Cortex-M and Cortex-A are devices with minimal capacities, minimal RAM and storage. CPU inference workloads must not only fit into a small space but frameworks must also be purpose optimized for the environment. µTVM and Tensorflow Lite Micro are both rapidly being developed to bring AI workloads on Microcontrollers. Collaborative activities across Linaro members are transforming these projects from early prototypes to productive environments for creating solutions.