Chiplets Get Physical: The Days of Mix-and-Match Silicon Draw Nigh (www.eejournal.com)

Cadence is set to make waves in the chiplet market with its announcement of the Physical AI Chiplet Platform, aimed at revolutionizing real-time AI processing for edge applications like autonomous vehicles and drones. This platform will be unveiled at the upcoming Chiplet Summit 2026, where David Glasco, the VP of Cadence's Compute Solutions Group, will discuss modular multi-die designs that integrate various chiplets into a cohesive system. Significantly, this approach not only addresses the growing interest in "physical AI" – which emphasizes AI that operates in the real world rather than in software – but also aims to standardize interactions between chiplets to ultimately allow for a more "mix-and-match" ecosystem. The platform seeks to streamline the chiplet design process by promoting a spec-driven framework that incorporates CPUs, AI accelerators, and customizable domain-specific chiplets, all interconnected through standardized technologies like UCIe. This shift from proprietary multi-die solutions towards interoperable systems is poised to foster collaboration across the semiconductor landscape. With partnerships established with companies like Arm and Samsung Foundry, Cadence aims to reduce the risks associated with chiplet integration while accelerating customers' time to market. Overall, this initiative not only reflects Cadence’s evolving role from a traditional electronic design automation vendor to a vital player in the burgeoning chiplet economy but also signifies a pivotal move towards a more accessible and versatile silicon future.