GlobalFoundries Bets on Body Biasing: FDX Fusion Targets Physical AI Without EUV - Tech Times

GF's FDX Fusion is a strategic reframe, not a technical retreat. By reviving Body Biasing—a technique standard at 90nm—to serve Physical AI workloads, GF shifts the value anchor from "smallest node" to "highest I/O bandwidth per watt." The battleground moves from ASML's EUV monopoly back to design and advanced packaging. Upstream, DUV multi-patterning yield management and CoWoS-class packaging gain outsized weight. Downstream, humanoid-robot and edge-inference customers see 15-25% BOM reduction, at the cost of longer verification cycles. On compliance, sidestepping EUV is itself a supply-chain hedge: capacity can lean on DUV lines in Taiwan, China and Malaysia, reducing single-source equipment risk amid escalating export controls. Competitively, TSMC won't follow—its N2/A16 narrative is built on "most advanced node." The real shock lands on Intel, whose 18A/14A pitch already claims non-EUV competitiveness. GF is stabbing in Intel's backyard. ASML's EUV premium gets structurally diluted in the edge-AI incremental market. 12-24 month outlook: expect at least two second-tier foundries (UMC, Samsung) to launch similar EUV-free AI processes. Physical AI silicon shifts from GPU-downsizing to dedicated ASICs; that niche could exceed $40B by 2027. Most critically, Body Biasing's revival forces EDA vendors to rebuild power-performance optimization algorithms—an impact far deeper than any single chip.