Stages of Decarbonization

A quick note on the shape of this argument. These stages are not clean chapters where one ends before the next begins — they are overlapping S-curves. California was installing batteries while still adding solar; the UAE is signing battery contracts in the middle of its solar sprint. What defines a stage is which resource is the marginal investment — the thing capital rushes toward because it is the cheapest way to add the next clean megawatt-hour. And each stage carries the seeds of its own saturation: the more of a resource you add, the worse the economics for the next unit of it. The interesting question for any country is not “will it decarbonize?” but “which stage is it in, and will it make the jump to the next one?”

Stage 1

The low-hanging fruit: the solar (and wind) sprint

In this stage, countries build as much solar (and wind) as they possibly can — not because it is clean, but because it is cheap. Even after accounting for its modest capacity factor, utility-scale solar is the cheapest source of new electricity for most of humanity, with onshore wind competing closely. That was not always true. The global weighted-average cost of utility solar fell roughly 90% between 2010 and 2024, from around $417/MWh to about $43/MWh (IRENA). And the finding is robust to recent headwinds: Lazard’s July 2026 report has US solar construction costs up 18% year-over-year on tariffs and rates — and still finds utility solar and onshore wind the cheapest new-build generation, unsubsidized. Credit where it is due: early movers — Germany’s feed-in tariffs, early adopters who paid the high prices — kept the learning curve fed until costs collapsed for everyone else. The low-hanging fruit had to be lowered before it could be picked.

The tell that this stage is driven by economics, not climate virtue, is who is doing it. The UAE and Saudi Arabia — among the highest per-capita emitters on Earth — are building some of the largest solar plants in history. Not to be green; that is a by-product. Every megawatt-hour of sunshine they use domestically is oil and gas they can export instead of burning at home. In July 2026, Abu Dhabi’s Masdar reached financial close on a $6.1 billion, 5.2 GW solar project — and the fact that it comes bundled with 19 GWh of batteries is a preview of Stage 2, which we will get to. When petrostates build solar for the margin, the economics argument has won.

This will not be universal. Solar is an extremely diffuse form of generation — orders of magnitude better than photosynthesis, but nothing like the power density of a thermal plant. Dense, land-constrained industrialized places hit a ceiling fast: Singapore generates only a low single-digit percent of its electricity from solar, targets about 2 GW-peak by 2030, and has essentially conceded the point by planning to import several gigawatts of clean power over subsea cables instead. The same diffuseness means utility-scale solar usually sits far from load centers, so this stage quietly accumulates a transmission bill that comes due later.

Solar is inflexible — it produces when the sun decides, not when the grid needs it. Keep adding solar with nothing else and every new panel produces into the same saturated midday hours. If nothing is done, a solar-only strategy becomes un-investible: prices in solar hours go to zero, then negative (subsidies and PPA structures keep generators bidding below zero). This is not a hypothetical — it is the observed endgame of Stage 1 everywhere solar has scaled:

Spain is living the same moment: it logged more than 500 negative-price hours in 2025 — double its 2024 total — and 397 hours in the first quarter of 2026 alone, versus 48 in the same quarter a year earlier. The average price in those hours deepened from −€0.12/MWh to −€1.80/MWh in a single year. Spanish solar asset values have sagged, and the financial press is asking whether the boom went bust. That is what the end of Stage 1 feels like from inside: the resource is still cheap, but the market for its output at 1pm is gone.

Stage 2

The rise of batteries

Batteries are a genuinely remarkable technology, and they kill many birds with one stone. They are the best frequency-regulation asset the grid has ever had — they took over ancillary markets almost immediately. They flatten prices by design: a battery’s whole business model is to find a price spike and arbitrage it away. They work without solar, but they work best with solar — charging on those near-zero (or negative) midday prices from Stage 1 and discharging into the evening peak, displacing peaker plants, the most polluting and most expensive gas units on the system. Stage 1’s failure mode is Stage 2’s fuel: cheap afternoons are exactly what makes battery arbitrage lucrative. The world has noticed — 307 GWh of batteries were added globally in 2025 alone, up 48% in a year, with China accounting for more than half and gigawatt-hour-scale projects now under way on every continent.

California is the furthest along here. The CAISO battery fleet went from a rounding error in 2020 to over 16 GW today, and you can watch it displace gas in real time:

Courtesy of this exceptional pace of battery installation, days with extreme price spikes have collapsed, and the average day now has a deep midday price canyon and a shrinking evening ridge — the exact spread batteries live on:

This stage has an internal progression: duration climbs. Every battery added shrinks the arbitrage pie for the next one — the pie is first come, first served. Short-duration batteries maximize $/kW and go in first; as the 2-hour opportunity saturates, the market moves to 4-hour, then 8-hour, maybe stretching to 12. The limit is rarely chemistry; it is economics. Because a lithium battery’s power and energy are coupled, doubling duration roughly doubles cost — at the same time as each added hour of storage chases a smaller remaining spread. Note also a subtle dependency: in the short run, battery economics depend on the peaker plants they are displacing. As long as peakers set the marginal evening price, arbitrage margins are fat. Once batteries themselves set the price — increasingly true in California — margins compress toward the battery’s own opportunity cost. The contribution of any resource towards reliability is measured using ELCC (Effective Load Carrying Capacity) or NQC (Net Qualifyinf capacity) in California. As batteries saturate the market, their ELCC drops, and they count lesser towards resource adeqaucy. Basically at this point, marginal battery addition is only incrementally beneficial to providing reliability to the grid.

So Stage 2, like Stage 1, saturates. Three structural limits, in increasing order of difficulty. First, the economics stop supporting ever-longer durations well before seasonal scale. Second, batteries cannot bridge inter-seasonal gaps — no one charges in July to discharge in January. Third, and most under-appreciated: batteries are conveyor belts, not sources. They move energy from cheap hours to expensive hours. On a cloudy, low-solar day there are no cheap hours to move — the conveyor has nothing on it precisely when the grid is most stressed. California saw a version of this on July 13 this year: weak solar meant weaker battery charge, thinner evening discharge, and more gas online to fill the gap. One bad day is fine. A grid designed around only sun and conveyors is not.

Stage 3

The hard part: the last stretch is a different problem

Stages 1 and 2 are the proven, cheap, at-scale path, and they will carry many grids a long way — I’d say to roughly 80% clean energy in favorable geographies. California is the live experiment: its clean share of demand has climbed from around 43% in 2020 to nearly 58% in 2025, on almost nothing but solar, wind, and batteries layered onto legacy hydro and nuclear.

Meanwhile the demand side has stopped cooperating. Electrification of cars and heating was the plan; data centers came out of syllabus. Load growth is back after two flat decades, and it is arriving fastest in exactly the places (India, Indonesia, much of Africa) where standards of living should and will rise. A resilient grid does not just match energy to demand day by day — it holds excess capacity in every hour. California’s planning reserve margin requirement sat around 15–16% for years; with a variable-heavy fleet, the effective margin regulators now target is drifting toward the mid-20s, because the capacity value of the marginal solar panel or 4-hour battery falls as you add more of them. Stage 3 is where you pay for reliability explicitly.

I think of Stage 3 not as one problem but four:

1. The bad-day problem (hours to a day): a cloud deck or a wind lull — manageable with the Stage 2 toolkit, at the cost of overbuild.
2. The dunkelflaute problem (multi-day): consecutive dark, still days that outlast any economical lithium duration.
3. The seasonal problem (months): winter demand against summer sun — a storage problem no conveyor belt solves.
4. The growth problem: serving large new load with high reliability while the reserve margin math gets harder.

The tempting answer is “just build more solar and batteries.” Abu Dhabi is running that experiment at project scale: Masdar’s round-the-clock plant pairs 5.2 GW of solar with 19 GWh of batteries to deliver 1 GW of firm power — a roughly 5x firming multiplier, for $6.1 billion, in one of the best and least seasonal solar resources on Earth. That math is the caution, not the blueprint. Even in the desert, making sunshine firm quintuples the hardware; try the same trick grid-wide, at latitudes with real winters and week-long cloud decks, and the multiplier explodes — you end up overbuilding for the worst week of the year and drowning in curtailment the other fifty-one. A single project can buy its way to firmness. A whole grid cannot. That is why Stage 3 needs genuinely different tools.

Long-duration storage attacks problems 2 and partially 3. The requirement is specific: decouple energy from power (huge kWh, modest kW), get the capital cost of the energy component very low, and accept mediocre round-trip efficiency and even high opex — a resource used a few hundred hours a year can afford expensive hours, it cannot afford expensive existence. The unflashy proven answer is pumped-storage hydro: century-lifetime assets, excellent grid response, and the overwhelming majority of all stored grid energy on Earth today (~180 GW globally, with China building it at extraordinary pace). Where the geography exists, it is the way to go. Where it doesn’t, the electro-chemical and thermal contenders — iron-air (Form Energy’s 100-hour battery, targeting ~$20/kWh on the energy component), compressed air, heat batteries — are racing to hit the same spec sheet. The arena is ripe for innovation and new technologies though.

Firm, dispatchable clean generation attacks problems 1, 3, and 4 at once: resources you can dial up on a cloudy Tuesday and run flat-out through a dark week — nuclear and geothermal being the inherently clean candidates. Studies have found that including firm low-carbon resources cuts the cost of a fully zero-carbon power system by 10–62% across scenarios, with the advantage growing as you approach 100%. CAISO’s own long-horizon transmission and busbar-mapping work says the same thing in procurement language: the 2045 portfolios are where new geothermal and long-duration storage finally show up in size. I’m personally rooting for resources that are inherently clean rather than clean-by-mandate (which is why I list nuclear and geothermal, not gas-with-CCS), and I remain intrigued by thorium’s long game — but the honest answer is the winners aren’t decided, and countries dawdling through Stages 1 and 2 have bought the R&D world some time to decide them. Firm resources also carry a geopolitical dividend: a country running on domestic heat — uranium’s energy density makes stockpiling trivial, geothermal needs no fuel at all — is not tied to anyone’s whims, a peril that has become vivid recently. I am torn when it comes to CCS though; because of its clean attribute being market/regulation dependent, it wouldn't be a preferred choice. However, gas is pretty important for reliability until we have these novel generators. The conundrum them becomes whether to go for CCGTs, whose relatively low flexibility fare well with carbon capture, but then it extends the fossil plants in operation; OR shut down existing CCGTs, and keep peak plants based on less efficient OCGTs. As of today, they are not very CCS friendly because of the distinct thermal cycles of the carbon capture and power generation process. However, because of intermittant operation, their capacity factors can be dialled down as we get more and more renewables onboard, a slow transition out of fossil fuels.

I believe that the real constraint in Stage 3 may be market design, not technology. A resource that runs 300 hours a year earns almost nothing in an energy-only market, no matter how essential those 300 hours are — the “missing money” problem. Stage 1 and 2 resources could ride merchant economics; every Stage 3 tool above needs capacity payments, reliability procurement, or long contracts to exist at all. Countries that get stuck at the end of Stage 2 will mostly be stuck for institutional reasons, not engineering ones. If you look at the CAISO interconnection queue, there is hardly any firm generator in there. This isn't a victory for solar and battery, but the delay in getting firm resources. No transition will be complete without those firm resources.

Getting stuck, and why the sequence matters

The stages sound inevitable when laid out like this. One must be careful that in order to achieve total transition, we have to view it as a complete process with these steps, and not bask in glory of completion of step 1 or 2. Stall at the end of Stage 1 and you get Spain’s current anxiety: negative prices erode solar revenues, investment slows, and the fossil fleet quietly stays. Stall at the end of Stage 2 and you get a grid that looks 80% clean on a nice April afternoon and burns a fleet of aging gas plants every dark week — clean on average, fossil at the margin, with reliability hostage to the weather. The countries that make it through will be the ones that start building each stage’s exit ramp while the stage is still booming: batteries procured while solar is still cheap to add, long-duration storage and firm generation contracted while batteries are still eating peakers.

The encouraging part is that the sequence exists because each stage funds and de-risks the next. Cheap solar created the arbitrage that pays for batteries. Batteries are creating the flat, reliable-until-it-isn’t grid whose remaining gaps define exactly what Stage 3 must build. The low-hanging fruit was lowered by two decades of stubborn research and early movers who overpaid so the rest of us wouldn’t have to. The same favor is owed to Stage 3’s technologies now — and the countries with the means to do that lowering should treat it as the highest-leverage climate spending there is.