For most of the twentieth century, energy was a solved problem. You burned something, you got power, you sold it. The grid was a machine designed for this: large, predictable generators pushing electricity in one direction, to passive consumers who flipped switches.

That world is ending. In its place, something far more complex and far more interesting is emerging, and understanding why it's happening, and what it requires to manage it, is one of the more important questions of the next decade.

Renewables are winning on price. And breaking the grid.

The single most important energy fact of the last decade is this: solar and wind are now the cheapest sources of electricity ever recorded.

The cost of utility-scale solar has fallen over 90% since 2010. Onshore wind is down 70%. New solar installations today generate electricity at a levelised cost of roughly $38 to $78 per MWh, cheaper than operating existing coal plants in most markets, let alone building new ones.

This is genuinely historic. For the first time, the cheapest way to generate electricity produces no fuel cost, no emissions, and near-zero marginal cost once built. The economic case for renewables is no longer an environmental argument. It is arithmetic.

But here is the problem that comes with this win. Solar panels only generate when the sun shines. Wind turbines only spin when the wind blows. Neither cares about when you need electricity.

The grid was built on the assumption that generation follows demand. You could turn a gas turbine up or down within minutes to match what consumers needed. Renewables invert this relationship entirely: supply now drives the curve, and demand must adapt, or the grid destabilises. Think of the grid like a water system. It needs constant, stable pressure to function. Too little pressure and nothing flows: lights go out. Too much and the pipes burst: a blackout from the other direction. Either way, the result is the same. The grid must be kept in near-perfect balance between supply and demand at every moment of every day.

The result is a phenomenon engineers call the duck curve named for the shape it draws on a graph of net grid demand throughout a typical sunny day. Solar floods the grid during midday hours, driving prices down, sometimes below zero. Then the sun sets. Demand surges as people return home, turn on heating, and plug in their cars. The grid must ramp from a midday oversupply of cheap electricity to peak demand in just a few hours a steeper, faster climb every year as more solar is installed.

In Germany, wholesale electricity prices already go negative for hundreds of hours per year; the grid has too much supply and nowhere to put it. In the same country, within hours of those surplus events, prices spike sharply during calm, overcast periods when wind and solar produce almost nothing simultaneously. The same cables, the same consumers, but prices swinging from negative €50/MWh to positive €300/MWh in the space of a day.

This is not a Germany problem. Every country adding significant renewable capacity is seeing the same pattern. The more renewables on the grid, the more volatile the prices, the more curtailment (where perfectly good clean electricity has to be switched off because the grid has more supply than it can handle and no way to store or redirect the excess) and the more stress on infrastructure designed for a different era.

Everything is being electrified. At once.

If the supply side is becoming harder to manage, the demand side is simultaneously becoming harder to predict. Three distinct waves of electrification are arriving at the same time, each large enough to matter individually, collectively unprecedented.

The household transformation. Europe is replacing the two largest energy-consuming systems in most homes (heating and transport) with electric equivalents. Heat pumps are replacing gas boilers. EVs are replacing combustion engines. Home batteries are being installed alongside rooftop solar panels. The EU has effectively mandated the end of new combustion engine car sales by 2035. Heat pump targets point to 60 million units across Europe by 2030, up from roughly 20 million today.

AI and data centres. Every time someone uses an AI assistant, runs an AI-powered business process, or generates content with a model, electricity is consumed, in a facility running 24 hours a day, 365 days a year. The IEA's World Energy Outlook 2025 projects global data centre electricity consumption to approach 1,000 TWh by 2035 and exceed 1,300 TWh by 2040, with AI workloads accounting for a growing share of that growth. Goldman Sachs estimates global data centre power demand will reach 1,131 TWh by 2030, roughly tripling from 2023 levels. Microsoft, Google, and Amazon each announced tens of billions in data centre construction in 2024 alone. This demand is relentless, round-the-clock, and growing faster than almost any other category.

Industrial electrification. Decarbonising heavy industry requires replacing fossil fuel combustion with electricity. Green hydrogen production, electric arc furnaces for steel, electrified cement production all require large amounts of electricity that previously came from gas or coal. Add to this the geopolitical pressure to reshore manufacturing to Europe, reversing decades of offshoring that quietly exported industrial electricity demand to Asia. More factories means more electricity demand, at a time when the grid is already under stress.

What about efficiency? The obvious counterargument is that we are simultaneously getting more efficient. LEDs use a fraction of the power of old filament bulbs. Modern industrial equipment is dramatically more efficient. AI itself can optimise energy use in buildings, factories, and logistics. These gains are real and should not be dismissed.

Why building more grid won't solve this.

Step back and the picture is stark. Renewable growth means the duck curve deepens every year, bringing more midday surplus, steeper evening ramps, more volatility. Electrification means demand surges arrive faster and harder when people return home. Both forces hit a grid designed for neither, simultaneously and at scale.

The intuitive response is to build more grid. More transmission lines, more substations, more cables. If the infrastructure cannot handle the load, upgrade the infrastructure.

This is not wrong, but it is too slow and too expensive to be the primary answer.

A major high-voltage transmission line in Europe costs €1 to €3 million per kilometre and typically takes 10 to 15 years from planning to operation, including permitting, land rights, and construction. Germany's Südlink project, a 700 km underground cable designed to move wind power from north to south, was first proposed in 2012 and is not expected to be operational until 2028. The European Commission estimates that meeting European decarbonisation targets requires €584 billion in grid investment by 2030. With four years remaining, the Commission's own planning figures put required spending at €90 billion per year — meaning roughly €450 billion still needs to be deployed before 2030. To put that in scale: it is roughly equivalent to Austria's entire annual GDP, to be spent in four years on infrastructure that typically takes a decade to permit and build.

Current investment trajectories fall far short. The devices, meanwhile, are arriving now.

But there is a less obvious opportunity hiding inside this mismatch, and it points toward a fundamentally different kind of solution.

The grid is radically underutilised for most of the day. Peak demand in a typical European country reaches 70 to 80 GW. Average demand over the same year is around 45 to 50 GW. That 30+ GW gap is headroom sitting unused most hours of every day. The infrastructure is sized for the worst moments: cold winter evenings when everyone arrives home, turns on the heating, and plugs in the car simultaneously. The rest of the time, it runs well below capacity.

This is the core insight. The problem is not insufficient total energy. The problem is that demand and supply are not synchronised. If you could shift even a fraction of peak demand move EV charging from 7pm to midnight, run heat pumps harder during cheap midday solar hours. The same grid infrastructure could handle dramatically more total throughput. Less curtailed renewable energy, less grid stress, no new cables required.

Research consistently finds that demand flexibility defers grid investment at a significant cost advantage over traditional infrastructure, though the ratio varies by context. At the conservative end, the U.S. Department of Energy found that virtual power plants cost $43/kW-year compared to $99/kW-year for a gas peaker plant, a 2.3:1 advantage. At the higher end, Con Edison deferred a $1.2 billion substation upgrade in New York by investing roughly $200 million in demand reductions — a 6:1 ratio. Across studies, the range runs from roughly 2:1 to 6:1 depending on location, asset mix, and what is being deferred. Flexibility is not just an energy market instrument. It is the faster, cheaper substitute for infrastructure we do not have the time or the money to build at the required pace.

Flexibility sitting dormant in millions of homes.

Before going further, it helps to calibrate scale. One gigawatt (1 GW) is roughly the output of a large nuclear power station, enough to power around 700,000 European homes. When we talk about tens of gigawatts of untapped flexibility, we are talking about the equivalent of dozens of power stations' worth of controllable capacity, sitting in living rooms and driveways across Europe, never asked to do anything except consume.

ENTSO-E estimates there is roughly 100 to 200 GW of controllable flexible demand already present in European homes and businesses. Currently, less than 20 GW of it is actively managed. The gap between what exists and what is used is where the opportunity lies.

What is this flexibility, concretely? It is the fact that a heat pump does not need to run at exactly 6pm; it can run earlier and achieve the same indoor temperature with no occupant aware of the difference. It is the fact that an EV does not need to charge the moment it is plugged in; it needs to be charged by 7am, and when it charges between 10pm and 6am is economically irrelevant to the driver. It is the fact that a home battery can absorb surplus midday solar and release it during the evening peak, rather than letting cheap electricity go to waste.

The economics per device are meaningful. Multiply the value generated per heat pump and battery across the 60 million heat pumps and 130 million EVs projected across Europe by 2030, and the aggregate market is not a niche instrument, it is a fundamental piece of energy infrastructure.

The critical bottleneck is not the devices they exist, they are being purchased, they are sitting in homes right now. The bottleneck is the software layer that connects them. A heat pump connected only to its manufacturer's app does not participate in flexibility markets. A utility that wants to aggregate 10,000 devices needs device-level communication, scheduling logic, market interfaces, compliance reporting, and real-time monitoring, across dozens of device types, across multiple markets, with different protocols and regulatory requirements in each country.

This is genuinely difficult software. It requires understanding both physical device behaviour, including the thermal dynamics of a heat pump and the charge curve of a battery, and energy market mechanics. Grid operators have market expertise but cannot build device software. Device manufacturers understand hardware but have no interest in becoming energy trading companies. Utilities have the customer relationships but not the engineering capability. The layer needs to be built independently.

The problem gets bigger. The role gets more critical.

Everything described so far reflects the situation today. The trajectory makes all of it more acute, not less.

Renewable penetration will continue to increase. The EU's target is 42.5% of total energy from renewables by 2030, up from roughly 24% today. More renewables means the duck curve deepens further, bringing more midday surplus, steeper evening ramps, and more curtailed clean energy. The grid stress that characterises high-renewable markets today becomes the default condition for every European grid within a decade. The flexibility requirement grows proportionally, and it compounds: each new solar panel installed increases the value of every controllable device already connected to the grid.

The home device transformation is only beginning. A household that adds an EV and a heat pump doubles its electricity demand. Add a home battery and a solar inverter (both increasingly common as prices fall) and that household becomes a genuinely complex energy node: generating, storing, consuming, and potentially selling back to the grid, all from the same address. Multiply that complexity by 100 million European households over the next decade and the management challenge is not incremental. It requires a different kind of infrastructure entirely.

AI's energy demand is still in its early stages. The inference workload, running AI across every business process, every device, and every interaction, has barely begun to land on the grid. Data centre construction announced in 2024 translates into operational power demand from 2026 onward, in quantities that will stress regional grids in ways we are only beginning to model. And the Jevons principle applies with particular force: AI makes energy-intensive processes cheaper and more accessible, reliably expanding total consumption rather than reducing it.

The countries, regions, and companies that solve this, that figure out how to have abundant, flexible, cost-competitive energy will be where the next wave of development happens. Those that don't will find energy acting as a ceiling on what they can build and how fast they can grow.

The solution to this is not purely physical. We cannot build our way out of intermittency with transmission cables alone, nor keep pace with demand growth through generation alone. What the energy system needs is a software layer that makes the existing grid intelligent that turns millions of distributed, unpredictable assets into a coherent, responsive network that absorbs surplus, shaves peaks, and moves value to where it is needed.

The network effects are real and compounding. Every device added to a managed fleet increases the total flexibility available to energy markets. More flexibility means access to faster and higher-value markets. More market access means more revenue per device for utilities, which makes enrolling more devices more attractive. The platform that achieves meaningful scale builds a position that is difficult to replicate not because the software cannot be copied, but because the network of traded devices cannot.

The grid stress is real. The demand surge is real. The infrastructure gap is real. And the solution hundreds of millions of devices already installed, already connected, waiting to be asked to do something useful is right there.

Making them useful is the work.