Power Delivery Becomes Analog Devices’ New Bet in AI Infrastructure

Power delivery is becoming a core constraint in artificial intelligence infrastructure, and Analog Devices’ agreement to buy Empower Semiconductor for about $1.5 billion shows why. The deal, announced on May 19, gives ADI a more direct position in one of the less glamorous but increasingly strategic layers of the AI buildout: getting dense, fast, efficient electricity to the processors doing the work.

The transaction also stands out because it is not simply another scale play in semiconductors. ADI is buying a specialist whose integrated voltage regulators and silicon capacitor technology are designed to reduce the energy footprint, component count and physical complexity around AI processors in data centers, according to the companies and Reuters. That makes the deal a useful signal about where future AI infrastructure spending may concentrate as compute demand rises.

Power Delivery Has Become a Front-Line AI Constraint

For much of the past two years, the AI conversation has centered on model capability, access to advanced GPUs and the availability of data center capacity. Those issues still matter, but the engineering bottlenecks have moved deeper into the stack. Once more processors are packed into tighter spaces, the question becomes how to feed them stable power quickly enough without adding too much heat, space or system cost.

That shift helps explain why a niche company such as Empower could become strategically valuable to a much larger analog and mixed-signal chipmaker. Rather than competing head-on in the most crowded part of AI silicon, the deal focuses on the electrical architecture that determines how well expensive processors actually perform in production environments.

Why Power Delivery Moves Closer to the Chip

Empower has spent years developing integrated voltage regulator technology aimed at vertical power delivery, a design approach that pushes power conversion physically closer to the processor. In practical terms, that matters because AI servers are becoming denser, more power-hungry and less tolerant of wasted energy between the source and the point of compute.

ADI said Empower’s architecture can shrink footprint, reduce height and cut component count while improving power density, efficiency and signal integrity. Empower separately markets its platform as a way to support kilowatt-class AI and high performance computing systems, where even modest improvements in power efficiency can translate into meaningful savings at rack and fleet scale.

The strategic appeal is straightforward. If leading AI chips continue to improve faster than the surrounding electrical systems, data center operators risk paying ever more for compute that cannot run at full potential. That makes power delivery less of a support function and more of a performance enabler.

The Bottleneck Extends Beyond Chip Supply

Reuters framed the transaction around a growing industry recognition that power bottlenecks can curb AI throughput. That matters because it broadens the competitive map. The AI race is no longer only about who can design the fastest accelerator or secure the most advanced wafers. It is also about who can remove the friction points that limit deployment once those chips arrive inside servers.

In that sense, the acquisition fits a wider pattern across AI infrastructure. Investors and operators are now spending heavily not only on processors, but also on packaging, interconnects, cooling, memory, networking and electricity. Power delivery belongs in that group because it sits directly between theoretical compute capacity and real-world system performance.

The result is that companies with strong positions in enabling hardware can capture more strategic value than their revenue size alone might suggest. That is especially true when hyperscalers and AI chip designers are looking for ways to scale faster without proportionally raising operating costs.

Why Analog Devices Wants Empower Now

ADI’s move looks less like an opportunistic tuck-in and more like a targeted extension of its existing strengths. The company already sells deeply into industrial, communications, energy and data center applications, and its value proposition has long rested on solving difficult analog problems inside larger systems. Buying Empower pushes that system-level role closer to one of the most important spending waves in technology.

The timing also matters. ADI announced the deal one day before reporting fiscal second-quarter results that showed revenue of $3.62 billion, year-over-year growth across all end markets and record bookings in its business-to-business segments. It also forecast third-quarter revenue of about $3.9 billion at the midpoint, suggesting the company is entering the acquisition from a position of operating strength rather than defensive necessity.

Power Delivery Fits ADI’s Core Business

Unlike companies trying to break into AI by branding ordinary products around the trend, ADI already has a credible technical and commercial pathway into the market. Its existing portfolio spans power management, signal conversion and other essential functions that sit around complex computing systems. Adding Empower gives it a more differentiated offering in one of the highest-value parts of that surrounding architecture.

The company said the acquisition would expand its next-generation high-density power portfolio for the AI era. That phrasing may sound broad, but it points to a specific ambition: to become more important not only to component buyers, but also to hyperscalers and AI silicon developers designing full platforms. In a market where large customers increasingly prefer fewer strategic suppliers, that can matter as much as any individual chip specification.

There is also a portfolio logic here. AI infrastructure customers are not buying isolated components; they are assembling systems whose performance depends on the interaction of power, thermals, signal integrity and compute density. A supplier that can address more of those interdependencies has a better chance of defending pricing and expanding account share.

Deal Terms Signal a Strategic Bet

Under the agreement approved by both boards, ADI will pay Empower stockholders $1.5 billion in cash. The companies said the deal is expected to close in the second half of calendar 2026, subject to customary conditions and the waiting period under U.S. antitrust law. That timeline suggests a manageable regulatory path, but it still leaves execution and integration work ahead.

One notable detail is that Empower chief executive Tim Phillips is expected to continue leading integrated voltage regulator technology efforts after closing. That points to the value ADI sees in retaining not only the intellectual property, but also the technical leadership and product roadmap behind it. In acquisitions built around specialized engineering, talent continuity can be as important as the purchase price.

The structure of the deal also shows that ADI is willing to pay for positioning before this corner of the market fully matures. If power delivery becomes a standard area of competition in AI servers, buying earlier can be cheaper and strategically cleaner than trying to replicate the capability internally after customer designs have already formed around another vendor.

What the Acquisition Says About the AI Infrastructure Market

The larger lesson from the deal is that AI economics are pushing value toward previously underappreciated parts of the hardware stack. Much of the market still talks as if data center competition begins and ends with GPUs, but the operating reality is more layered. Every watt saved, every board simplified and every constraint removed can influence how quickly expensive infrastructure turns into usable capacity.

That is why the deal deserves attention beyond ADI’s own earnings cycle. It signals that the next phase of AI competition may reward companies that solve narrow but consequential hardware problems. Those companies may not always have the loudest public profile, yet they can become essential to scaling the broader ecosystem.

Power Delivery Will Shape Data Center Economics

Empower’s own description of its technology focuses on lowering the energy footprint and total cost of ownership of data centers. ADI’s framing is similar: it wants to deepen its role as a system-level power partner for hyperscalers and silicon developers. Both messages reflect the same underlying reality that AI infrastructure is now constrained by economics as much as by innovation.

As racks become denser and workloads more persistent, power efficiency compounds into a business issue, not merely an engineering metric. A design that reduces wasted energy or shrinks supporting hardware can influence operating expenditure, floor layout, cooling requirements and the pace of deployment. In large AI fleets, even small percentage gains can matter financially.

That does not mean power delivery suddenly eclipses compute, memory or networking. It does mean the sector is becoming more interconnected. Winning in AI infrastructure increasingly depends on how well companies coordinate several difficult subsystems at once, and power is now firmly one of them.

Hyperscaler Programs Matter More Than the Headline Size

Perhaps the most important signal in ADI’s release was not the purchase price itself, but the claim that Empower’s silicon capacitors are already in production and that its integrated voltage regulator programs are advancing with leading hyperscalers and AI silicon providers. If accurate, that places Empower closer to commercial deployment than a headline about a specialized startup might imply.

For ADI, that matters because customer proximity can shorten the path from acquisition logic to revenue relevance. It suggests the company is not just buying an idea for future labs, but a technology set already being evaluated inside the organizations that will shape AI data center design over the next several years.

There are still reasons for caution. The companies did not disclose Empower’s revenue base, and integration always carries risk in focused semiconductor acquisitions. But even with those caveats, the transaction looks meaningful because it reflects where AI infrastructure bottlenecks are moving next and how established chip suppliers are repositioning to capture that shift.

Analog Devices is betting that the next profitable layer of the AI buildout will include the electrical plumbing around the processor, not only the processor itself. Readers can follow more technology and infrastructure coverage like this across Berrit Media.