Came here to point this out as well. Previous flexible and plastic-based CPUs have also had a significantly low clock speed due to the scale of the components being massive versus traditional Si dies. That requires the circuit to operate at sub-MHz speeds since the device is massive (in area) and the signal paths are long. It seems like a novel device only suited for low performance applications that don't need a lot of resources so I'm not sure why they put ML features on it as a priority. Hardly seems like ML is needed here at all. And what about RAM? Big features don't lead to efficient or performant RAM devices at all. I just don't see the need for such a development when we currently have extremely low-cost and micro-sized microcontrollers available currently which blow the doors off this things performance abilities. Padauk chips are in the range of $0.01 per unit at 1000 units.