The global buildout of AI infrastructure is creating an energy problem. Training runs for frontier models consume tens of megawatts for weeks at a time. Inference workloads are growing faster than anyone predicted. And every new AI datacenter added to the grid is, by default, a large, inflexible load that draws power without contributing anything back to the energy system that supplies it.
This trajectory is unsustainable — not because the world lacks the generating capacity, but because modern power grids were not designed for the concentration of demand that AI infrastructure introduces. The answer is not simply more generation. It is smarter infrastructure that integrates with the grid rather than straining it.
The Grid Stability Problem
Power grids operate on a fundamental principle: supply must equal demand at every instant. When a large, constant load appears on a distribution network — as a datacenter does — the grid operator must ensure that sufficient generation is available and that frequency remains within tight tolerances. The higher the load, the more vulnerable the system becomes to frequency deviations caused by sudden changes in supply or demand.
Battery energy storage systems directly address this challenge. A BESS participates in frequency containment reserve (FCR) and manual frequency restoration reserve (mFRR) markets, providing rapid-response balancing services that keep the grid within its operating parameters. The response time of a modern lithium iron phosphate BESS — less than 200 milliseconds — is orders of magnitude faster than any thermal generation asset. This makes battery storage the most effective tool available for real-time grid stabilisation.
Co-Location Changes the Economics
The conventional approach to AI infrastructure treats energy as a cost line: procure power, consume it, pay the bill. Battery storage co-located with compute changes this equation fundamentally. The BESS earns revenue by stabilising the grid — participating in reserve markets, arbitraging price spreads, and providing ancillary services. This revenue stream offsets the energy costs of the compute operation and, in favourable market conditions, generates positive returns on the storage asset itself.
Co-location also creates operational synergies. Compute workloads can be made interruptible — shaped around renewable generation curves and grid signals. When the grid is stressed, compute curtails and the battery injects; when renewable supply is abundant and prices are low, compute runs at full capacity while the battery charges. The result is an infrastructure site that behaves as a flexible, responsive participant in the energy system rather than a rigid, passive consumer.
Revenue Diversification
A compute-only facility has a single revenue stream: the fees charged for compute services. A co-located compute-plus-BESS facility has at least three: compute service fees, frequency reserve market revenues, and energy arbitrage. This diversification materially improves the risk profile of each site. When compute demand softens, the BESS continues to earn. When reserve market prices spike — as they do during grid stress events — the storage asset captures that value directly.
For the Riveon Group, this multi-revenue architecture is not an optimisation layered on after the fact. It is the foundational design principle. Every site in the group’s portfolio is engineered for co-located compute and storage from the outset, with the NEXXUS EnergySuite orchestrating dispatch across all asset classes in real time.
Infrastructure That Gives Back
The AI economy will be built on physical infrastructure. The question is whether that infrastructure will be a burden on the energy systems it depends on, or a contributor to their stability and resilience. BESS co-location is not a niche strategy — it is the only credible path to scaling AI infrastructure without destabilising the grids that power it.
Battery systems earn revenue by stabilising the grid. They absorb surplus renewable generation that would otherwise be curtailed. They provide the rapid-response reserves that keep frequency within tolerance. And they transform datacenter sites from pure energy consumers into active, grid-positive participants in the energy transition.
The AI economy needs infrastructure that gives back to the grid, not just takes from it. BESS is how that happens.