The Alpha chip technology was used in the design of multiple generations of Google's TPU.
The Alpha chip project developed a deep reinforcement learning agent capable of generating superhuman chip layouts.
Alpha chip technology has been used in the production of the last four generations of Google's TPU AI accelerator chips.
A one-day delay in the release of a chip like the NVIDIA Blackwell is estimated to cost the company approximately $225 million in lost opportunity cost.
In its second phase, Recursive Intelligence plans to become a platform that can take a workload, such as a new Claude model, and automatically design a custom chip architecture to accelerate it.
The third phase of Recursive Intelligence's strategy involves vertical integration, where the company would build its own chips and co-evolve them with its own AI models.
Recursive Intelligence aims to make its chip design tools 100,000 times faster than existing commercial tools.
Recursive Intelligence has developed a static timing analysis (STA) engine that is 1,000 times faster than commercial equivalents while maintaining high-fidelity correlation.
Recursive Intelligence aims to create a "designless" era for semiconductors, analogous to how TSMC enabled the "fabless" era.
Recursive Intelligence is a company focused on using AI for chip design and designing chips specifically for AI workloads.
The research on the Alpha chip was published in the journal Nature.
Alpha chip technology was used in the design of Google's Axion data center CPUs and chips for Pixel phones.