Flex to Acquire EPC Power, Adding Leading Power Conversion Capabilities for AI Data Centers and Grid Applications

Flex will acquire EPC Power for $4.4B (subject to adjustments), positioning the company for next-generation 800V data center power architectures as AI drives higher power densities. EPC Power is expected to generate ~$800M of revenue in calendar 2026, with organic revenue growth of ~40% in 2027 and EBITDA margin expanding by double-digit percentage points to ~30% in 2027. Flex also plans to separate its Cloud and Power Infrastructure segment into an independent company in Q1 2027, using a debt/equity mix to fund the deal.

Analysis

This is less about one acquisition than about Flex trying to move from being a capacity partner to an architectural gatekeeper in AI infrastructure. The market should focus on mix shift: if Flex can own more of the power stack, it can capture higher-margin, less commoditized content and reduce dependence on box-build economics. That creates a plausible path to multiple expansion, but only if the integration does not pull the business back toward lower-quality hardware execution and working-capital intensity.

The second-order winners are the electrical infrastructure incumbents with software/control layers and qualification depth, especially ETN and GEV, because 800V adoption expands the addressable spend pool and raises switching costs. The likely losers are smaller, single-function power component vendors and low-differentiation EMS names that lack system-level control; in an AI buildout, customers will pay up for fewer suppliers that can certify reliability end-to-end. The hidden risk is that higher integration complexity slows deployment cycles, so the revenue upside may arrive later than the stock can price today.

Near term, the stock can trade on financing details rather than strategy. If the capital raise leans too heavily on equity, dilution could offset the strategic premium; if it is mostly debt, leverage and execution risk rise into the spin timeline. Over 1-3 months, watch for the merger proxy and spin-off tax structure; over 6-18 months, the real test is whether the new platform shows sustained margin lift without customer concentration or buildout delays.

Contrarian view: the consensus may be underestimating how cyclical AI infrastructure still is. The market is treating this as secular volume growth, but the actual risk is a 2027 supply bulge if every OEM chases the same 800V thesis at once. If order conversion or backlog disclosure disappoints, the premium for integrated power systems can compress quickly, especially if the AI capex narrative cools.

Key Decisions for Investors

• Long FLEX on pullbacks over the next 1-3 months; the setup is a sum-of-the-parts rerating if spin-off mechanics are clean. Risk/reward improves if the stock gives back the initial pop before the proxy filing.
• Use a FLEX Jan-2027 call spread rather than outright common to express upside from the spin and architecture shift while capping dilution risk from the financing package.
• Overweight ETN and GEV versus broader industrials for a 6-18 month AI power-cycle beneficiary trade; they should capture incremental spend even if Flex execution slips.
• Avoid chasing the headline move until financing terms are public; if the company issues significant equity or the merger spread widens materially, that would be a signal to fade the rally.
• Watch the next data-center capex prints from hyperscaler suppliers; if AI buildout commentary weakens or 800V adoption sounds more aspirational than booked, reassess the FLEX premium quickly.