Rapidus targets customers underserved by TSMC, but can it convince them to take the risk?

Rapidus's real bet isn't 2nm yield—it's design-service lock-in. Pulling Synopsys into EDA adaptation and Infosys into engineering outsourcing builds a full-stack delivery loop independent of TSMC. Once a customer's IP-to-silicon workflow is calibrated to this ecosystem, switching costs become prohibitive. That's far more strategic than selling raw wafer capacity. The 'underserved' framing exposes TSMC's structural blind spot: mid-volume, long-tail custom silicon gets systematically deprioritized during CoWoS capacity crunches. Rapidus is targeting this orphaned segment, not head-on competing on flagship nodes. TSMC's most probable counter isn't price cuts. It's accelerating multi-year capacity lock-ins with top-tier customers while shaving NRE thresholds on 7nm/5nm nodes—killing Rapidus's cost-advantage narrative before it matures. Samsung will likely accelerate its 2nm roadmap to avoid being squeezed between Japan and Taiwan, China simultaneously. Over 12–24 months, Rapidus almost certainly won't hit 2nm yield targets by 2026. But the 17-company design alliance is the real moat. Japan's $15B subsidy is, at its core, a fiscal lever to inject 'customer trust'—the single most expensive variable in foundry economics. That's a smarter play than Intel spending $20B on fabs and still losing customers to TSMC.