Untether AI

Untether AI is creating a new class of specialized processor targeting an extreme point on the energy-complexity scale. By eliminating as much complexity as possible while still being able to perform the operations necessary to do neural net inference and putting that processing as close as possible to the memory, we can get close to theoretical minimum power numbers for the given process (initially targeting 28nm). Neural net inference will become increasingly important on all types of devices. These devices operate with strict energy budgets which in some cases are too restrictive for any current inference technology and in other cases limit overall intelligence due to the maximum inference throughput. Untether AI is uniquely positioned to solve this problem with architecture that maximizes inference output for minimum power and is completely scalable.

Data last updated: June 2026