Ericsson CEO bets on AI and in-house RAN chips as 6G, satellite competition loom

Ericsson's in-house RAN chip push is a forced pivot from systems integrator to vertically stacked player. The 2018 HiSilicon decoupling proved supply-chain sovereignty is existential, not theoretical. This move directly pressures Qualcomm and MediaTek in baseband, forcing RF front-end and optical module suppliers to renegotiate pricing power. AI inference migrating to the RAN edge means 6G base-station compute will abandon the legacy CPU+DSP paradigm for NPU-native scheduling. On compliance, Ericsson's reliance on advanced-node foundries in Taiwan, China exposes it to the same decoupling risk Huawei faced. In-house design reduces external dependency, but sub-28nm yield and cost curves will compress margins near-term. Competitively, Samsung already commands vertical integration via its own fabs; Nokia remains externally dependent. Starlink and AST SpaceMobile are eroding terrestrial RAN coverage economics—Ericsson's indoor-5G bet is defensive contraction, not offensive expansion. 12–24 month outlook: 3GPP Release 21 will lock in 6G standardization, making AI-native RAN silicon a mandatory bid criterion. Ericsson's in-house chip likely won't achieve volume replacement before 2026, with gross margins squeezed 5–8 points in the interim. Satellite-terrestrial convergence shifts from supplemental to parallel architecture.