Compute optimisation strategies are becoming more advanced via event driven architectures and Function-as-a-Service. With the growth of Artificial Intelligence and increasingly availability of Inference-as-a-Service and GPU-as-a-Service providers, a similar compute optimisation path is being travelled by AI workloads. While such orchestration of pay-as-you-go compute is driving massive efficiency for consumers of cloud-based compute services, it is creating huge inefficiencies in utilisation levels of compute hardware and poor monetisation efficiency of available electricity within the data centre. It is also creating huge energy usage fluctuations within the data centre, leading to fluctuating demands upon the energy grid, and therefore grid instability. In this session, we describe a blueprint for Data Centre design that balances these conflicting priorities using workload and device orchestration, and energy usage optimisation through the integration of a flexible-load consumer. By balancing these demands we demonstrate optimum monetisation of available energy, and a methodology for creating a consistent and stable baseload on the grid, driving grid stability.