Tech news in 3 minutes

ChipAgent Raises $130 Million and Why it Matters

11 h ago

At DAC 2026, Undo showcased technology that provides AI-driven chip design tools with critical runtime context, enabling evidence-based root cause analysis up to 100x faster and reducing AI hallucinations—addressing a key bottleneck for semiconductor verification and debug workflows. Undo’s deterministic recording captures every variable, event, and I/O across a program’s execution, allowing coding agents like Cursor, Claude, and Copilot to reason about dynamic behavior rather than relying on static logs or generic LLM intelligence. CEO Dr. Greg Law explained that this runtime context transforms AI from a “clever guesser” into a reliable problem-solver for complex C/C++ and SystemC codebases. Two use cases illustrate the impact: GPU model architects can diagnose functional defects faster, boosting simulation coverage from 50–60% to higher levels and improving chip quality; system-level verification teams—where up to 70% of design investment goes into verification and 35% of total effort into debug—can reduce debug time by enabling AI to autonomously identify root causes across test benches, fast models, software loads, and RTL hardware. Undo’s approach directly addresses the gap between model intelligence and runtime context, which the company identifies as the current bottleneck for scaling AI in chip design.