跳到正文
Data Center Knowledge 数据中心行业· Shane Snider·· 2026-09-04AI 评分91

「英中对照」Flex拟44亿美元收购EPC Power,加码AI数据中心800V架构

Flex Pays $4.4B for EPC Power as AI Data Centers Push 800V Architecture

AI 导读

Flex将以44亿美元收购EPC Power,为其数据中心基础设施业务增加800V直流、构网型及其他电力转换技术;交易预计2026年四季度完成,但仍需监管批准等交割条件。

正文 · 原文

Shane Snider,

Senior News Writer,Data Center Knowledge

September 4, 2026

4 Min Read

Flex will acquire EPC Power for $4.4 billion, adding 800V DC, grid-forming, and other power-conversion technologies to its data center infrastructure business, the companies said Friday. The deal is expected to close in the fourth quarter of 2026, subject to regulatory approvals and other closing conditions. Flex said EPC Power will join its Cloud and Power Infrastructure segment, which Flex plans to spin off as an independent public company in the first quarter of 2027.

EPC Power develops power-conversion systems for data centers, utility-scale energy storage, and microgrids, including grid-forming technology, digital rectifiers, and DC-DC converters, with a roadmap for solid-state transformers. Flex said EPC Power’s platform is designed to connect 800V DC loads more directly to grid voltages and can consolidate functions traditionally provided by UPS systems and AC power distribution. EPC Power said its 800V architecture is designed for next-generation AI data centers.

Related:The 200 GW Moment: Reinventing the Grid for the AI Economy

800V DC: Raising Rack Power While Cutting Losses

For data center operators, 800V DC could change how power is delivered in high-density AI facilities. Higher voltage allows more power to move at lower current, reducing conductor requirements and electrical losses. It can also reduce some of the conversion equipment between the grid and the rack.

As rack power rises, the electrical systems serving those racks become a larger part of the facility design. EPC Power’s technology targets the power-conversion portion of those systems, rather than the computing equipment itself.

Power Conversion Becomes a Strategic Layer

Neil Osnato, founder of Persistence Analytics Group, said the $4.4 billion purchase signals that power conversion is becoming a more strategic part of AI infrastructure.

“The $4.4 billion price itself is a signal that power conversion is moving from a supporting component to a strategic control point in AI infrastructure,” Osnato said.

He said higher rack and campus power densities make the way electricity is converted between the grid, on-site generation, storage, and computing loads more important.

“AI infrastructure is becoming as much a power-systems problem as a compute problem,” Osnato said.

Power Electronics Change the Grid Interface

EPC Power’s technology also targets rapid changes in power demand associated with AI workloads. Its Agile Grid-Forming platform is designed to respond to changes in milliseconds. The company said its systems can integrate energy storage and on-site generation while supporting grid connections and microgrid applications.

Related:Can Supercapacitors Smooth AI’s Power Swings?

Osnato said those capabilities can give a facility greater control over how its electrical loads appear to the grid. Power electronics and storage can smooth rapid changes in demand, he said, while converters can help manage the electrical behavior seen at the grid interface. But he cautioned that the capabilities should not be treated as a substitute for addressing the underlying requirements of a large load.

“Load smoothing is not load elimination,” Osnato said.

For utilities evaluating large AI facilities, Osnato said the relevant information goes beyond a site’s stated maximum demand. Utilities should consider ramp characteristics, power-electronic controls, storage capability, on-site generation, ride-through behavior, grid-forming and grid-following modes, harmonics, and fault behavior. He added that utilities must verify claimed capabilities during interconnection. “Declared capability is not demonstrated capability,” Osnato said.

For example, if a developer says a large campus can smooth its load, ride through disturbances, or provide grid support, utilities need evidence showing what the capability is, under what conditions it works, how long it lasts, and what changes would require it to be revalidated, he noted.

Related:Efforts to Curb Data Center Speculation Gain Ground Across the US

EPC Scales US Manufacturing to 27-40 GW

The acquisition follows EPC Power’s expansion of its US manufacturing footprint. In July, the company opened a 167,000-square-foot manufacturing facility in Fountain Inn, South Carolina. EPC Power said the facility nearly tripled its production capacity, with 27 GW of annual capacity initially and the ability to scale to 40 GW. The facility produces M and MRACK series power inverters for grid, storage, and AI data center applications and is expected to create about 275 jobs.

EPC Power said its annual US manufacturing capacity is expected to exceed 30 GW in 2027.

Deal Sets Up Q1 2027 Spin-Off Plan

Flex said EPC Power is expected to generate about $800 million in revenue during calendar 2026. Flex also said it expects organic revenue growth of about 40% in 2027 and an EBITDA margin of approximately 30%. The acquisition will be financed through a combination of debt and equity.

Goldman Sachs Alternatives and Cleanhill Partners acquired a majority stake in EPC Power in 2022, following an initial investment in 2021.

About the Author

Shane Snider

Senior News Writer, Data Center Knowledge

Shane Snider is Senior News Writer at Data Center Knowledge, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. His reporting focuses on the operational, economic, and environmental forces reshaping digital infrastructure, including AI factories, utility constraints, liquid cooling, renewable energy procurement, and next-generation data center architectures. He has won recent Azbee awards for news series and government reporting. Based in Raleigh, North Carolina, Snider covers how hyperscalers, utilities, chipmakers, and infrastructure providers are responding to the rapid rise of AI workloads and global compute demand.

You can reach Shane at [email protected] or on LinkedIn.

来源:Data Center Knowledge 数据中心行业 · datacenterknowledge.com