Infineon Buys C2i for AI Data Center Power Chips

Infineon Technologies said on Monday, August 24, 2026 that it has agreed to acquire C2i Semiconductors, a Bengaluru-based startup that designs digital power-management chips for AI data centers. Financial terms were not disclosed, and Infineon expects the transaction to close in the third quarter of calendar 2026. The deal is small in headline terms but pointed in intent: it targets one of the least glamorous and most binding constraints on the AI buildout — getting clean, fast, precisely regulated power to the processors doing the work.

The German chipmaker framed the purchase as a way to deepen its lead in power solutions for AI data centers and to establish a new center of excellence for digital power technologies in India, folding C2i’s engineering team into Infineon’s global power business.

What C2i builds

C2i specializes in software-defined multiphase controllers and smart power stages — the components that sit between a rack’s power supply and a compute chip and manage the final, delicate step of delivering electricity at exactly the right voltage and current. In a modern AI server, that step is brutal: an accelerator can swing from near-idle to full draw in microseconds, and the power-delivery network has to respond instantly without sagging or overshooting, or the chip throttles, errors, or trips.

The startup’s distinguishing bet is putting software control into that layer. Traditional voltage regulators are largely fixed-function analog designs; C2i’s approach makes the controller programmable, so the power system can adapt in real time to the fast, shifting load profiles that AI workloads generate. That flexibility is worth increasingly more as accelerators grow hungrier and their current spikes grow sharper.

C2i’s portfolio also includes work on vertical power delivery, including Substrate Integrated Voltage Regulators (SIVR) — an architecture that moves voltage regulation physically closer to the processor, integrating it into the package substrate rather than placing it out on the board. Shortening the distance current has to travel cuts resistive losses and improves how quickly the system can react to load changes. As per-chip power climbs toward and past the kilowatt range, moving regulation under the die is one of the few levers left to keep delivery efficient.

Why power delivery is the new bottleneck

For most of the AI hardware story, attention has fixed on the processors themselves — GPUs and the scarce high-bandwidth memory stacked beside them. But as clusters scale, the surrounding infrastructure has become the binding constraint. Getting power in and heat out now shapes what a data center can physically hold, and both are straining against the limits of older designs.

The numbers driving this are stark. Individual accelerators now draw enough current that delivering it efficiently — without wasting a meaningful fraction as heat in the conversion stages — is a first-order engineering problem, not an afterthought. Every watt lost in power conversion is a watt that must also be cooled, compounding the strain on facilities already wrestling with the economics of AI data centers. That is why the same buildout has pulled in adjacent specialties, from advanced liquid cooling to on-package power regulation.

Infineon’s pitch is that it can pair C2i’s digital-power expertise with its own scale and its broad portfolio of power semiconductors — spanning silicon, silicon carbide (SiC), and gallium nitride (GaN) technologies — to build more complete power-delivery systems for AI infrastructure. SiC and GaN, so-called wide-bandgap materials, switch faster and run more efficiently than plain silicon at high voltages and power levels, which is exactly the regime AI racks operate in. The combination the company is describing runs from the wall to the die: wide-bandgap devices for the high-power conversion stages, and C2i’s programmable controllers and SIVR work for the precise final delivery to the chip.

The strategic logic

The acquisition fits a clear pattern in how incumbents are responding to the AI capital-expenditure wave. Rather than chase the accelerators themselves — a market dominated by a handful of players — Infineon is reinforcing an adjacent, essential layer where it already holds a strong position and where demand rises in lockstep with every new cluster. Power semiconductors are a picks-and-shovels business: whoever wins the accelerator race, the racks they fill still need power delivered to them.

Buying a specialist startup is also a way to acquire capability and talent quickly in a fast-moving niche. C2i brings a focused team and a differentiated software-defined approach that would take time to build in-house; Infineon brings manufacturing scale, a mature portfolio, and the customer relationships to put that technology into volume production. Adam White, who leads Infineon’s power business, said the deal will “further strengthen Infineon’s leadership in power solutions for AI data centers and create a new center of excellence for digital power technologies in India.”

That India angle is more than a footnote. Standing up a center of excellence in Bengaluru taps a deep local pool of semiconductor and systems engineering talent and reinforces India’s growing role in the global chip-design supply chain. For a European company, it is also a way to diversify where its most advanced power-design work happens. The move lands amid a broader reordering of the chip world, in which governments and companies alike are tightening their grip on strategically important silicon and the infrastructure around it.

What it means

The C2i deal is a small transaction pointing at a large shift: power delivery has become a competitive front in AI hardware, not a commodity afterthought. As accelerators grow more power-hungry and their current demands spike more violently, the ability to deliver clean, fast, efficient power right up to the die is turning into a genuine differentiator — one that affects how densely a data center can pack compute and how much energy it wastes doing so.

Who wins. Infineon strengthens a durable position in a market that grows with every AI rack deployed, regardless of which accelerator vendor is inside. Data center operators and server builders get access to more integrated power-delivery systems, which should help with the density and efficiency pressures squeezing their facilities. And India’s chip-design ecosystem gains another anchor tenant and a visible vote of confidence.

Who feels pressure. Rival power-semiconductor suppliers now face a competitor with a stronger digital-power and vertical-delivery story, and independent voltage-regulator specialists may find the biggest players consolidating the most attractive AI-adjacent niches around them. Expect more deals like this one — incumbents buying focused startups to complete their AI-infrastructure portfolios — as the capital pouring into AI compute keeps flowing downstream into every layer that touches a rack.

What to watch next. Whether SIVR and other vertical-power architectures move from roadmap to volume production, and how quickly; whether Infineon can integrate C2i’s software-defined controllers into shipping products without losing the startup’s edge; and whether power delivery starts showing up explicitly in the specifications data center operators use to choose their hardware. If it does, the least visible chip in the rack will have become one of the most contested.