Prometheus for Grid Balancing: Sovereign Load and Generation Forecasting for Energy Operators

By Micky Irons, founder and CEO of Mickai

The forecasting problem grid operators cannot send to the cloud

Every energy operator runs on prediction. How much demand the network will carry at 17:00 on a cold Tuesday. How much wind and solar will actually arrive against the day-ahead schedule. How much reserve to hold, how much to dispatch, and what it costs when a balancing call goes wrong. These forecasts move money and they move stability. A forecast that drifts by a few percent across a control area becomes constraint payments, frequency excursions, and in the worst case load shedding.

The instinct over the last decade has been to hand this to large public-cloud machine learning. For a grid operator that instinct collides with the law. Transmission and distribution operators sit squarely inside the UK and EU critical national infrastructure perimeter. Under the NIS Regulations they carry operator-of-essential-services duties over the security and resilience of the systems that run the grid. SCADA telemetry, real-time settlement data, network topology and demand profiles are exactly the operational data a regulator expects you to keep inside controlled boundaries.

Public-cloud AI breaks that boundary in a way many procurement teams underestimate. The US CLOUD Act reaches data held by US-headquartered providers wherever in the world it physically sits. A grid operator that pushes balancing telemetry into a hyperscaler model has, on paper, exposed national-infrastructure data to foreign-jurisdiction compulsory access. That is not a theoretical concern for an operator of essential services. It is a finding waiting to happen.

So the operator is caught between two pressures. Forecasting quality demands modern machine learning. Regulation and national-security duty forbid sending the data that feeds it off premises. Most vendors ask you to choose one. We built so you do not have to.

What Prometheus does inside the wall

Prometheus is the forecasting Studio inside Mickai, the sovereign AI operating system. Mickai is AI that regulated businesses own and run inside their own walls, on-prem and air-gapped, with every action written to a tamper-evident, post-quantum-signed audit record we call the OAR. Prometheus brings that posture to energy forecasting.

It runs three core jobs against an operator's own historical and live data, all without that data leaving the building.

Demand forecasting. Short-term and day-ahead load prediction across the control area, conditioned on weather, calendar effects, settlement periods and local behaviour, so the desk plans against a forecast tuned to its own network rather than a generic curve.

Generation forecasting. Wind, solar and intermittent output predicted against scheduled positions, so the gap between what was promised and what will physically arrive is sized before it becomes a balancing problem.

Balancing and reserve. The net imbalance position and the reserve required to hold frequency and meet constraints, surfaced early enough to act on rather than react to.

Because Prometheus runs on the operator's own infrastructure, it trains and infers on the full-resolution operational record. There is no sampling down to a privacy-safe subset, no anonymisation tax, no contractual ringfence around what the model is allowed to see. The model sees the real network because the real network never leaves the room.

Why on-prem and air-gapped is the whole point

The forecasting accuracy a grid desk needs comes from proximity to the data. The compliance posture a national-infrastructure operator needs comes from never letting that data cross a jurisdictional boundary. On-prem and air-gapped is the only configuration where both are true at once.

Every Prometheus run, every model update, every forecast issued is written to the OAR. The audit record is tamper-evident and post-quantum-signed, so an operator can show a regulator not only what the forecast was but that the record of it has not been altered. For an operator-of-essential-services duty under the NIS Regulations, that provable chain of custody over the decision record is the difference between asserting control and evidencing it.

This is the same architecture that lets Mickai serve other regulated sectors under PRA SS2/21, UK GDPR special-category rules, the NHS DSP Toolkit, the EU AI Act high-risk regime and ITAR and EAR. The wedge is large because the constraint is structural: roughly 0.85 million UK businesses and around 5 million across the EU are effectively barred from sending their most sensitive workloads to public-cloud AI. The sovereign AI market sized at around USD 40 billion in 2025 is projected toward USD 148 billion by 2032. Energy forecasting is one of the cleanest expressions of that need, because here the regulated data and the national-security data are the same data.

A platform, not a point tool

Prometheus is one Studio among many in Mickai. The same on-prem substrate runs Nemesis for fraud and AML, Plutus for finance, Tyche for underwriting, Nomos for compliance, Astraea for legal, Panacea for clinical, Pythia for business intelligence and Aletheia for audit, alongside Trust Agent, the AMT agentic marketing team, Vinis voice and OAR-as-a-Service. A grid operator that brings Prometheus inside the wall is adopting an operating system that can also run its compliance, audit and business-intelligence workloads on the same sovereign footing, under the same audit record.

That platform is built and live, not a roadmap. It rests on a defensible position: 104 filed UK patent applications, roughly 2,340 claims, held by Mickai LTD with inventor Micky Irons. Filed not granted, which gives us priority and a prior-art moat over how sovereign, audited, on-prem AI is done. As a third-party momentum signal, Micky Irons ranked number 4 on Crunchbase as verified in June 2026, with the Mickai company in the top 1 to 2 percent globally. We are a UK company with Birmingham manufacturing secured, building to scale.

I want to be clear about posture. Mickai is an ally to the wider AI ecosystem, not a hyperscaler killer. The hyperscalers will keep serving the workloads that belong in public cloud. Our category is the workloads that legally cannot go there, and grid balancing is one of them. That is a dual-buyer thesis: the regulated operators who need to own their AI, and the larger players who would rather partner with the sovereign layer than try to retrofit one. It is a category a hyperscaler would want to own.

The window for selected partners

We are working now with a small number of energy operators who want sovereign forecasting inside their own walls and want to shape how it lands in their control rooms. We are running with chosen partners rather than open intake, because the operators who move first will set the reference architecture the rest of the sector follows.

If you run a transmission, distribution or generation operation and the gap between forecasting ambition and compliance reality is a problem you recognise, I would like to talk. Reach me directly at micky@mickai.co.uk.

FAQ

Does Prometheus send any grid data to the cloud? No. Prometheus runs on-prem and air-gapped on the operator's own infrastructure. Training and inference happen on the full-resolution operational record inside the building, and that data never crosses a jurisdictional boundary.

How does Prometheus help with NIS Regulations and CLOUD Act exposure? Because the data never leaves the operator's controlled environment, it is never exposed to foreign-jurisdiction compulsory access under the CLOUD Act. Every run and forecast is written to the tamper-evident, post-quantum-signed OAR, giving an operator-of-essential-services a provable chain of custody over its decision record.

What does Prometheus actually forecast? Three core jobs: demand and day-ahead load across the control area, wind and solar generation against scheduled positions, and net imbalance plus the reserve required to hold frequency and meet constraints.

Is Prometheus a standalone tool or part of a platform? It is one Studio inside Mickai, the sovereign AI operating system. The same substrate runs compliance, audit, finance, clinical and business-intelligence Studios under the same audit record, so an operator adopts a platform, not a point tool.

Is Mickai competing with the hyperscalers? No. Mickai is an ally to the wider AI ecosystem. Its category is the regulated and national-infrastructure workloads that legally cannot run in public cloud, which the hyperscalers do not serve.

Frequently asked questions

Does Prometheus send any grid data to the cloud?

No. Prometheus runs on-prem and air-gapped on the operator's own infrastructure. Training and inference happen on the full-resolution operational record inside the building, and that data never crosses a jurisdictional boundary.

How does Prometheus help with NIS Regulations and CLOUD Act exposure?

Because the data never leaves the operator's controlled environment, it is never exposed to foreign-jurisdiction compulsory access under the CLOUD Act. Every run and forecast is written to the tamper-evident, post-quantum-signed OAR, giving an operator-of-essential-services a provable chain of custody over its decision record.

What does Prometheus actually forecast?

Three core jobs: demand and day-ahead load across the control area, wind and solar generation against scheduled positions, and net imbalance plus the reserve required to hold frequency and meet constraints.

Is Prometheus a standalone tool or part of a platform?

It is one Studio inside Mickai, the sovereign AI operating system. The same substrate runs compliance, audit, finance, clinical and business-intelligence Studios under the same audit record, so an operator adopts a platform, not a point tool.

Is Mickai competing with the hyperscalers?

No. Mickai is an ally to the wider AI ecosystem. Its category is the regulated and national-infrastructure workloads that legally cannot run in public cloud, which the hyperscalers do not serve.