DFT's Expanding Role in Software-Driven Silicon: From Structural Test Closure to System-Assurance Evidence

The 2026 Siemens EDA and Wilson Research Group Functional Verification Study includes new DFT questions that provide a current-state view rather than longitudinal trend data. The individual practices reflected in the data are not necessarily new, but the results quantify them within a common industry study and, when read together, suggest that DFT verification is increasingly connected with mainstream verification infrastructure, software-driven diagnostics, and broader silicon-lifecycle evidence. This expansion is consistent with the broader shift toward software-driven, operationally complex, and assurance-sensitive silicon described in the companion paper Five Signals from the 2026 Functional Verification Study.³ In that environment, scalable DFT validation and functional fault grading can connect manufacturing-test coverage with software-driven diagnostics, in-system test, and broader silicon-quality evidence.

Introduction: DFT is Still Structural, but No Longer Isolated

For decades, the center of gravity in DFT has been clear: make internal logic controllable and observable, generate patterns that detect modeled defects, meet coverage and test-time goals, and deliver production-quality test content. That foundation remains essential. Nothing in the 2026 study suggests that scan, ATPG, compression, memory test, test access, or structural fault coverage have become less important. What may be changing is the boundary around the DFT problem. Modern test features interact with clocking, reset, power intent, hierarchical integration, secure lifecycle states, embedded processors, firmware-controlled diagnostics, multi-die assembly, and post-silicon debug.

A structurally complete DFT implementation can still create system risk if test-mode behavior, control sequences, power transitions, X behavior, access policy, package-level interactions, or recovery paths are not understood before silicon. Structural DFT itself is also evolving through more power-aware, hierarchical, physically informed, and in-system methods; the focus here is the expanding boundary around its verification and evidence.

The new DFT questions in the 2026 study are useful because they do not describe a single tool or methodology. They describe timing, integration, execution, ownership, challenges, and the use of functional patterns. The novelty is empirical synthesis, not invention of the underlying techniques: many of these practices have been documented individually, but the study provides a common dataset for examining how they appear together in current DFT verification practice.

Read together, the results suggest that DFT is becoming part of a larger evidence chain connecting pre-silicon verification, manufacturing test, software-driven diagnostics, safety and security analysis, multi-die integration, and silicon learning. Read together, they suggest that DFT is becoming part of a larger evidence chain connecting pre-silicon verification, manufacturing test, software-driven diagnostics, safety and security analysis, multi-die integration, and silicon learning. The companion papers provide a broader functional-verification context; this paper focuses on the implications for DFT.^1-3