Dispatch intelligence for the batteries already on the grid.
Studio Helper — self-initiated concept
Halcyon Grid is a fictional company. No figure here is a real result.
The problem
Most installed storage runs on logic fixed at commissioning — a charge window, a price threshold, a schedule someone wrote once. Wholesale price moves every five minutes. The two are not the same problem.
of US utility-scale battery capacity lists price arbitrage among its uses — but only 41% runs it as the primary strategy.
EIA, Sept 2025 · public domain
MWh of wind and solar curtailed in CAISO in 2024, up 29% year on year. Solar was 93% of it. Clean power thrown away while storage sat idle.
EIA, May 2025 · public domain
of all hours in 2025 cleared at negative prices in France, Germany, the Netherlands and Spain. In Spain the count doubled in a single year.
Ember, European Electricity Review 2026 · CC BY 4.0
The gap is not hardware and it is not capital. It is that the asset does not know what the next four hours are worth.
Why now
US utility-scale battery capacity has grown at roughly 70% a year for three years, and another 54 GW is already in the pipeline. Globally, 108 GW was added in 2025 alone — 40% more than the year before.
An optimisation layer needs assets to sit on top of. Three years ago there were not enough to build a company on. There are now, and every one of them is commissioned with software that will never improve on its own.
EIA, Aug 2026 (public domain) · IEA, May 2026 (CC BY 4.0)
Dashed bar is announced pipeline, not installed capacity. Global storage could reach 1,200 GW by 2030 — but that figure is the IEA's Net Zero Emissions pathway, a normative scenario rather than a forecast, and is shown here only for scale.
The insight
Every megawatt-hour of storage installed in the last five years is still there, still connected, and still being dispatched by whatever logic was loaded at commissioning. Most of it follows a fixed schedule or a single price threshold.
That installed base is larger than anything a new entrant could finance, and it is available today. Changing what an asset does requires no steel, no permit and no grid connection — only permission and an API.
We do not build batteries. We change the decisions the existing ones make.
What we do
Halcyon Grid sits between price and hardware. It forecasts, it decides, it instructs, and it measures what actually happened — then feeds that back.
No hardware. Nothing is installed on site. We speak the protocol the controller already speaks.
No capex for the owner. The asset is theirs; the uplift is shared.
Days, not quarters. Integration is a credential exchange and a shadow-mode period.
One asset, one day
Modelled day on a real market shape. Not an observed trade.
Market
US utility-scale storage operational
EIA, end-2025
lists arbitrage among its uses
EIA, Sept 2025
platform fee per MW-year
Our pricing — illustrative
US platform revenue, addressable today
Recurring, before performance share
Europe, same arithmetic
The EU fleet passed 10 GW in 2025, more than double the 4 GW of 2023, with a pipeline above 40 GW. On the same attach rate and the same fee that is a further €40M of addressable platform revenue, growing faster than the US.
Ember, European Electricity Review 2026 · CC BY 4.0
Why the number grows on its own
The denominator is installed capacity, and it is compounding at roughly 70% a year in the US. We do not need to win share faster than the market to grow — though we intend to.
What we have deliberately excluded
Behind-the-meter and EV fleets, which are larger again and structurally similar. They are not in this figure. A market slide that counts everything is a market slide nobody believes.
Business model
Line one · contracted
Platform fee, billed monthly, three-year term. This is the only number we call ARR.
Line two · variable
Against a counterfactual baseline agreed with the owner before we touch a setpoint. Reported separately, never blended into ARR.
Where a megawatt-hour of uplift goes
Gross uplift captured
To the asset owner
Halcyon Grid
The objection, and the answer
“Performance revenue is low quality.” Correct, if it is all you have. So the platform fee is contracted and covers gross margin on its own; the share is upside. Every customer also carries a floor and a cap, which makes the variable line forecastable within a band rather than a hope.
Pricing and split are illustrative — this company does not exist.
Traction
Every figure on this slide is illustrative. Halcyon Grid is a concept company; these are modelled against published Series A benchmarks, not achieved results.
ARR, contracted platform fee only — performance share excluded.
Benchmarks for context: Series A median ARR €2.5M, NRR 101%, gross margin 77%, CAC payback 15 months. Sources: CRV, Benchmarkit 2025.
Moat
01 · Data compounds
We do not observe the market, we act in it. Each instruction produces a measured outcome against a known counterfactual — the one training signal a data vendor cannot buy.
02 · Registration is slow
Operating in a wholesale market means registration, compliance and settlement integration per market. It takes quarters. We have spent them.
03 · The baseline stays with us
Performance is measured against a history we hold. Switching does not just mean re-integrating — it means losing the record that proves what the asset is worth.
04 · Density beats scale
Forecast accuracy improves per market, not per customer. The tenth asset in a market makes the other nine better — a local network effect a global competitor cannot shortcut.
Why not the battery OEM
They sell hardware and optimise for warranty exposure. Cycling the asset harder is a cost to them and a revenue to the owner.
Why not the trading desk
Their model requires owning or tolling the asset. Owners who want to keep the asset — most of them — are not served.
Why not in-house
A single owner sees one portfolio in one market. The forecast they can build is bounded by their own fleet.
Go to market
Land
A single site, free, in shadow mode. We produce the instructions we would have sent and report what they would have earned. Two weeks, no risk, no integration to undo.
Sales cycle 6–10 weeks
Expand
Shadow-mode evidence converts to the portfolio, because the owner now has a number for what the rest of the fleet is leaving behind. This is where the 128% retention comes from.
Median 4× MW at 12 months
Channel
Asset managers and independent power producers carry us into fleets we would never reach one site at a time. They keep the relationship; we keep the dispatch.
Two signed, four in diligence
The motion is deliberately bottom-up. Storage owners will not sign a control agreement on a slide deck — but they will let you watch, and the watching produces the evidence that closes them.
Illustrative
Team
This deck is a concept, so no individuals are named or invented. What a real deck would carry here is the composition below and the reason this specific group is the one to back.
Grid operations
Someone who has been on the control-room side of a dispatch instruction and knows why an owner says no. Carries market registration and the relationships that shorten it.
Quantitative trading
Short-horizon forecasting and execution under uncertainty, from a desk where being wrong cost money. Owns the optimiser and the risk limits around it.
Production ML infrastructure
Models that run every five minutes against live assets, with rollback. This is a reliability problem before it is a machine-learning one.
Headcount illustrative
The ask
Series A · 24 months of runway to 2,000 MW under management
Lead sought · board seat offered
Existing investors participating pro rata
Use of funds
Engineering & data science · €7.6M
Market operations & compliance · €3.8M
Commercial & partnerships · €3.4M
Infrastructure & security · €2.1M
Working capital · €1.1M
By month 6
Two further markets registered and live.
By month 12
1,000 MW under management. €12M ARR.
By month 18
Channel contributing a third of net new MW.
By month 24
2,000 MW. Series B on metrics, not narrative.
Round size, allocation and milestones are illustrative.
The batteries are already built. We decide what they do next.
Studio Helper — self-initiated concept · Halcyon Grid is a fictional company