SAN JOSE, California: Analog Devices said on May 19, 2026, that it will acquire Empower Semiconductor for roughly $1.5 billion in an all-cash transaction, expanding its push into the chips that feed AI accelerators. The deal was disclosed in an 8-K filing with the Securities and Exchange Commission and a joint press release from the two companies. Both boards have approved the deal, which is expected to close in the second half of calendar 2026 subject to Hart-Scott-Rodino antitrust review.
Empower, based in Silicon Valley, builds integrated voltage regulators, known as IVRs, and silicon capacitors. Those chips sit physically next to GPUs and AI accelerators and handle the last stage of power delivery, converting and routing energy at the point of consumption. ADI is folding the technology into what it calls a grid-to-core power platform, a portfolio meant to cover everything from utility-scale conversion at the data center fence to power management on the silicon die itself.
Why ADI Is Paying for Power Delivery
The acquisition is ADI’s largest move into AI infrastructure to date and reflects a shift in where chipmakers see scarcity. Through 2024 and 2025, hyperscalers and the Magnificent Seven competed primarily on GPU supply. By early 2026, the binding constraint had moved downstream. Power density, not raw wattage, became the limit on how much compute can be packed into a single rack.
“AI infrastructure is fundamentally reshaping how power must be delivered, with energy now the most persistent constraint to scaling next-generation systems,” ADI chair and chief executive Vincent Roche said in the company’s announcement. Roche framed the deal as a way to help customers “rearchitect their power systems and achieve the compute densities next-generation AI demands,” and noted the technology applies “well beyond AI data centers to any domain where energy constrains what is possible.”
Empower has been moving in that direction commercially. Earlier this year the company announced a collaboration with Marvell Technology to develop integrated power solutions for Marvell’s custom silicon platforms, the kind of accelerator chip that hyperscalers are designing in-house for AI workloads. Empower’s flagship Crescendo IVR series is engineered to be roughly five times smaller than traditional board-level designs, with faster transient response and higher efficiency, according to Empower’s product documentation.
What does the ADI Empower acquisition mean for AI infrastructure?
The deal signals that AI capital is now flowing downstream from GPUs into the picks-and-shovels layer beneath them. Power delivery, cooling, and on-die conversion are emerging as the new bottlenecks for hyperscalers, and ADI is paying a premium to own that layer rather than license it. Founders in adjacent categories should expect more strategic M&A on the power-electronics side over the next 12 months.
Empower chief executive Tim Phillips described the company’s mission as solving “the hardest problem in AI power delivery.” That framing now becomes ADI’s positioning. The combined entity will own intellectual property across the full power path, which matters because hyperscalers increasingly buy power architecture as a system, not as discrete components. Marvell, Nvidia, and the in-house silicon teams at Amazon, Google, and Microsoft each design their accelerators around assumptions about how power will reach the die. ADI is now selling into that conversation with a single integrated stack.
The financial structure also says something about ADI’s confidence. An all-cash $1.5 billion deal from a company with a market capitalization north of $110 billion is not a financing stretch, but it is a clear capital allocation choice. ADI is funding it from cash on hand rather than issuing stock, which suggests management views Empower’s pipeline as a near-term contributor to revenue rather than a speculative bet.
What Founders Should Watch Next
The first signal to track is regulatory. The Hart-Scott-Rodino waiting period will run through the summer, and antitrust enforcement of vertical chip acquisitions has tightened since the Justice Department’s review of recent semiconductor deals. ADI’s filing language emphasizes “customary closing conditions,” but a second request from the Federal Trade Commission would push the close into late 2026 or early 2027.
The second signal is competitive response. Texas Instruments, Infineon, and STMicroelectronics all compete with ADI in power-management chips and have not yet made a comparable bet on integrated voltage regulators. Whether any of them moves on Empower’s smaller competitors, including Vicor and ferroelectric-capacitor specialists, will indicate how quickly the rest of the power-electronics market follows ADI into AI-specific architectures.
The third signal is adjacent. ADI’s deal lands two weeks after Cowboy Space raised $275 million to build orbital data centers and one week after Kevin O’Leary’s nine-gigawatt Stratos project drew renewed scrutiny. Each of those bets reflects the same underlying thesis. The AI buildout has shifted from “we need more chips” to “we need to get power to the chips we already have.” Expect the M&A flow, the venture capital, and the policy attention to track that shift over the rest of 2026.
