Copper vs. Code
Grid congestion already costs the Dutch economy billions a year and puts security of supply at risk, because electricity demand is rising faster than grids can be built. New cables take the better part of a decade. Software that sees, predicts and steers the grid can free up capacity in the wires we already have within one to two years, and some of it in months. This note explains how, how much, and where the investable value sits.
The age of electricity has hit a wall of copper
Heat pumps, electric trucks, industrial electrification and AI data centres are all arriving on the same grid at once. The IEA expects data-centre electricity use alone to roughly double to about 945 TWh by 2030, slightly more than Japan uses today.1 Worldwide, more than 2,500 GW of solar, wind, storage and large-load projects are stuck in grid connection queues.2
The Netherlands is the sharpest case in Europe. More than 14,000 businesses are waiting for a new or larger connection, and since 1 July 2026 households in congested areas join the same queue.3 A government review puts the cost to the Dutch economy at roughly €10–40 billion a year.4 Businesses that cannot connect cannot electrify, and a grid run at its limit has less headroom when demand spikes or a line fails.
The root cause is a clock mismatch. Everything that plugs into the grid is built in one to five years. The grid itself takes five to fifteen.
Source: IEA, Electricity 2026 [2].
The grid is mostly empty. AI can finally see where.
Grids are sized for the few worst hours of the year, and are under-used the rest of the time. Operators leave wide safety margins because they cannot see or predict what happens on their lines. Cheap sensors, machine-learning forecasts and software that steers batteries, chargers and factories now make it possible to run the same wires much closer to their true limit. Four data points show what that is worth.
Queued projects that grid-enhancing technologies could connect
The IEA estimates that dynamic line and transformer ratings, power-flow control and topology optimisation could connect this much stalled capacity, with lead times of one to three years. Adding flexible connection agreements lifts it to 1,200–1,600 GW.
Cost of a virtual power plant versus the alternatives
A Brattle study for Google found that aggregating home batteries, EVs and smart devices into a virtual power plant delivers the same peak capacity for 40–60% of the net cost of a gas peaker or grid-scale battery.
Value added to wind power by better forecasts
DeepMind predicted output from 700 MW of Google's wind farms 36 hours ahead. Committing power to the grid a day in advance raised the value of that wind energy by roughly a fifth.
What the market pays for energy software at scale
Octopus Energy spun out Kraken, its AI-powered utility platform, at this valuation in December 2025. It is contracted to serve over 70 million accounts and runs a large residential virtual power plant.
One congested substation, one winter weekday
Below is a stylised 44 MW Dutch substation. Its evening peak already sits at the safe operating limit, so it can accept almost no new business connections. Switch on the software levers to see how much room they free up, and when that room appears.
The +20–30% at the top of this note is what better ratings add to a single line. At a substation, shifting demand and adding a battery stack on top of that, so the room can grow well beyond it. Part of that extra room is non-firm: it holds only while forecasts are right and flexible users respond.
Illustrative model by AIXTerra, not a forecast. Load profile, 44 MW rating and lever sizes are stylised. Ranges are anchored on the IEA's dynamic transformer rating estimate of +5–15% [2], a forecast-driven reduction in the operating margin, and a flexible share of 4–12% of load. Capacity freed this way is partly non-firm: it depends on forecasts holding and on contracted flexibility responding.
Software does not replace the next cable. It buys the years until that cable arrives, and those years are where much of the economic damage sits.
A problem with a price tag, and rules that now pay to solve it
per MWh of blocked demand
Ecorys's average estimate of the net societal cost of congestion in the Netherlands, even after companies take their own measures.8 Customers feel that as lost growth, not as an abstract grid issue.
Flexibility gets priority
The ACM's prioritisation framework gives “congestion relievers”, such as batteries and flexible users, earlier access to scarce capacity. Non-firm connection contracts are becoming mainstream.3
to first revenue
Software and orchestration ventures sell into an acute, paid problem, with recurring revenue and far less capital per euro of growth than hardware or materials.
This is the complement to our Dry Lab thesis. Materials ventures compress a long discovery clock; AI-powered energy ventures start earning early by making existing assets work harder. A portfolio with both balances long-dated upside with earlier cash flow.
Forecasts are becoming a commodity. Control is not.
A good load or price forecast is increasingly available off the shelf. Value accrues to the companies that turn predictions into trusted action on physical assets, at scale.
Generic forecasts and dashboards
Useful but easy to copy, and increasingly bundled by large utilities and cloud providers. Hard to defend on its own.
Proprietary asset data and integrations
Every connected charger, heat pump or transformer adds data and switching capability a competitor does not have. Integrations with hardware and grid operators take years to build.
Orchestration with market access and operator trust
Platforms that grid operators rely on to dispatch flexibility, and that earn money in energy and balancing markets. Revenue grows with every megawatt under control, not with headcount.
The risks a careful investor should price in
Regulation sets the revenue
Flexibility and congestion markets are young. Tariff or market design changes can move the business case up or down quickly.
Slow utility sales cycles
Grid operators buy carefully, with long pilots and procurement. Ventures need the runway and references to get through them.
Services dressed as software
Many “AI for energy” firms grow linearly with consultants. We avoid businesses whose revenue scales with headcount.
Big incumbents move in
Utilities, OEMs and hyperscalers are building their own platforms. Startups need a data or integration edge they cannot easily replicate.
AI-native ventures that make the energy system work harder
AIXTerra backs European ventures that use AI to unlock capacity, flexibility and value in the energy system we already have, making it cheaper and more secure, and that scale with software economics. We invest early, with initial tickets of €0.5–3 million from pre-seed to Series A.
- Early in the AI-energy nexusOur partners backed Dexter Energy (AI forecasting) and Sympower (flexibility aggregation) at pre-seed and seed, years before the current AI wave.
- Operated, not only investedThe team includes energy operators and AI engineers who have built, scaled and exited ventures in this market.
- Industrial pullPartnerships with leading energy and industrial corporates give portfolio companies a route to pilots and customers.
- A clear red lineWe back platforms whose revenue grows with megawatts under control, not with consulting hours.
A separate note sets out a specific co-investment opportunity in this space.
- IEA, Energy and AI, April 2025.
- IEA, Electricity 2026, Grids chapter, 2026.
- VNO-NCW, Afspraken over snelle oplossingen netcongestie, Feb 2026; Ondernemen.nl on the 1 July 2026 rules.
- Rijksoverheid, IBO Schakelen naar de toekomst, May 2025 (citing BCG).
- Brattle Group for Google, Real Reliability: The Value of Virtual Power, 2023.
- Google DeepMind, Machine learning can boost the value of wind energy.
- CNBC, Dec 2025; Octopus Energy Group announcement.
- Ecorys for the Ministry of Climate and Green Growth, De netto maatschappelijke kostprijs van netcongestie, Aug 2025; Energie-Nederland summary.
Important information. This note is issued by AIXTerra Ventures for information and discussion with qualified investors only. It is thought leadership, not an offer or solicitation to buy or sell any security or fund interest, and not investment, legal or tax advice. Illustrative figures and models are simplified assumptions, not forecasts or indications of future returns. Early-stage investing carries a high risk of losing all capital invested and is illiquid. Past performance of the team is no guarantee of future results. Any investment is made solely on the basis of the relevant offering documents.