coordination. The arrangement can improve transient response, spread thermal loading, and reduce output ripple when correctly controlled.
The controller determines how those phases behave as processor demand moves. C2i’s technology uses digital control and software algorithms to adjust conversion and regulation in real time, allowing the power system to respond more precisely to changing load while balancing stability, efficiency, and device temperatures.
That complements Infineon’s existing position in power semiconductor hardware. Its portfolio spans silicon, silicon carbide, and gallium nitride devices, giving designers different combinations of voltage capability, switching frequency, conduction loss, thermal behaviour, and cost across the conversion chain.
The C2i acquisition adds more of the intelligence used to coordinate those components rather than simply another switching technology. Infineon describes its data-centre approach as extending from grid to core, covering the electrical path from the incoming power system to the low-voltage supply immediately around the processor.
C2i also brings expertise in vertical power delivery, an architecture intended to shorten the distance over which very high currents have to travel across a server board. Conventional layouts often place regulator stages beside the processor, forcing current laterally through printed-circuit-board conductors before it reaches the package.
That becomes increasingly awkward at higher processor power. Resistive losses increase with current, copper occupies valuable board area, and the electrical path itself can limit transient performance. Moving elements of power conversion closer to or beneath the processor reduces that route.
C2i’s development work includes Substrate Integrated Voltage Regulators, or SIVR, in which voltage-regulation functions can be incorporated into the substrate beneath the computing device. The technology remains part of a developing architecture rather than a universal server standard, but it addresses a physical constraint that becomes harder as AI processor power density rises.
The acquisition comes after AI power demand has already become a major part of Infineon’s growth. The company reported record quarterly sales in its third financial quarter and has disclosed substantial customer commitments associated with data-centre power capacity.
That context makes C2i a technology-extension deal rather than an attempt to buy entry into an unfamiliar market. Infineon already supplies many of the power devices used around AI infrastructure; C2i adds control algorithms and architecture expertise intended to make those devices operate as more tightly integrated systems.
The distinction also avoids reducing the transaction to another headline about AI market growth. Data-centre electricity demand is a large commercial opportunity, but the engineering bottleneck sits in converting and regulating that power efficiently enough to reach processors without producing excessive loss, heat, or instability.
Increasing processor current places pressure on components throughout the server. Voltage regulators have to respond rapidly, interconnects have to carry more current, cooling systems have to remove the resulting heat, and board layouts have to accommodate the power-delivery hardware without displacing memory, networking, or compute components.
A software-defined controller offers another degree of optimisation, but its performance depends on the complete physical system around it. Algorithms have to be matched to particular power stages, packaging, inductors, capacitors, thermal conditions, processor requirements, and board layouts.
Bringing that work into Infineon gives C2i access to a larger device portfolio, manufacturing base, and customer network. Infineon gains a specialist engineering team in India and says the acquisition will establish a new centre of expertise for digital power technologies within its R&D organisation.
The group already employs approximately 2,800 people in India. Integrating C2i expands that presence in a technical area where product development increasingly spans semiconductor design, control software, packaging, and system-level application engineering.
The transaction still has to close and the technology then has to progress through customer design and qualification cycles. AI hardware moves rapidly by semiconductor standards, but server power systems remain reliability-critical infrastructure whose failure can take expensive computing equipment offline.
Infineon has spent years improving the efficiency and performance of individual power semiconductor devices. C2i extends that work into the architecture controlling them — an increasingly important distinction as the electrical path between the grid and an AI processor becomes one of the limiting factors in how much computing power can be fitted into a rack.
