The startup Verkor.io has marked a milestone in the semiconductor industry by using its AI agent system, Design Conductor, to design a complete processor core based on the RISC-V architecture. The process, which started from a document of just 219 words, generated a verified, fabrication-ready schematic in only 12 hours, a speed that surpasses traditional timelines of 18 to 36 months for commercial chips by several orders of magnitude.
Accelerated design flow: from text to wafer in half a day ⚡
The Design Conductor system operates as an autonomous agent that interprets technical specifications in natural language and translates them directly into hardware descriptions (RTL). In this case, the input document contained functional instructions for a 32-bit RISC-V core, including the control unit and data path. The AI executed tasks of logic synthesis, formal verification, and gate optimization in record time. To contextualize this advancement in the field of 3D microfabrication, let's imagine a visualization of the flow: from the initial text, through netlist generation, to the physical placement of standard cells on a wafer. The lithography simulation, which would normally require weeks of manual adjustments, was completed in minutes thanks to machine learning algorithms that predict and correct fabrication defects. The result is a design ready for tape-out, with a transistor density comparable to chips designed by human teams over months.
Implications for the future of chip design 🔬
This achievement poses a paradigm shift in the industry. While human design remains crucial for high-level architecture and innovation, AI demonstrates that technical execution can be almost fully automated. For semiconductor manufacturers, this means shorter development cycles and reduced costs, especially for prototypes and application-specific chips. In the context of 3D modeling, the ability to generate and verify complex schematics in hours allows iterating on physical designs with unprecedented agility, accelerating the transition towards advanced lithography processes and 3D packaging. The question is no longer whether AI can design chips, but how to integrate this speed into the global semiconductor supply chain.
What implications does the fact that an AI can design a RISC-V core in 12 hours have for the future of 3D microfabrication, considering that three-dimensional chip integration usually requires extremely optimized architectures and much longer development timelines?
(PS: at Foro3D our favorite lithography is printing layers of filament)