Why Solar Generation Keeps Growing Exponentially

ELI5 / TLDR

Every year for the past decade, solar has added about 25-30% more electricity to the world than the year before. Everyone keeps expecting it to slow down — that’s what new technologies do, they grow fast then level off — and every year it stubbornly refuses. In 2025 it grew 30%, the fastest rate in eight years, and accounted for three-quarters of all the new electricity the world consumed. The knock-on effects are starting to show: fossil generation actually fell globally, including in China and India, and cheap batteries are quietly removing the one thing that used to cap how much solar a country could absorb.

The Full Story

The host, Shayle Kann of Energy Impact Partners, opens by calling himself an “electricity bull” and saying he’s not sure the market is bullish enough. His guest is Nick Fulgham, a data analyst at Ember, the think tank behind the annual Global Electricity Review. The whole episode is a tour of who’s growing, who isn’t, and why the surprises are bigger than the headlines suggest.

Capacity is a promise, generation is the receipt

First, a distinction that the whole episode rests on. Capacity is how much a power plant could produce if it ran flat out — the nameplate number. Generation is how many electrons it actually delivered. They’re different because the sun sets and the wind drops. Critics of solar love to point at capacity figures and say “but that’s not real power.” So Ember deliberately tracks generation — the receipt, not the promise.

On that stricter measure, 2025 was a standout. Solar generation rose by 636 terawatt-hours. A terawatt-hour is a billion kilowatt-hours; to anchor it, 636 of them is roughly twice the entire annual electricity demand of the UK, added in a single year. That was a 30% jump over the year before — the highest growth rate in eight years.

Why a 30% growth rate is the real shock

Here’s the counterintuitive bit. When something is tiny, growing 30% is easy. When it’s already big, you expect the percentage to shrink even as the absolute numbers climb. Solar is now big — a major player in most of the world’s largest grids — and yet the percentage isn’t shrinking. It went up.

Most technologies follow an S-curve: slow start, explosive middle, then a flattening as the market saturates. Picture the letter S lying against a wall. Solar has been on the steep middle part for 17 years and shows no sign of bending over.

“I have a gut feeling every year that this is the last year where we’ll see this exponential growth continue… and then 2025 again we saw the same growth rate maintained and even increased.”

Ember has a nice yardstick for “how fast does an energy source go from a bit player to a giant” — the time to climb from 100 terawatt-hours (the point where a source stops being a curiosity) to around 2,000. No source in history has made that jump faster than solar: roughly 10-12 years, and it’s already past 2,700. Wind is just behind. Nuclear made a similar sprint in the 1970s and 80s — and then flattened almost completely and barely grew on a net basis afterward. The open question is when solar’s S finally bends, because an earlier or later turn changes everything about when the world reaches 30%, 40%, 50% solar. Right now it’s only at 8-9% globally.

The narrative violation: fossil fuels went down

Because solar (plus a solid 200 TWh from wind) covered nearly all of 2025’s extra demand, something rare happened: global fossil generation fell — only the fifth time this century, and the first since 2020. The deeper surprise is that coal fell in both China and India, the two countries everyone associates with building new coal plants.

China is the headline. Its fossil generation has now been flat for a couple of years — not a one-off. The trick is the gap between capacity and generation again: China keeps building coal plants (a hangover from blackout scares in 2020-21, plus the fact that China can throw up a coal plant in three or four years where others take much longer), but those plants run less and less. Coal’s capacity factor — the share of the time it actually runs versus its theoretical max — keeps falling. Coal has gone from being the base load (the steady floor of power that’s always on) to a flexible backstop that throttles down hard in the sunny middle of the day and steps in during the hot evenings.

“The coal fleet increasingly has a change in role from the base load provider that always needs to deliver power throughout the entire day to meeting demand in shoulder hours.”

There’s a subtlety: China is also building so much solar that not all of it shows up as generation. Some gets curtailed — switched off because the grid can’t use it that moment, like a tap left running with nowhere for the water to go. Curtailment is probably higher than China officially reports, partly because the rules for which plant gets dispatched still favor coal and take years to rewrite. The optimistic flip side: once those market rules catch up, there’s a backlog of already-installed solar that can start generating without building anything new.

India isn’t going to be China-with-a-15-year-lag

The lazy assumption is that India today is China 15 years ago, so it’ll repeat China’s coal binge. Fulgham argues no, for two reasons.

One: India is far more electricity-efficient — it produces more than twice the GDP per unit of electricity, because it leans on services rather than heavy manufacturing. It simply doesn’t need as much power to grow.

Two: India arrives 15 years late to a world where solar and wind already exist and are cheap. Its renewables-per-person is already five times what China’s was at the equivalent stage. 2025 gave a preview — a mild monsoon meant less air-conditioning, demand growth was unusually low, record renewables filled the gap, and fossil generation dipped. That won’t be structural for a few more years, but Ember expects India’s coal generation to peak at three to four times lower than China’s peak — and possibly as early as 2030, far sooner than the 2040s most forecasters assumed.

Batteries: the threshold that disappeared

The grid integration story is the most interesting part. There used to be a natural ceiling on solar: pump too much in at midday and you either curtail it or you can’t ramp other sources down fast enough. Batteries dissolve that ceiling by shifting sunshine to the evening.

Battery installations hit 250 gigawatt-hours in 2025, up 46% in a year. Ember built a clean metric: take the world’s daily added solar (a touch under 2 TWh) against the new battery capacity, and you find batteries can now time-shift 14% of all new solar generation away from midday — a big step up.

The punchline is that you don’t need 100% battery coverage to tame the midday glut. About 50% is enough: charge the batteries with the midday surplus, release it morning and evening, and effectively solar stops growing in the middle of the day and starts growing in the shoulder hours (the ramps either side of the peak) where the grid actually has room. Chile and Australia are already above 50%. The US sits at 20%, China 18%, the EU just 9%. The implication: batteries need to grow two to three times faster than solar to catch up — and cheap batteries have effectively turned solar into a dispatchable source you can call on when you want.

It won’t be smooth. Batteries earn money from arbitrage — buy power cheap, sell it dear — and that depends on price volatility. Texas (ERCOT) had a battery boom, then a couple of calm-weather years with low price spreads, so merchant batteries underperformed and the next wave got harder to finance. Expect boom-bust cycles market by market, but solar and batteries make each other’s economics work, so the long arc is up.

Natural gas and nuclear: the exceptions

Nuclear is still essentially flat globally; almost all the recent growth is in China, and even there it grows nowhere near as fast as wind and solar. In China and India there’s so much coal still in the stack that a steady nuclear plant can displace base load without worrying about curtailment — a different problem than in France, where reactors now dial down at midday to make room for solar.

Gas is the genuine outlier, and it’s almost entirely a US story. Global gas generation rose a modest 30-40 TWh in 2025 — about 18 times less than solar. The US (and Canada) have absurdly cheap domestic gas, partly a byproduct of oil fracking, so the AI data-center load boom is being met heavily with new gas plants. Everyone else imports gas as LNG, and recent shocks — the Strait of Hormuz scare, Russia in 2022, wobbles in Qatari supply — have spooked importers like Japan, Korea, and Europe into building renewables faster to escape it. So the US curve bends up; the global one doesn’t.

Key Takeaways

Solar generation grew 636 TWh in 2025 — roughly 2x the UK’s annual electricity demand, and ~3/4 of all new global electricity demand.
The 30% year-over-year growth rate was the highest in eight years, defying the expected S-curve flattening; the 10-year average is 27%.
Solar is the fastest source ever to climb from 100 to 2,000 TWh (10-12 years); it’s now past 2,700 TWh but still only 8-9% of global electricity.
Capacity ≠ generation. Capacity is nameplate potential; generation is actual delivery. Ember tracks generation to silence the “solar isn’t real power” critique.
Global fossil generation fell in 2025 — only the 5th time this century — with coal falling in both China and India.
China keeps building coal (fast build times, a hangover from 2020-21 shortages) but coal’s capacity factor keeps dropping; coal is shifting from base load to flexible peaking/shoulder duty.
Curtailment (dumping unusable solar) in China is likely higher than reported; outdated dispatch rules favor coal and take years to reform — but that means a backlog of generation can be unlocked later for free.
India’s electricity intensity is less than half China’s (services vs heavy manufacturing), and its renewables-per-capita is already 5x China’s at the equivalent stage.
India’s coal generation is forecast to peak 3-4x lower than China’s, possibly by 2030 — not the China path delayed, a genuinely different trajectory.
Battery installs hit 250 GWh in 2025, +46% YoY; ~14% of new global solar can now be time-shifted.
Only ~50% battery-to-solar coverage is needed to remove the midday peak entirely; Chile/Australia are above 50%, US 20%, China 18%, EU 9%.
Batteries make solar effectively dispatchable, pushing solar growth into shoulder/evening hours (visible in California and Australia).
Battery economics depend on arbitrage spreads, so expect boom-bust cycles (e.g. Texas/ERCOT) market by market.
Natural gas growth is almost entirely a US phenomenon, driven by cheap domestic shale gas and AI data-center load; gas importers (Japan, Korea, Europe) are accelerating renewables to escape LNG dependence.

Claude’s Take

This is a clean, numbers-first conversation between two people who clearly read the source report and don’t oversell. No hype words, no exclamation marks, just a steady walk through one think tank’s data. The honest framing up front — “this is a table setter, the real thesis is coming in a few weeks” — is a point in its favor; they’re not pretending a data review is a grand theory.

The genuinely useful ideas here are the metrics, not the cheerleading. The 100-to-2,000 TWh “time to scale” comparison and the 50% battery-to-solar ratio are the kind of mental yardsticks that survive past the episode. The repeated honesty about curtailment and about capacity-vs-generation also inoculates the discussion against the usual solar boosterism — they keep flagging where the official numbers probably overstate reality.

What to keep a skeptical eye on: it leans entirely on Ember, which is a clean-energy advocacy-adjacent shop, so the framing tilts optimistic by construction. The exponential-curve claim is real but the speaker himself admits his “this is the last year” instinct has been wrong for a decade — which cuts both ways: it could keep being wrong, or the bend could finally arrive in 2026 as he half-predicts. And “fossil fell in India” rests substantially on a mild monsoon, which they’re upfront about — that’s weather, not yet structure. Net: a high-signal, low-drama briefing. An 8 — not a 9 because it’s explicitly scaffolding for a thesis that hasn’t been delivered, and single-source.