Why Automation Multiplies the Workforce and Then Moves It Somewhere With Cheaper Electricity
A robot is not a worker. It is a capital good with a power cord, and the cord is the part that determines where it gets plugged in.
This distinction is the whole argument. The political class in the West has convinced itself that automation is the mechanism by which manufacturing returns: build the robots, and the factory can come home, because the robot does not need a wage that clears a Western mortgage. The reasoning is half right, and the half that is wrong is the half that matters. The robot does make the plant independent of the wage. That is precisely why the plant no longer needs to be near the worker. Once labour stops being the location variable, the location variable becomes the electricity bill.
The robot is a workforce multiplier. It multiplies output per worker, which is the definition of productivity, and it does so at a scale no previous capital good has matched. It also multiplies the consequences of a decision made decades ago and not quickly reversed: the grid.
I. The Multiplier, Stated Properly
Start with what the robot actually does to labour, because the popular account gets this backwards in both directions.
The robot does not mainly destroy jobs. It destroys tasks and expands output. The empirical literature on robot adoption finds measurable productivity gains with limited aggregate employment loss, a falling manufacturing employment share, and rising total output. The plant that installs the robot produces more with the same headcount, or the same with fewer, and the difference shows up as output per hour rather than as a headline unemployment statistic.
The Chinese data is the cleanest natural experiment, because China ran the largest automation programme in history and published the labour numbers alongside it.
China’s manufacturing workforce fell from 115 million in 2013 to below 85 million in 2025. That is a loss of more than 30 million factory jobs. Over the same period, Chinese exports hit record highs and the country’s share of global manufacturing rose to roughly 27–30 per cent.
Read those two facts together. Thirty million fewer manufacturing workers. A larger share of global manufacturing output than the United States and the European Union combined. That is what a workforce multiplier looks like when it is measured honestly. Output went up. Headcount went down. The multiplier accrued to the capital owner and the export account, not the worker.
Foxconn replaced 60,000 workers at a single plant with automated systems. At BYD’s Zhengzhou complex, 98 per cent of the welding is done by robots and a new electric vehicle rolls off the line every 46 seconds, overseen by a handful of human workers across a site of more than ten square kilometres. The humans are still there. There are just very few of them, and they supervise rather than weld.
The multiplier is real. It multiplies output per worker. It does not multiply the number of workers, and the political promise that treats those two statements as equivalent is the source of the entire confusion.
II. The Location Variable Flips
For forty years, the argument for manufacturing in China was the wage. Chinese workers were cheaper, so the plant went where the workers were cheap. That era ended because the robot became cheap instead.
The average cost of a newly installed industrial robot has fallen below $3,300 in some markets, according to IFR data. At that price the robot is a commodity input, not a capital project. The hardware is sold globally by FANUC, ABB, KUKA, Yaskawa and a growing roster of Chinese manufacturers. There is no proprietary advantage in owning one.
When the robot is a commodity and the wage is no longer the constraint, the remaining variable is the cost of running the robot. That variable is electricity, and it is not a commodity in the same sense, because electricity cannot be shipped. A factory can buy a robot from anywhere. It must buy the electron from the grid it is attached to.
The physics of the automated plant sharpens this rather than softening it. A dark factory runs around the clock, with no shift changes, no lunch breaks, and no climate control calibrated for human comfort. It consumes electricity the way a data centre does: continuously, predictably, and in enormous volume. The robot does not reduce the energy intensity of manufacturing. It concentrates it. A plant with 98 per cent robotic welding and one vehicle every 46 seconds is, functionally, a large electrical load with a production function attached.
That is the flip. When labour dominates the cost structure, you locate where labour is cheap and tolerate the electricity. When capital and energy dominate, you locate where energy is cheap and tolerate the wages. The robot did not abolish geography. It changed which geography matters.
III. The Price of the Electron Is the New Wage
The numbers are not close, and they are not disputed.
BloombergNEF’s comparison of final industrial electricity prices across 37 markets puts US businesses at an average of $110 per megawatt-hour** in 2025, Chinese businesses at **$89, German companies at $226**, and British companies at **$327. On a cents-per-kilowatt-hour basis, US industry paid about 8.62 cents in 2025 against roughly 22.64 cents for German industry in the second half of the year. Chinese industrial electricity runs in the 6 to 9 cent band.
The Asian comparison is tighter and more instructive. India’s industrial tariff is around $95 per megawatt-hour**, against an estimated **$60–80 per megawatt-hour in China, Vietnam, Thailand and the United States. Vietnam’s business tariff is around $0.079 per kilowatt-hour**, Thailand’s around **$0.123, Malaysia’s around $0.139. This is why the factories are moving to Thailand and Malaysia. This is also why the Western tourist is no longer the priority in Thailand. This is why the Thai woman is no longer begging for his attention. He’s dating a Chinese lineman, now. The Chinese guy is still in his prime working age and besides his earnings, he employs half of her extended family too. The Western man could compete in the same field, but his own habits and inertia prevent him from improving, so he resorts to complaining and manosphere.
Put those numbers next to the robot. A robot that costs $3,300 and runs continuously for a decade is, to a first approximation, a machine that converts electricity into output at a fixed rate. The plant’s competitiveness is the price of that conversion. A German manufacturer paying 22.64 cents is paying roughly 2.6 times the American price and 3 times the Chinese price for the same input, on a machine that never sleeps and never takes a lunch break.
For Finland competing in this field would be bordering ridicuously easy child’s play: accept the data center, require a part of the compute serve domestic capacity and use the excess heat in process industries, particularly in precision fermentation, and voilà, almost all fiscal worries GONE for the next forty years and the demographics would improve, but habits and inertia, and Diesel Identity.
That is not a wage gap. It is an energy gap, and it does not close with a training programme.
IV. Where the Robots Actually Went
The installation data is the revealed preference of the capital, and it is unambiguous.
In 2025, factories worldwide installed more than 600,000 new industrial robots, an 11 per cent increase, taking the global operational stock to a record 5 million units, more than double the level seven years earlier.
China accounted for 59 per cent of global installations, with 354,000 units, up 20 per cent on the year and almost 60,000 above the previous annual record. Chinese manufacturers supplied 195,000 of those units, a 55 per cent domestic market share.
The United States became the world’s second-largest market, overtaking Japan, with almost 38,500 installations, up 12 per cent and the third-highest annual total on record. Japan installed 36,219, down 19 per cent. South Korea installed around 30,000, down 1 per cent. Germany, the largest European market, fell 8 per cent to fewer than 25,000 units.
India installed almost 10,500 units, up 15 per cent, after growing at an average of 27 per cent a year between 2020 and 2025.
China’s operational robot stock passed 2 million units in 2024, the largest of any country. Its manufacturing robot density reached roughly 470 units per 10,000 manufacturing workers, against 307 in the United States. South Korea leads the world at 1,220 per 10,000 manufacturing employees, with Singapore at 818, Germany at 449 and Japan at 446.
The pattern is not subtle. The country with the cheapest industrial electricity on the list is installing more robots than the rest of the world combined. The countries with the most expensive electricity are installing fewer, and in several cases fewer than they did the year before.
This is not a story about Chinese industrial policy alone, although the policy is real. It is a story about the arithmetic of a machine that runs on watts. The robot went where the watt was cheap.
V. The Grid Is the Binding Constraint, and It Does Not Respond to Subsidy
The Western response to this data has been to subsidise the factory. That is the wrong instrument, because the constraint is not the factory.
The United States is now in a demand-growth regime it has not occupied for twenty years. The Energy Information Administration expects electricity demand to grow roughly 1.3 per cent a year through 2050, more than double the 0.6 per cent rate of the previous two decades. Data centres alone could grow from 4 per cent of US electricity load in 2023 to 9 per cent by 2030, according to EPRI estimates. The grid is being asked to absorb a new class of load that arrives faster than conventional industrial demand, demands enormous capacity, and operates with a near-constant profile.
China, in the same period, added 543 gigawatts of generation capacity in a single year — 315 GW of solar and 119 GW of wind — taking total installed capacity to 3.89 terawatts, up 16.1 per cent. That is roughly twice the total electricity capacity of Germany, added in twelve months.
The comparison is not a policy comparison. It is a physical comparison. A subsidy can pay for a fab. It cannot conjure a gigawatt that does not exist, and it cannot shorten an interconnection queue by writing a cheque. The Chinese robot fleet runs on a grid built at a pace the Western grid is not currently matching, and the robot does not care which government is in office.
The robot is mobile. The substation is not.
VI. The Political Delusion
The political promise is that automation brings the factory home. The arithmetic says automation is the technology that made the factory homeless in the first place.
A robot does not need a visa, a green card, a union, or a vote. It needs a connection. Once the connection is the binding constraint, the factory’s location is determined by the grid, and the grid is determined by a planning process no trade policy can accelerate by more than a few years.
So the Western political class is caught in a promise it cannot keep. It promised the return of the manufacturing wage. It delivered the automation that removes the wage from the cost structure. It promised the return of the plant. It cannot deliver the plant, because the plant now wants a cheap electron and the Western grid is priced at two to three times the Chinese level.
The result is a cultural grievance dressed as an industrial policy. The voter is told the factory is coming back. The factory is not coming back, because the factory is not a building any more. It is a load, and the load is going where the load is cheapest to serve.
Also the voter from the perspective of the employer, is already in the retirement pipeline. The public administration would prefer him to work so they could save on his pension, but the factory owner sees someone he has to retire in a decade. This is a stereotype because habits and inertia cause it to work on a population level, even if you personally were sharp and fit.
VII. The Multiplier and the Distribution
If the robot is a workforce multiplier, the question that follows is who receives the multiplied output. The Chinese data gives the answer, and it is not flattering to any of the parties involved.
Thirty million fewer manufacturing workers. A larger share of global manufacturing output than the US and EU combined. Record exports. That is the multiplier, and it accrued to the capital owner, the exporter, and the state that supplied the cheap power. It did not accrue to the worker who was multiplied out of the plant.
The Western version of the same transition will produce the same distribution, with one difference: the Western worker has a political system that will try to compensate him, and the compensation will be fiscal. That is a transfer, not a production function. It buys time and it buys peace, and it does not buy a factory, because the factory is now a load and the load is going where the electrons are cheap.
Expensive, subsidized workers bundled with expensive power in Finland and Germany can’t compete with equally skilled, cheaper workers with cheaper power in Kenya and Nigeria. German industry is already moving. Finnish isn’t because habits and inertia cause it not to.
The humanoid robot sharpens this rather than softening it. China is already at mass production: UBTech’s Liuzhou plant is designed to produce one humanoid every 10 minutes. AgiBot shipped about 9,700 units in the first half of 2026, Unitree more than 7,000, and the Ministry of Industry and Information Technology projects 2026 output above 100,000 units. The humanoid is the general-purpose version of the same capital good, and it will follow the same cord.
VIII. What the West Can Actually Do
The fatalist version of this argument says the West has no cheap electrons and therefore no future in automated manufacturing. That version is wrong, and it is wrong in a way that matters, because the West is not short of cheap electrons. It is short of cheap electrons where the load already is, and the distance between the two is a border, a bottleneck, and a permitting queue.
The geography is not the problem. The transmission is.
VIII.1 The Cheap Electron Is Already Inside the West
The United States has the Sunbelt, and the Sunbelt is not a metaphor.
Texas is the clearest case. ERCOT is an energy-only market with its own interconnection queue, its own planning regime and comparatively light permitting, and the result is that Texas has led the country in solar and battery additions for several years running, with industrial electricity prices among the lowest in the developed world. The state added more grid-scale storage than any other in recent years and continues to attract load — data centres, fabs, hydrogen and electrified industrial projects — precisely because the power is available and the queue moves.
The same logic runs through Arizona, New Mexico and Nevada: high irradiance, cheap land, and a solar resource that produces at industrial scale for most of the year. The US industrial average of roughly 8.6 cents per kilowatt-hour in 2025 is roughly a third of the German level, and the Sunbelt sits below that average.
This is why Elon loves Memphis, Texas & Louisiana. Power is cheap.
Europe has the equivalent, and it is the Iberian peninsula.
Spain and Portugal have among the lowest wholesale electricity prices in Europe, and in the solar hours they are frequently the lowest on the continent — at times clearing near zero or negative as the midday PV fleet oversupplies the domestic load. Spanish industrial tariffs are materially below German ones. The resource is there, the generation is there, and the market design is there. What is not there is the ability to move the electron to the German factory that needs it.
That is the bottleneck, and it is the same bottleneck as the American interconnection queue in a different costume: the Pyrenees. France–Spain interconnection capacity is small relative to the two countries’ generation, and expanding it has been discussed for two decades and delivered in increments. The cheap electron exists and cannot cross the mountain at scale.
So the West’s position is not “no cheap power.” It is “cheap power behind a bottleneck.” That is a transmission problem, and transmission problems are capital-intensive, slow, and — crucially — solvable. They are not solvable on a political cycle, which is why the political class prefers to talk about factories.
VIII.2 The Cheap Electron Is Also Directly South of the Border
Here the geography becomes politically uncomfortable, because the cheapest electron in the Western hemisphere is not in the United States. It is in Mexico.
Mexico has the irradiance of the Sunbelt with lower land costs and lower labour costs, and it has a manufacturing base already built along the northern border. The nearshoring wave of the last several years has pushed industrial investment into Monterrey, Querétaro, Saltillo and the border corridor at record levels, and the constraint on that growth has been electricity supply rather than demand. CFE’s generation adequacy and the permitting regime for new industrial connections have been the binding limit, not the willingness of manufacturers to locate there.
Read that carefully, because it is the whole essay in one sentence: the factory did come back to the region. It did not come back to the country. It went to the near shore, where the electron is cheaper and the connection is available.
VIII.3 The Mediterranean, and the Elephant
The same pattern runs through Europe’s south.
The Mediterranean basin — Spain, Portugal, Greece, and increasingly the southern shore — has the resource. The Saharan and North African irradiance is roughly double that of central Europe, on the order of 2,000–2,500 kWh per square metre per year against roughly 1,000–1,100 in Germany. The capital cost of a solar plant there is lower per megawatt-hour than almost anywhere in Europe, and the land is available.
The interconnection projects that would move that power north are real and already moving:
ELMED, the Tunisia–Italy link, is a 600 MW high-voltage direct current cable across roughly 200 kilometres of the Sicilian channel, co-funded by the EU and targeted for the late 2020s. Xlinks, the Morocco–UK project, proposes on the order of 10.5 GW of Moroccan solar and wind with 5 GW of storage and a 3,800-kilometre HVDC cable to supply a substantial share of British electricity, with a target in the early 2030s and a financing history that shows how hard these projects are to close. GREGY, the Greece–Egypt link, is in the same family. Morocco’s Noor Ouarzazate complex, at around 580 MW of concentrated solar capacity, has operated for years as the demonstration that the resource is bankable at utility scale.
These are not speculative. They are approved, co-funded, and in various stages of procurement. They are also slow, capital-heavy, and dependent on subsea cable manufacturing capacity that is globally constrained — the same class of serial bottleneck that the SMR and the fusion pilot suffer from.
The largest cheap-electron resource available to Europe is not in Europe at all, and everyone in Brussels knows it. The reason the Sahara has not been connected is not physics. It is the same reason the Pyrenees crossing has taken twenty years: a capital project that crosses a border, a regulatory regime and a security perimeter requires a coalition that outlasts several governments, and the European coalition does not reliably outlast one.
The factory will not wait for them. It will locate where the electron is available today, which means the Sunbelt, the Mexican border and Iberia, with the Maghreb as the second act.
VIII.4 The Bulge and the Irradiance
The Sahara is the export story. The population bulge is the domestic story, and it is further south.
West Africa and East Africa are the two places on earth where the demographic curve and the irradiance curve coincide, and the coincidence is the whole argument.
Africa’s population is roughly 1.5 billion today and is projected to reach 2.3 billion by 2043 and around 2.5 billion by 2050. The median age is under 20, against 44 in the European Union, and roughly 40 per cent of Africans are under 15. More than a quarter of the world’s people will be African by mid-century.
West Africa is the demographic centre of gravity. Nigeria is projected to reach 377 million people by 2050, which would make it the third most populous country on earth. Niger, with a fertility rate near 7 children per woman, is projected to reach 68.5 million from 21.5 million in 2017. The Sahel is the fastest-growing place in the world, and it is also the poorest.
East Africa is the second bulge. Ethiopia is projected to reach 225 million by 2050 from roughly 139 million today. The East African Community’s population is projected to rise from 331 million in 2023 to 647 million by 2050, with the DR Congo, Tanzania and Uganda all entering the global top twenty.
Now put the map of the sun on top of the map of the people.
West African irradiance runs from roughly 1,850 to 2,350 kWh per square metre per year, with Niger and Mali at the top of the range. Around 87 per cent of the ECOWAS region has long-term direct normal irradiance above 1,800 kWh/m²/year, and the Sahel and Sahara sit above 2,000. That is roughly double the German figure.
East African irradiance runs 1,800 to 2,200 kWh/m²/year across the region, with Kenya at around 2,400 and Ethiopia at roughly 5.2 kWh/m²/day. Kenya’s natural conditions for solar are among the best in sub-Saharan Africa.
The two curves do not merely overlap. They reinforce each other. The places with the most people coming are the places with the most sun.
VIII.5 The Grid That Will Not Arrive
Here is the constraint that makes the African case different from the European one.
600 million Africans lack access to electricity, roughly 43 per cent of the continent, and sub-Saharan Africa accounts for about 85 per cent of the global access gap. Nigeria alone has more than 80 million people without power, the largest number of any country in the world. In West Africa, electrification is below 58 per cent in urban areas and below 25 per cent in rural areas.
The grid is not going to close that gap on the schedule the population is growing. The population is doubling faster than the transmission network is being built, and the capital required to build it is not being committed at the required scale. Africa receives less than 3 per cent of global energy investment despite holding 60 per cent of the world’s best solar resources. It has 60 per cent of the best solar resources and about 1 per cent of installed solar capacity.
So the African household does what the Finnish household did not have to do. It does not wait for the grid. It buys the panel and the battery, because the alternative is kerosene, a diesel generator, and a business that closes at dusk.
That is the demand, and it is not a subsidy story. It is a survival story.
The installation data confirms the market is not waiting for the policy. Africa installed 4.5 GW of solar in 2025, a 54 per cent increase on 2024 and the continent’s fastest year on record. South Africa led with 1.6 GW, followed by Nigeria at 803 MW, Egypt at 500 MW and Algeria at 400 MW. The Global Solar Council forecasts more than 31.5 GW of cumulative capacity by 2029.
Kenya is the proof of concept for the East African model: more than 80 per cent of its power comes from renewables, and it has 173 MW of solar on a grid where solar is still less than 5 per cent of capacity.
The distribution matters more than the headline. Utility-scale accounted for 56 per cent of 2025 capacity, and the remaining 44 per cent came from distributed sources — and the Global Solar Council states plainly that the distributed figure is “clearly underestimated,” because the volume of panels imported to the continent far exceeds what utility-scale projects can absorb. The African market is not a utility-scale market with a rooftop add-on. It is a distributed market with a utility-scale layer on top, because the grid is the bottleneck and the roof is not.
The binding constraint is finance, not resource. Financing costs in Africa remain three to five times higher than in developed markets, which suppresses projects that would otherwise be viable. That is the same class of serial bottleneck as the subsea cable and the interconnection queue: a capital constraint no amount of irradiance can relieve.
This is where the Special Economic Zones genuinely help: they make cheap, Western financing magically happen.
VIII.6 The Strategy, Corrected
The West’s cheap-electron strategy has been framed as an export play: the Sahara powers Europe, the Mediterranean is the bridge, the cable is the bottleneck. That is one story, and it is real.
The larger story is that the cheapest electrons and the fastest-growing demand are in the same place, and that place is not waiting for a cable to Europe.
West Africa and East Africa will be served by distributed solar and storage because the grid will not arrive in time and the diesel is unaffordable. That is the largest new electricity market in the world, and it is being built one rooftop and one mini-grid at a time, with panels imported from the same global manufacturing base that the Finnish household is waiting on a subsidy to access.
The realistic strategy has four parts.
First, price the electron, not the factory. The binding constraint is industrial electricity and the connection to it. The levers that move it are grid permitting, transmission build-out, and the cost of capital for generation. A subsidy to a factory that cannot get an interconnection is a transfer to a consultant.
Second, build the copper. The West’s cheap electrons are behind the Pyrenees, the interconnection queue, and the subsea cable. Those projects determine where the robot goes, and they are underwritten by transmission policy rather than industrial policy. The country that shortens its queue and lays its cable keeps the load. The country that does neither exports the load to the neighbour.
Third, compete where the robot is not the whole cost. High-mix, low-volume, regulated, security-critical and proximity-critical production still rewards proximity, skill and institutional trust. The robot multiplies output per worker; it does not multiply the value of a certification, a clinical trial, or a defence clearance. That is where the expensive-electron economies still win, and it is a real advantage rather than a consolation prize.
Fourth, accept the near shore and the far south. The honest policy is to integrate the cheap-electron periphery into the domestic industrial base rather than pretend it can be relocated into the expensive-electron core. That means treating northern Mexico, Iberia, the Maghreb and — over a longer horizon — West and East Africa as part of the supply geography, with the grid, the customs regime and the security architecture built to match. It is not the promise that was made. It is the one the arithmetic supports.
The West’s choice is not whether to build the wire to the Sahara. It is whether to treat the African market as a development charity, a resource colony, or the industrial customer it is about to become. The first two framings produce a cable and a photo opportunity. The third produces a supply chain.
The sun is not the constraint. The financing is. And the financing is a decision, not a physical limit.
IX. The Close
The robot is the most productive capital good the industrial economy has ever deployed, and it is indifferent to the country it is deployed in. It does not care about the flag on the factory. It cares about the voltage and the price per kilowatt-hour.
That is why the automation boom is concentrating in the one large economy that built the generation capacity to serve it, and why the Western reshoring promise is running into a constraint that is physical rather than political.
The West is not without cheap electrons. It has the Sunbelt, Iberia, the Mexican near shore, and the largest untapped resource on the planet directly across the Mediterranean and down the African coast. What it lacks is the wire, the queue, and the coalition that outlasts a government.
You can print a subsidy. You can print a tariff. You can print a press release announcing the return of the factory.
You cannot print a gigawatt.
The robot follows the watt, and the watt is going to the Sahel, the Rift Valley, the Sunbelt and the northern Mexican border — wherever the people are and the power is cheap. The West can build the wire and be the market that supplies it, or it can watch the allocation go to the buyers who cannot wait.
References
International Federation of Robotics. (2026). World Robotics 2025: global operational stock 5 million units, +9%; more than 600,000 installations, +11%; China 354,000 units and 59% of global deployments; US ~38,500 units, +12%; Japan 36,219, −19%; Korea ~30,000, −1%; Germany <25,000, −8%; India ~10,500, +15%.
International Federation of Robotics. Robot density: Korea 1,220; Singapore 818; Germany 449; Japan 446; China ~470 per 10,000 manufacturing workers; US 307.
BloombergNEF, via Oenergetice. (2026, October 18). Final industrial electricity prices, 37 markets, 2025: US $110/MWh, China $89/MWh, Germany $226/MWh, UK $327/MWh.
Germany Insolvencies / industry data. (2026, June 1). US industrial electricity average 8.62 US cents/kWh in 2025; German industry ~22.64 cents/kWh in H2 2025.
Xpert.digital. (2026, February 17). Industrial electricity: China 6–9 cents/kWh, US ~8 cents/kWh, France 9–11 cents/kWh.
Economic Times / Indian power ministry. (2026, January 21). India industrial tariff ~$95/MWh; China, Vietnam, Thailand and US estimated $60–80/MWh.
Acclime Vietnam. (2026, September 9). Vietnam business electricity ~$0.079/kWh; Thailand ~$0.123/kWh; Malaysia ~$0.139/kWh.
Frontieras North America. (2026, May 27). BYD Zhengzhou: 98% robotic welding, one EV per 46 seconds; Xiaomi Changping dark factory; China manufacturing workforce 115 million (2013) to below 85 million (2025); average installed robot cost below $3,300 in some markets; dark factories consume electricity like data centres.
China electricity data, National Energy Administration via Mercom and Chinese industry reports. (2026, January). Installed capacity 3,890 GW at end-2025, +16.1%; 315 GW solar and 119 GW wind added; total additions ~543 GW.
US Energy Information Administration, via RBC Wealth Management. (2026, October 2). US electricity demand growth ~1.3% annually through 2050 versus 0.6% over the prior two decades; data centres from 4% of load in 2023 to up to 9% by 2030 (EPRI).
TD World / IEEE PES T&D. (2026, May 11). Hyperscale data centres arrive faster than conventional industrial load, demand enormous power, and operate with near-constant profiles.
Visual Capitalist. (2026, August 10). China’s share of global manufacturing reached 27% in 2025, ten percentage points ahead of both the US and the EU.
36Kr and industry reporting. (2026, September). UBTech Liuzhou humanoid plant, one robot every 10 minutes; AgiBot ~9,700 units in H1 2026; Unitree 7,000+; MIIT 2026 output projection above 100,000 units.
CNBC TV18 and Chinese media. (2026, September 8). China shifting AI data centres west to Inner Mongolia, Guizhou and Gansu for cheaper power; more than half of the data-centre project pipeline now in the north and northwest.
ENTSO-E and Spanish market operator data. Iberian wholesale electricity prices consistently among the lowest in Europe, with midday solar-hour prices at or below zero on high-production days.
European Commission. ELMED Tunisia–Italy 600 MW HVDC interconnection: co-funded project, targeted for the late 2020s.
Xlinks. Morocco–UK project outline: approximately 10.5 GW of Moroccan solar and wind, 5 GW of storage, roughly 3,800 km HVDC, target early 2030s.
GREGY Interconnector. Greece–Egypt HVDC link proposal.
Moroccan Agency for Sustainable Energy (MASEN). Noor Ouarzazate complex, approximately 580 MW of concentrated solar capacity.
Global Solar Atlas / World Bank. Global horizontal irradiance: North Africa and the Sahara approximately 2,000–2,500 kWh/m²/year; central Europe approximately 1,000–1,100 kWh/m²/year.
CFE and Mexican energy ministry reporting, via industry press. Industrial electricity supply and permitting as the binding constraint on Mexican nearshoring investment in Monterrey, Querétaro, Saltillo and the border corridor.
United Nations Economic Commission for Africa. (2025, September 23). 600 million Africans without electricity; Nigeria more than 80 million without power; Africa receives less than 3 per cent of global energy investment despite 60 per cent of world solar potential.
UN Sustainable Development Group. (2025, January 27). “600 million” without reliable access; more than 80 per cent of the global electricity access gap.
Illuminem. (2026, February 5). Africa holds nearly 60 per cent of the world’s best solar resources; West Africa electrification below 58 per cent urban and below 25 per cent rural; IRENA potential of 10 TW annually.
Institute for Security Studies, Africa Futures. Africa population roughly 1.5 billion in 2024, projected 2.3 billion by 2043; median age under 20; 40 per cent under 15; TFR 3.8 in 2024.
PopulationPyramid.net / UN World Population Prospects. Nigeria 2050 projection 359–377 million; median age 24.9.
UN Economic Commission for Africa. Niger population 21.5 million (2017), 68.5 million (2050).
Population Reference Bureau / UN. Ethiopia 2050 projection 225 million.
ISS Africa Futures. EAC population 331 million (2023) to 647 million (2050); DR Congo, Tanzania and Uganda in the global top 20 by 2050.
MDPI and ECOWAS sources. West Africa annual irradiation 1,850–2,350 kWh/m²; Niger and Mali highest; 87 per cent of ECOWAS above 1,800 kWh/m² DNI.
Copernicus / East Africa irradiance studies. East Africa GHI 1,800–2,200 kWh/m²/year; Kenya around 2,400; Ethiopia around 5.2 kWh/m²/day.
Global Solar Council, Africa Market Outlook 2026–2029, via pv magazine. Africa installed 4.5 GW in 2025, +54 per cent; South Africa 1.6 GW, Nigeria 803 MW, Egypt 500 MW, Algeria 400 MW; distributed share underestimated; financing costs three to five times developed markets; 31.5 GW by 2029.
UNECA. Kenya more than 80 per cent renewable generation.
SDG knowledge platform. Kenya solar installed capacity 173 MW, under 5 per cent of grid capacity.
OECD / AfDB / UNOPS, Africa’s Urbanisation Dynamics 2025. Urban population doubling from 700 million to 1.4 billion by 2050; cities absorbing 80 per cent of population growth.