Skip to content
hw.dev
hw.dev/signal/onsemi-fab-right-divest-tarlac-mountain-top-silex-2026
Signalonsemi

onsemi Divests Two CMOS Fabs in Fab Right Push to Concentrate on SiC and GaN

onsemi is shedding commodity CMOS capacity -- Tarlac to Greatek, Mountain Top to Silex -- to concentrate manufacturing investment in silicon carbide and GaN, a clear signal that specialized process nodes are stratifying away from general-purpose fabs.

#semiconductor#manufacturing#supply-chain
Read Original

onsemi is selling its Tarlac, Philippines assembly site to Greatek Electronics and its Mountain Top, Pennsylvania fab to Silex Microsystems. The company frames this explicitly as a "Fab Right" strategy: shed volume commodity manufacturing, direct capital toward the process nodes where onsemi actually differentiates -- silicon carbide for EV powertrains and GaN for power conversion. Expected savings are $35M per year once the Mountain Top transaction closes in January 2028.

The interesting part is not the cost savings. It is the admission, in press-release language, that general-purpose CMOS back-end capacity is not where onsemi wants to compete. Tarlac and Mountain Top handle packaging and commodity IC work. That work is going to an OSAT player (Greatek) and a specialized MEMS foundry (Silex). onsemi keeps long-term supply agreements with both sites, so the capacity doesn't disappear -- it just becomes someone else's problem to optimize.

For hardware teams building EV inverters, motor drives, or high-frequency power supplies, the signal is that onsemi is making a focused bet. Compound semiconductor roadmaps -- SiC yield improvements, GaN voltage scaling, packaging advances for high-temperature operation -- are where onsemi's remaining capex goes. The fabs doing commodity silicon are structurally not part of that story. If you are designing around onsemi silicon carbide or GaN, the supply chain is getting more focused, not broader.

The strategic loser in this move is not immediately visible. But any design team that chose onsemi for broad-range CMOS availability -- rather than for the compound semi specialization -- is now supply-chain-dependent on a vendor that has explicitly declared it is not optimizing for that use case. Qualify a second source before 2027 or re-evaluate the BOM.