Architect Labs says AI agents designed its Frostbite CPU end to end
Architect Labs has introduced Frostbite, a general-purpose CPU with a new instruction set that it says its AI systems discovered, designed and verified end to end, the company says. Its systems iterated autonomously over the compiler, instruction set, architecture, microarchitecture, RTL, verification and final layout, using swarms of agents working from a formal specification that served as the single source of truth for the chip.
Frostbite is built around interaction nets, a model in which a program is a graph and every computation step is local, so parallel work comes from the program itself rather than from threads or out-of-order hardware. Synthesised on a 7 nm process at 500 MHz in the final layout, one Frostbite quad reached a geometric mean of 4.59 times the performance per watt of a single core of a modern in-order superscalar CPU of similar area across 25 workloads, and 3.83 times on the 15 workloads held out from the agents' search. The company notes these are still pre-silicon power numbers; its software model of the chip at first diverged from the real design by almost 250% in the worst case before an automated recalibration.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.
Also, consider following us on social media:
More from Data Centers
Nov 22 Nscale's $35 billion US IPO faces scrutiny over ByteDance ties Nov 22 ASIO chief warns data centres and AI expose Australia to foreign interference Nov 21 South Korea's President Lee urges speed in AI race at SK and AWS data centre site Nov 21 Spotify founder Daniel Ek backs AI researcher's fusion startup Europa Fusion Nov 21 NXP launches unified robotics platform for physical AI, starting with dronesSilicon Brief
Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.
Industry analysis
2025 Global Business Services Agenda: Gen AI Takes Center Stage
This industry analysis by The Hackett Group explores the transformative impact of generative artificial intelligence (Gen AI) on global business services (GBS) in 2025. The study highlights the shift from exploration to acceleration of Gen AI initiatives, with 89% of executives advancing these projects to improve customer satisfaction, innovate products, and reduce costs. The report also discusses the challenges and strategies for successful Gen AI adoption, emphasizing the need for a technology-enabled operating model and the importance of reskilling the workforce.
Read moreYou may also like
OpenAI says its models improved AMD's flagship GPU design by up to 14%
Q.ANT unveils NPU Gen 2+ photonic processor and runs an LLM with optical Fourier transforms
FuriosaAI Raises $620 Million to Scale Inference Chip Production
HyperBench Launches AI Agent Platform for Hardware Testing
DeepX tapes out 2nm DX-M2 edge AI chip built for 100B-parameter models at 5W
Daily AI Brief: the AI news that matters, in your inbox.