Vital

Intelligence is becoming cheap and power at the right hour is becoming expensive. Solar has made daytime electricity abundant. In 2025 Germany had more than 570 hours of negative power prices and Spain more than 550. On sunny days, noon power trades near zero, sometimes below. Meanwhile, demand for power that is there at the peak keeps growing. Data centers run all night. Heat pumps pull hardest on cold evenings. EVs charge after work. Factories are coming home. The grid can't keep up. US connection queues held 2,061 GW of plants and batteries at the end of 2025, against a whole existing fleet of 1,374 GW. Britain's old queue passed 700 GW. A large transformer can take up to four years to arrive. So value is moving from the MWh itself to the MWh at the right hour and in the right place. Batteries, flexible demand, long cables and firm power plants (gas, nuclear, coal, hydro with a reservoir, geothermal and biomass) all compete for that gap. Who earns it, who pays for it, and who carries the price swings is the most interesting money question in energy. Most commentary stops at more renewables or more demand. The hour-by-hour price, which decides who actually gets paid, gets much less attention. That price is what I'll write about here: how the US and European grids work, why their prices move the way they do, and how the plants and wires get financed.

Before lending to a solar farm, banks hire an engineer to set a bad-year number: the output the farm should beat in nine years out of ten. In one Solargis example for southern Spain, the engineer works from 30 years of sunshine records, and the bad year comes in about 7% below a normal one. Banks size the loan so the farm can still repay in that bad year. Battery storage does not have such record. A battery earns by buying power cheap and selling at the evening peak and nobody can forecast those price gaps for 2035. So banks lend less or ask for a contract that fixes the income. That’s why Italy’s 15-year storage contracts drew bids two thirds below the cap. But Terna bought only 10 GWh in that round and turned away more than three quarters of what developers offered. We need a new way to finance battery storage without a contract like that. Until then, money is the bottleneck.

Italy offered up to $41,800 a year for battery storage. Developers bid $14,600. In September 2025 Terna, Italy’s grid operator, asked for 10 GWh of battery storage in the south and on the islands. Terna offered to rent that capacity for 15 years, at up to $41,800 per MWh a year. Bids came in at more than four times the capacity on offer, and the winners averaged about $14,600. The winners could bid that low because 15 years of fixed payments from a grid operator let them borrow cheaply. The cheapest borrowers won most: Enel and Eni, both partly state-owned, took about 70% of the capacity. Cells are cheap now. Money to build with them often isn’t, and that’s the bottleneck for battery storage. Give developers 15 years of fixed income and they bid two thirds below the cap. The cost of capital, more than the price of cells, decides how much storage Europe builds.

Cheap panels, expensive money: why solar in Brazil costs 76% more Solar power in Brazil cost $79 per MWh in 2021, on IEA estimates. With Europe's cost of capital, it would have cost $45. A solar farm is paid for almost entirely up front. So the interest rate on that money weighs on every kWh it sells. Rates differ mostly because of what lenders charge to lend in a country, which tracks its credit rating. The gap grows when loans are in dollars or euros and revenue is not. In Brazil and most of Africa, the cheapest kWh now comes from cheaper money: contracts that hold, revenue in the same currency as the debt, policies that don't reverse. That work gets a fraction of the attention panel prices do.

The grid is the new bottleneck US grid queues hold 2,061 GW of plants and batteries waiting to connect. The entire US fleet today is 1,374 GW. Most of it will never be built. Of everything that applied between 2000 and 2020, 13% was running by the end of 2025 and 75% had dropped out. Projects that did make it waited about five years.

Every new solar panel lowers what solar earns Yesterday's chart was one day in Spain: free at noon, $260 at 7:45 p.m. Repeat that for years and solar ends up cannibalizing its own revenue. In 2019 a MWh of Spanish solar sold for slightly more than the average price for power from any source. In 2026 so far, it sells for half. It's not just Spain. German solar earned about half the average price in 2025. In California it's under a third. Every solar farm produces in the same hours. Once there is enough of it at noon, gas drops out and something cheaper sets the price: other solar, hydro, or an offer below zero. So solar sells most of its output into hours its own output made cheap. Therefore, solar targets in MWh measure the wrong thing. They count how much energy solar makes, but the market pays for when it makes it.

Free at noon, $260 at 7:45 p.m. ;) Tomorrow’s power prices in Spain set in today’s auction: about $0 from noon to 5 p.m., about $260/MWh at 7:45 p.m. Same grid, same day. The only thing that changed is the sun. Every solar farm sells into the zero, which is why Spanish solar now earns about half the average price. Whoever can store power buys it at noon and sells it after sunset: pumped hydro, batteries, and the traders who run them.

The price of power is decided in the capital stack as much as in the factory Baltic Power, a 1,140 MW wind farm off Poland, borrowed 80% of its cost from banks. The owners put in 20%. Banks get paid first, so they charge the least. Owners get paid last, take the first hit if revenue falls short, and want a much higher return for it. So the bigger the share banks will fund, the cheaper the money and the cheaper the power. Banks lend 80% only when revenue is locked in, here by a 25-year contract that outlasts the loan. Want cheaper electricity? Make more of it lendable.

How solar became bankable and how Spain broke it In 2000 Germany promised solar owners the equivalent of 47 US cents for every kWh, for 20 years. A bank could read a meter, multiply by a price fixed in law, and see 20 years of cash. That's what made solar lendable. It passed the three tests a lender cares about. A meter the owner doesn't control measures the output. Decades of weather data predict it. And the law forced the grid operator to buy every kWh at that fixed price, so there was always a buyer. Spain copied the model in 2007. Banks lent for 20 years at about 1 point over Euribor, the euro interbank rate. Then Spain cut the promise. In 2010 it capped paid output at 1,250 hours a year, even for plants already running. In 2013 it scrapped the tariffs. The panels and meters were fine. The payer changed the rules. Where promises held, debt kept getting cheaper. Solar loan margins fell 27% in Europe and 32% in North America from the early 2010s to 2020. I would call Spain's cut one of the most expensive savings a European government has made. It saved on tariffs once and taught every lender to charge for the chance that a state rewrites its promises.

A wind farm can *pay* $60 to sell its power and still make money When the price is negative $60/MWh, every seller pays $60 for each MWh it delivers. Most plants switch off. Some keep running. They are paid per MWh by the state. Say a farm gets $70 for every MWh it produces. At negative $60 it still clears $10, so it keeps producing into a grid that already has too much power. In Germany that money comes from the federal budget. In the US it is a federal tax credit. That's why negative prices run deeper and longer than they need to. The plants that could end them by switching off are paid to keep going.

How a day-ahead electricity auction clears At noon Central European Time order books for the next day close across most of Europe. One algorithm Euphemia then sets a price for every bidding zone for every 15 minutes of the next day at once and allocates cross-border line capacity in the same run. Final results come out at 12:57 p.m. Each zone gets one price per 15 minutes paid to every seller and charged to every buyer. Prices can go below zero, down to a hard floor. The rules lower that floor automatically when prices get close to it. PJM, the grid operator for 13 states from New Jersey to Illinois, closes its day-ahead market at 11 a.m. Eastern. Generators offer three numbers: what it costs to start, what it costs per hour just to stay on at minimum output, and a price for each extra MWh. PJM picks the cheapest set of plants to run tomorrow that doesn't overload any power line. Where a line is full, cheap power can't reach the far side, so prices differ by location: thousands of separate prices, published by 1:30 p.m. So, a European plant's day-ahead income for tomorrow is fixed by early afternoon today. Coming from DeFi, I read this as a batch auction: everyone clears at one price, so every seller's best bid is its real cost.

Why noon power got cheap California and Spain have the same problem at noon. Too much solar, not enough demand. Every panel peaks at once. When output passes what the grid can use or export, the price falls to zero and then below. Spain had more than 550 hours of negative day-ahead prices in 2025. Most were shallow, averaging about minus $2/MWh. California's grid operator curtailed about 4 TWh of wind and solar in 2025, up from 3.4 TWh in 2024. Solar was 93% of the 2024 total, mostly in spring, when sun is strong and demand is mild. Why would anyone sell below zero? Some plants get paid regardless of the market price. A subsidy per MWh, a tax credit or a fixed-price contract can make running at a negative price better than stopping. Then the sun sets. Demand stays high into the evening, solar output drops to nothing, and the price climbs, usually set by gas. The midday dip deepens as more solar arrives, because each new panel produces in the same hours as the last one. That dip is the battery's buying window. California had about 15.8 GW of batteries installed at the start of 2026. In 2025 they routinely supplied almost a fifth of demand at the evening peak. Each new solar farm sells into a cheaper noon than the last, so its owners and lenders can count on less income per MWh. Batteries are paid by the same gap: the deeper the noon dip and the higher the evening price, the more each charge and discharge earns. What bothers me is a subsidy paid per MWh whatever the market price. Say a solar farm gets a fixed $60 for every MWh. At a market price of minus $2 it still clears $58, so it keeps running in exactly the hours the grid has too much power. And the subsidy costs the most then: it has to cover the full $60 plus the $2 below zero. My view: no subsidy should be paid for power produced when the price is negative. The plant would switch off or store the power instead. Germany already does this for new solar since 2025.

Capacity factor and why it is the wrong comparison Capacity factor is what a plant produces divided by what it would make running flat out all year. In the US in 2024, the US Energy Information Administration put nuclear at 90.8%, gas combined-cycle at 60.5%, wind at 34.3% and solar at 23.2%. So 1 GW of nuclear makes about 8 TWh a year and 1 GW of solar about 2 TWh. It is only half the comparison. The other half is timing. Solar's 2 TWh all arrive in daylight, in the same hours as every other panel. When every panel produces at once, the midday price collapses. In May 2025, a MWh delivered evenly around the clock in Germany fetched about $76 on average. Solar, selling mostly into those cheap midday hours, averaged about $22.50, according to S&P Global. A MWh is worth what its hour pays. Compare sources by what they earn and what they replace, hour by hour. For lenders and buyers of a solar farm, the figure that counts is what its output earns in its own hours, not yearly output times the average price. Valued at that around-the-clock average in May 2025, solar's income would have looked more than three times larger than it was. Capacity factor is the most overused number in energy arguments, and I see it used badly in both directions. On its own it measures energy, not value. The number I want next to it is the capture price: what a plant earned per MWh in the hours it actually ran.

Who sets the price Every hour, the market lines up offers from cheapest to most expensive and runs down the list until supply meets demand. This is the merit order. The last plant needed sets the price for everyone. A solar farm with almost no running cost is paid the same as the gas plant that closed the gap. That's why gas matters so much in Europe even where it supplies a small share of the energy. Across the EU, fossil fuels set the price in roughly half of all hours in 2025, down from about 70% in 2020, according to the European Commission. The spread between countries is wide. In 2026 so far, gas set the price in close to 90% of hours in Italy, about 40% in Germany and 15% in Spain. US markets work the same way, with one difference. The organized markets price each point on the grid separately, so the last plant needed at a given node sets the local price. When lines fill up, prices at different nodes pull apart. Renewables push gas out of the stack in sunny and windy hours. Gas comes back to set the price when they fade. Batteries are now taking some of those hours, charging at noon and selling into the evening peak. In California in 2025, batteries routinely supplied almost a fifth of evening peak demand. So in Italy, where gas sets the price in close to 90% of hours, the wholesale power bill moves with the gas price no matter how much energy comes from solar or wind. Each hour that renewables or batteries push gas out of the stack is an hour every buyer pays a lower clearing price.

Electricity consumption correlates with the ability to finance power projects. More capital for power projects -> more power generation -> more electricity consumption -> greater GDP per capita.

Measure, predict, enforce: how a cash flow becomes bankable In 2016 a 100 MW solar farm in Ontario sold investment-grade bonds that pay off by 2035. Lenders accepted it because its income passed three tests: someone other than the owner measures it, history predicts it, and a contract makes it payable. Measure A lender needs a number the owner can't change: revenue meters read by the grid operator, market settlement statements, an independent engineer's report. Data an owner reports about its own asset is weaker evidence, even when it's accurate, because the asset is grading itself. Predict Debt runs 10 to 20 years, so the lender needs history covering that span. Solar has it: decades of weather records, which engineers turn into a P90, the output a plant should beat in nine years out of ten. The Ontario plant, Kingston Solar, was rated on that case and produced 105% of it in its first 11 months. Enforce Kingston sells under a 20-year fixed-price contract with Ontario's grid operator, and the bonds are repaid inside that term. Germany's feed-in tariffs from 2000 did the same for a whole country.Batteries pass half the testsWhether a battery was available is measured in fine detail by grid operators, backed by warranties and paid for by capacity and frequency contracts. Trading revenue is measurable but not predictable. Modo Energy puts the average Texas battery at $193 per kW of capacity a year in 2023, $56 in 2024 and about $29 projected for 2025. No lender can build a P90 on a sixfold drop in three years. So what The three tests decide how much debt a project can carry, and debt is the cheapest money in a power project. Income that passes all three gets long, cheap loans, so the plant can sell its power cheaply. Income that fails one gets less debt, costlier equity or no financing at all. That is why the industry is working to make battery income pass: tolls that pay a fixed fee, insurer-backed revenue floors, and long contracts such as Italy's 15-year storage auction. How fast storage gets built depends as much on that work as on the price of batteries.

US grid batteries, per the US Energy Information Administration. End of 2020: 1.65 GW. End of 2025: 43.6 GW. Mid-2026: nearly 52 GW, with 14 GW more planned before year end. About 26 times in five years.

The transformer and cable shortage A large power transformer is built to order around a core of grain-oriented electrical steel. Few factories make them. In North America a single company makes that steel for distribution transformers. Wood Mackenzie put lead times for large power transformers at 128 weeks in Q2 2025, and 144 weeks for the step-up units that connect power plants to the grid. The International Energy Agency says large transformers can take up to four years and high-voltage cables two to three, with both lead times almost doubled since 2021. The US imports about 80% of its power transformers, and Wood Mackenzie expected a 30% supply shortfall in 2025. Delays make existing grid connections more valuable. A site that already has its transformer and an approved grid connection can be running years sooner than one starting from scratch.

How a battery gets paid A grid battery earns money in three ways. The first is trading. It charges in cheap hours and discharges in expensive ones. Some energy is lost on the way. US utility-scale batteries returned an average of 82% of what they stored in 2019, according to the US Energy Information Administration. The spread has to beat those losses before the battery earns anything. The second is balancing. Grid operators pay batteries to stand ready to push power in or pull it out within seconds when frequency drifts from 60 Hz in the US or 50 Hz in Europe. They are paid for being available, and again for the energy they deliver. The third is capacity payments, where they exist. In the US, PJM (13 states from New Jersey to Illinois), New York and New England pay for firm capacity, as do Britain, Poland, Italy and Belgium in Europe. Texas has none. Germany has none yet; the European Commission approved a German scheme in September 2026 that starts in 2031. Balancing markets are small, and they fill up fast. Once balancing is full, trading becomes the main job. Then a battery's income depends on how far apart the cheap and expensive hours sit, day after day.