Where the learning curve actually lives — inside the factories
China mainly dominates these markets because it has produced a long-term industrial strategy for these technologies and has honed an optimised, modern supply chain as a result.
This module covers the following five lessons:
In this lesson you will walk through the five-step journey that turns sand into electricity, and see how each step contributes to the overall learning curve.
The learning curve from Module 5 doesn't happen in a spreadsheet. It happens in factories. Sand becomes electricity through a chain of five physical processes — each with its own learning curve:
The first three steps are about silicon. Quartz sand is refined to 99.9999% pure polysilicon. This step is energy-intensive enough that polysilicon plants are built where electricity is cheap. The polysilicon is then melted and grown into a single crystalCalled the Czochralski method, after the Polish scientist who found it in 1916 — by accident, when he dipped his pen into a molten metal bath instead of an inkwell.: a small seed crystal is dipped into the molten silicon and drawn slowly upward over about three days, pulling out one continuous 600 kg ingot. Diamond wire saws then slice each ingot into wafers thinner than two human hairs, where every micrometre of kerf lossThe silicon wasted as fine dust when a diamond-wire saw cuts an ingot into wafers. Cut it thinner and less silicon is lost as dust. saved means more cells per kilogram of silicon.
The last two steps make the wafer useful. The cell conversion stage adds doping (impurities that form the electrical junction), metallisation (printed silver conductors), and passivation (coatings that reduce energy losses). Those three steps together are what lift a plain wafer to 23–25% efficiency. Module assembly wires the cells together, sandwiches them between glass and a backsheet, frames the result, and fits a junction box. A single modern factory produces enough cells in a year to power roughly 200,000 homes.
PERC (Passivated Emitter and Rear Cell): the 2010s breakthrough. Adding a reflective dielectric layer on the back of each cell sends photons that would otherwise pass straight through the silicon back for a second chance. Worth about 1% absolute efficiency for a few cents per watt. PERC pushed commercial cell efficiency past 22% and was the industry standard through to roughly 2023.
TOPCon (Tunnel-Oxide Passivated Contact): the 2020s upgrade. An ultra-thin oxide film between the silicon and the rear contact slashes the electrical losses where metal meets wafer, lifting module efficiency to 23–24%. Now the dominant new-build cell architecture, and the main reason 2024–2025 modules are 1–2 percentage points more efficient than 2022 modules at roughly the same price.
Step through the chain below — each card adds the cost share of that process and where in the world it's concentrated. Together the five processes account for roughly 70% of total system cost.
Process: Quartz sand is refined into metallurgical-grade silicon, then purified further still — to 99.9999%, or about one stray atom in a million. That last step is what makes polysilicon so energy-hungry, and it is the raw material for every solar module.
Cost share: ~15% of total system cost
Key fact: China produces 85% of global polysilicon, concentrated in Xinjiang and Inner Mongolia where electricity is cheap.
The remaining ~30% is Balance of System: racking, wiring, inverters, permitting, labour. We unpacked BoS in Module 6, Installation is Key, which explains why two countries paying the same dollar-per-watt for modules end up with very different installed costs.
Polysilicon → Ingot → Wafer → Cell → Module. Each step has its own learning curve. The aggregate of all five improving simultaneously produces the ~24–27% learning rate we saw in Module 5.
The learning curve is not a spreadsheet abstraction. It's the cumulative effect of bigger furnaces, thinner wafers, more efficient cells, and faster assembly — compounded across billions of units.
In this lesson, you will zoom in on the physical improvements — bigger furnaces, thinner wafers, faster lines — that turn the abstract learning curve into real cost reductions.
The learning curve from Module 5 can feel abstract — a significant drop in price for every doubling of production. But inside the factories, it's tangible. Let's look at what actually changes.
Bigger furnaces improve silicon yield. A larger crystal puller produces a bigger ingot from the same melt, reducing energy and time per kilogram of silicon. Thinner wafers use less material per cell. In 2005, a standard wafer was 200 micrometres thick. Today it's under 150 — and heading lower. Every 10 micrometres saved means more wafers per ingot, more cells per kilogram of silicon.
Faster coating lines increase throughput. Modern cell production lines run at speeds that would've been unimaginable a decade ago, with automated handling systems that rarely touch a human hand. Higher automation reduces labour cost per unit and improves consistency — a machine doesn't have bad days.
Higher cell efficiency means more watts per square metre of silicon. PERCPassivated Emitter and Rear Cell — a reflective rear layer bounces missed photons back through the silicon for a second chance. The designs overtaking it are TOPCon and heterojunction, both of which cut losses where the metal contacts meet the wafer. cells pushed commercial efficiency past 22%. The designs replacing them are now past 25%. Each percentage point of efficiency is worth real money, because you need fewer panels, less racking, and less installation labour per kilowatt.
Each of these improvements is small on its own — perhaps 1–3% per iteration. But compounding across thousands of iterations and billions of units, they produce the three-order-of-magnitude cost drop we are witnessing. This is not magic. It's manufacturing innovation and discipline, applied at scale, over decades.
Move the slider in the interactive figure below to see how four key manufacturing metrics have changed over the last 25 or so years.
You can see the same metrics below in a side by side comparison.
In 2025, global solar module installation hit roughly 650 GW. At around 600 W per module, that's about 1.1 billion modules a year — one for every eight people on Earth, rolling off production lines annually.
If we extrapolate forward, cumulative installed modules will pass one per person worldwide by the late 2020s. And if manufacturing keeps scaling on its recent trajectory, by the early 2030s, the industry could be producing one solar module per person, every year. No other energy hardware has ever been built at this cadence. The closest analogy is smartphones, which are around 1.4 billion per year.
In this lesson, you will see how China came to dominate the solar supply chain, and what risks and responses that concentration creates.
Here's a fact that makes some people uncomfortable: China controls over 80% of the solar module supply chain. Polysilicon, ingots, wafers, cells, modules — China dominates them all. However, the picture is more nuanced than the 80% headline, as modules are not the whole system.
How did this happen? It's tempting to frame it as a geopolitical story, but it's really an economic one. China invested heavily in solar manufacturing when solar was expensive, and most other governments wouldn't touch it. That early bet, backed by industrial policy and cheap capital, created scale. Scale drove costs down. Lower costs attracted more investment. The same learning curve that makes solar cheap made Chinese factories dominant, and other manufacturing centres struggled to compete.
The numbers are striking. The largest polysilicon plant in China produces more than the entire rest of the world combined. Chinese cell factories operate at scales that make equivalent Western facilities look like pilot plants.
But the module is not the whole system. An installed utility-scale solar plant breaks roughly into modules (~30% of total cost), inverters (~5%), Balance of System hardware (~20%), and installation plus soft costs (~45%). Each component has its own supply geography. The chart below shows where each comes from — bar width is the share of total installed cost, and the fills show the origin mix. Read it as a global utility-scale average; Western markets diverge from this in ways the prose below covers.
Bar length is that component's share of the total cost. The percentages inside each bar show where that component is made.
Two of those four categories follow a very different geography from the module chain. The Balance of System (the racking, mounting, cabling, combiner boxes, trackers, and transformers) is bulky, heavy, and usually supplied locally. Shipping a simple aluminium frame across an ocean rarely makes economic sense when you can fabricate it thirty miles away. BoS is overwhelmingly local or regional. Installation labour and soft costs, (the labour, permits, interconnection, and developer overheads) are 100% local by definition.
Inverters sit somewhere between the two. Globally, Chinese firms still lead the market, Huawei and Sungrow alone shipped more than half the world's inverters in 2024, but Western alternatives operate at scale. These include SMA (Germany), Fronius (Austria), Enphase and Tesla (US), and Power Electronics (Spain). In Western utility-scale markets specifically, Chinese inverters are a minority.
Weighted across the whole installed system, Chinese content averages roughly 35% — a very different number from the 80% that describes the module chain. So when we talk about "manufacturing concentration," it is worth being precise: the module supply chain is highly concentrated; the finished installed system is not. The risks of single-source dependency, the diversification efforts already underway, and the inverter cyber-security story are considered in Module 8.
China's dominance is not a subsidy story or a labour-cost story. It's the learning curve applied with discipline: early investment → scale → lower costs → more investment → more scale. Two decades of that feedback loop produces a 97% share of ingots and wafers.
The implication: diversification is possible, but it requires starting from behind on the same learning curve. It will cost more in the short term. Whether that's worth paying for is a political choice, not an economic one.
In this lesson, you will trace the journey from raw minerals to a finished lithium-ion battery system, and see why the choice of cathode material is as much a supply-chain decision as a chemistry one.
Batteries have their own supply chain, and it's no less impressive. Like solar's, it runs as a clear four-step sequence — each step with its own learning curve and its own geography:
It starts underground. Lithium comes from brine evaporation ponds in South America or hard-rock mining in Australia. Graphite — mined or increasingly synthetic — ends up as the anode. The cathode minerals depend on which chemistry the cell is built around: nickel and cobalt for NMCNickel Manganese Cobalt — a lithium-ion cathode chemistry that delivers ~250 Wh/kg energy density at higher cost per kWh. The standard for EVs where range matters most. cells, iron and phosphate for LFPLithium Iron Phosphate — cheaper, longer-cycling, and safer than NMC, in exchange for ~30% lower energy density. Now dominant in stationary storage and short-range EVs.. The cathode is the most expensive single component in the cell, and it fixes most of the battery's behaviour — energy density, cycle life, safety, and ultimately cost. All of these raw materials are refined into battery-grade chemicals before anything is coated, and that refining step is increasingly concentrated in China.
NMC (lithium nickel manganese cobalt oxide): higher energy density, used in premium EVs and applications where weight matters. Supply chain exposed to cobalt (Democratic Republic of Congo) and nickel.
LFP (lithium iron phosphate): slightly lower energy density but cheaper, longer-lasting, safer, and built from abundant minerals. Now dominant in grid-scale storage and increasingly in standard-range EVs.
Cell manufacturing then takes the cathode and anode powders, coats them onto thin metal foils, stacks them with separators, fills the housing with electrolyte, and seals it. Each cell runs through "formation" — days of slow charge-discharge cycles that activate the chemistry. Contamination at the micrometre level ruins a cell, which is why these are some of the cleanest factories on Earth.
Pack assembly groups cells into modules, modules into packs, and packs into complete systems with thermal management and a Battery Management System — the brain that watches every cell's voltage, temperature, and state of charge.
Step through the chain below — each card adds the cost share of that step and where in the world it's concentrated.
Process: Lithium from brine (South America) or hard rock (Australia). Nickel and cobalt for NMC (Indonesia, DRC). Iron and phosphate for LFP (Australia, Morocco, China). Graphite for anodes (China, Mozambique, Madagascar). Refined into battery-grade chemicals.
Cost share: ~30% of cell cost
Key fact: China refines ~70% of the world's lithium and controls a similar share of cobalt and graphite refining.
LFP's rise is partly a supply-chain story because the cells do not contain cobalt or nickel. Both are classed as critical minerals, with concentrated supply and known geopolitical pinch points. Skipping them means cheaper raw materials and lower political risk. It's a big reason LFP now dominates new grid-scale installations.
In this lesson, you will see how the same kind of physical, factory-floor changes that drove the solar learning curve are now playing out inside lithium-ion battery plants — at a faster pace, and visible in every product from a wall-mounted home battery to a 20-foot utility container.
The learning curve from Module 5 can feel abstract for batteries too. Inside the factories, it's tangible. Let's look at what's actually changed.
Denser cells. Cell-level energy density has roughly doubled since 2010, from around 100 watt-hours per kilogram of cell to around 200. Better cathode formulations, thinner separators, and tighter electrode geometry all add up.
Longer-lived cells. Cycle life — the number of full charge-discharge cycles a cell can deliver before it drops to 80% of its original capacity — has roughly tripled. A modern lithium-ion cell built for stationary storage is routinely warranted at 8,000 cycles. That's more than two decades of daily cycling.
Denser packs. Cell-to-pack architecture has eliminated the intermediate module layer entirely. The same cells, packed more directly into the housing, deliver roughly twice the kilowatt-hours per litre of finished pack. Every product gets denser, not just the big ones.
Bigger factories. The largest single battery plant in 2010 ran at a few gigawatt-hours per year. The largest single plant in operation today — CATL's Plant 2 at Fuding — runs at over 120 GWh per year, with multiple lines under one roof. The same cells flow into home batteries, commercial skids, and utility containers alike.
Each of these improvements is small on its own — a few percentage points per generation. But compounding across billions of cells, over fifteen years, they produce the cost drop we charted in Module 5. It's the same shape as solar's, on a shorter clock.
The route is familiar. Chinese firms, principally CATL, BYD, and EVE, invested heavily in lithium-ion when EVs were a niche bet and grid-scale storage barely existed. Industrial policy, cheap capital, and an early call on the cheaper iron-phosphate cell chemistry produced the same scale-and-cost feedback loop we saw in solar. CATL alone now ships more battery cells per year than the entire non-Chinese world combined.
Move the slider in the figure below to see how four key battery-manufacturing metrics have changed from 2010 to 2025. All four are cell- or process-level, so the story holds whether you're thinking about a Powerwall on a wall or a multi-megawatt-hour container on a grid site.
The same metrics show up below in a 2010-vs-2025 spec sheet, with the headline outcome, pack price, on the highlighted bottom row.
The slider and table above show the inputs — the gains at the cell and factory level. The table below shows the outputs — what those gains let you buy, at each scale, in each era.
Those cell and process-level gains translate into different products at each scale we operate at. It's the same cells, from the same factories, but with different packaging.
In 2025, global lithium-ion cell production hit roughly 1.6 terawatt-hours — enough new battery cells every year for about 1 in 6 people on Earth to get a battery big enough to run their fridge for a day.
The announced gigafactory pipeline pushes production toward 5 TWh per year by the early 2030s. At that rate, the industry will be making enough cells annually for one fridge between two people.
In this lesson we'll see how the cost stack of an installed utility-scale battery plant breaks down — and why its weighted Chinese content sits noticeably higher than installed solar's.
With that industrialisation story in mind, let's see where the supply chain actually concentrates. The materials chain is even more concentrated than the cells. Roughly 95% of LFP cathode active material is refined in China, and a similar share applies to battery-grade graphite for the anode. Western and Korean alternatives exist — LG Energy Solution, Samsung SDI, Panasonic — and US gigafactories are coming online. But they're starting nearly a decade behind.
What does the supply chain look like for a finished BESS? An installed utility-scale battery plant breaks roughly into the cells and DC pack (~45% of total cost), the Power Conversion System (~10%), AC Balance of System hardware (~20%), and installation plus soft costs (~25%). The chart below mirrors the solar version from the previous lesson — bar width is the share of total installed cost, and the fills show the origin mix for a global utility-scale average.
Bar length is that component's share of the total cost. The percentages inside each bar show where that component is made.
The picture rhymes with installed solar but tilts harder towards China. Cells dominate the cost stack in a way modules don't, so the cell chain's ~80% Chinese share carries far more weight on the total. The PCS story is the inverter story: Sungrow and Huawei lead globally, with SMA, Power Electronics, and a handful of Western specialists providing the alternative. Containers, transformers, switchgear, cabling, foundations (the Balance of System), is mostly local, on the same logic as solar's BoS.
Weighted across the whole installed system, Chinese content runs close to 50%. Meaningfully higher than installed solar's ~35%.
One detail the chart hides, but mentioned in the text above. LFP cathode active material itself is ~95% Chinese-refined. This is part of why battery supply-chain diversification has become a sharper political question than solar's, and why Inflation Reduction Act (IRA)-driven US gigafactories, the EU's Critical Raw Materials Act, and Korean and Japanese capacity expansion are all moving at the speed they are.
Solar and batteries share the same ~80%-China headline at the cell/module level. The difference is what that 80% is attached to.
In an installed solar plant, the module is ~30% of total cost — concentration is partially diluted by inverters, BoS, and (especially) installation labour.
In an installed BESS, cells alone are around 50%, and there's no equivalent of solar's large install-and-soft costs slice.
That's why batteries — not modules — have become the politically sharper supply-chain conversation.
Seven questions ahead. If you're unsure on any of them, jump back through the module and try again. There's no limit on attempts.
Manufacturing is where the learning curve actually lives. The solar supply chain — polysilicon, ingot, wafer, cell, module — has five steps, each with its own learning curve; together they produce the ~24–27% aggregate rate from Module 5. The battery supply chain — minerals, cathode/anode, cell, pack, system — follows a similar pattern. Chemistry choices (NMC vs LFP) are as much supply-chain decisions as technical ones.
Manufacturing concentration in China (>80% of most steps, 97% for ingots and wafers) is not a geopolitical mystery. It's the learning curve made industrial: early investment, enormous scale, and two decades of discipline. That concentration brings low costs — and real single-source risk. Diversification efforts (US IRA, EU gigafactories, India PLI) are starting from behind.
The physical basis for the learning curve is tangible: bigger furnaces, thinner wafers, faster coating lines, higher cell efficiency, greater automation. Each improvement is small; compounded across billions of units, they deliver the three-order-of-magnitude cost drop.
Next up: In Module 8, we turn to the component that ties the system together — the inverter — along with cybersecurity, embodied carbon, and the full lifecycle picture.
Congratulations — Module 7: Supply Chains and Manufacturing is complete!! You now understand the five-step solar supply chain, why manufacturing is concentrated in China, how the learning curve works physically inside factories, and how end-of-life recycling closes the loop.
