Somewhere today, an investment committee is pricing a grid-scale battery. In front of them is a revenue forecast. If that forecast treats the battery as a price-taker — dispatching against prices that ignore what every other battery in the fleet will do — it will overstate revenues by 100–300% at the fleet Great Britain plans to have on the system by 2030.

That is not a rounding error. It is the difference between a project that services its debt and one that doesn't. And versions of it are being made quietly, at scale, across the industry because the standard tools solve for prices first and storage second, when the two are one problem.

I founded Compounding Energy to close that gap, and the others like it.

The transition is a capital-allocation problem

I've spent my career building models that real decisions rest on. I'm an applied mathematician by training. After a career in academia and government service, I founded Vibrant Clean Energy (VCE) in 2016, where I created the WIS:dom capacity-expansion model — used for many studies across the industry — and co-authored some of the more widely cited work on the economics of high-renewable power systems. I exited by selling VCE to Pattern Energy, worked there for several years, then relocated to the UK. Over fifteen years of building and running national-scale grid models, I kept meeting the same trade, everywhere: rigour exchanged for speed. Zonal approximations where the grid is nodal. Simplified dispatch where the physics bind. Black-box methodologies you were asked to trust rather than interrogate.

That trade was tolerable when renewables were marginal. It isn't now. Global investment in the energy transition runs at roughly $2 trillion a year, and the transition is, at bottom, a capital-allocation problem: that money has to land in the right assets, in the right places, at the right time. Models that miss the system's real economics don't just cost their users money. They misallocate the transition itself — and every financing built on returns that were never available makes the next one harder to raise. Decarbonisation deserves numbers you can bank.

There's a second reason, and I'll be direct about it. I want to demonstrate that doing this properly is the winning commercial strategy: that the rigorous tool beats the convenient one in the market, not just in the journals. If Compounding Energy makes that case by existing, it will have done part of its job before a single forecast is read.

Why "Compounding"

Small errors compound. Every decision in energy compounds. A price forecast that ignores cannibalisation compounds into a revenue model that misses the equilibrium, which compounds into a financing built on a spread that the fleet itself will erode. By the time the error surfaces, it has a balance sheet attached.

But careful engineering compounds too. Weather physics feeds generation forecasts; generation feeds price formation; prices feed dispatch and revenue; and a fast native solver underneath makes the whole loop cheap enough to close properly, every day. Each product we build makes the next one better. The name is a promise about both directions.

What's live today

CECadence answers the question that opened this article. It solves fleet build-out, prices and battery dispatch jointly, to a fixed point — the equilibrium where the fleet's own behaviour is already priced in — across half-hourly wholesale and the ancillary stack, by duration cohort. Its annual mean revenue has landed within 5% of outturn in three of the five graded windows from 2022 to mid-2026, including the pre-registered held-out 2025 at 1.6%, and every run is reproducible from scenario, version and timestamp. It also reports the number most tools hide: the gap between the naive forecast and the equilibrium one.

CEGridSight is physics-first price forecasting to 14 days: a pan-European unit-commitment model running weather through demand, wind and solar, dispatch, carbon and into price, half-hourly across 43 zones — 17 in Great Britain, 26 EU bidding zones — with a full solve four times a day and a 30-minute nowcast between. Its skill is graded continuously against the settled market index, and we publish the grading.

CompoundVision forecasts wind and solar generation from first principles for more than 26,000 farms across Great Britain, the EU and the US — around 988 GW — hourly out to 16 days, with probabilistic bands that widen honestly when an asset's specifications are inferred rather than known.

CEAtlas is the lens that ties it together: a global map of generation assets, interconnectors and load, live locational prices across the US ISOs and EU day-ahead markets, site suitability across seven technologies, real-time lifecycle carbon, and a REST API. Increasingly it's the single interface for the entire Compounding Energy product suite: CEGridSight forecasts, CompoundVision generation and early CENovaSage results all render there today. The free tier is live now.

The engine room

Forecasting at this resolution is an optimisation problem before it is anything else, so we built the optimiser. CEMeridian is our native solver, written in Rust, engineered for continental-scale power-system problems; and it is in production. Every CECadence solve already runs on it, and it is in trial beneath CENovaSage. On DC-OPF benchmarks it runs 1.9–4.4× faster than HiGHS — the strongest open-source solver — single-threaded, and GPU-parallel contingency screening runs 31×. It holds its own against leading commercial solvers on the same instances. Speed here isn't vanity. It's what makes equilibrium solves and full probabilistic sweeps cheap enough to be routine rather than occasional.

Held to the standard

The next major release is CENovaSage: nodal capacity expansion. Where to build, at the resolution the grid actually operates. Built on adaptive nested Benders decomposition. It is working. It is not released. It is in full validation, because it is precisely the tool whose outputs a lender's technical adviser will probe hardest, and it ships when it passes that scrutiny — not before. Production-cost, resource-adequacy and techno-economic modules follow through the same gate.

That gate is the company in one sentence: methodology documented and published, skill graded against outturn, every claim reproducible, and nothing released until it meets the standard. If a number of ours is wrong, we want to be the first to know and the first to say so.

Who this is for

Battery developers and fleet operators. Investors and lenders underwriting energy assets. Utilities and system operators. Anyone whose decisions turn on the power system's real economics rather than a convenient approximation of them.

The platform is live at compoundingenergy.com — CEAtlas has a free tier you can use today. If your decisions rest on these numbers, I'd like to hear what you need.

The grid is being rebuilt once. Let's get the numbers right.