Chapter 6: Capital-Intensive Sectors: Heavy Industry, Process Substitution, and Stranded Asset Risk

The No. 4 blast furnace at Handan Iron and Steel was tapped for the last time on a cold morning in March 2021. It had been relined six years earlier at a cost the company's engineers put at roughly 400 million yuan, and its designed campaign life ran another fifteen to twenty years. The shutdown order came from a provincial air-quality directive capping crude steel output across Hebei. But the decision to comply without a fight came from somewhere quieter: a spreadsheet in the planning department that priced the furnace's remaining output against a carbon price that did not yet exist. The reline had been booked as routine capital expenditure. By the spring of 2021 it had become a sunk cost with a truncated recovery horizon, and a small early entry in what will be, over the next two decades, the largest revaluation of industrial capital since the Ruhr began closing its mines.

This chapter is about that revaluation. Steel, cement, aluminum, petrochemicals, flat glass, ceramics—these sectors sit at the awkward center of China's decarbonization problem. They account for a disproportionate share of national emissions. Their assets are large, lumpy, long-lived, and immobile. And they are exposed to international trade in ways that furniture and apparel are not. The evidence assembled here points to a conclusion that runs against the grain of how light manufacturing upgraded: heavy industry will not green itself through incremental, compliance-driven improvements. It will move through a series of discrete, expensive, irreversible process substitutions, each of which creates winners, losers, and a category of assets that no longer earn their cost of capital.

Three arguments organize the chapter, and I want to state their limits up front. The decarbonization of heavy industry is a problem of capital stock turnover, not operating efficiency; the binding constraint is the timing and financing of replacement, not the optimization of existing plants. Process substitution—hydrogen steelmaking, electrified cement kilns, inert-anode aluminum—is technically feasible but economically contingent on input prices and policy supports that are themselves uncertain, which means the rational firm faces a real-options problem rather than a straightforward investment decision. And the stranded-asset risk that follows from these substitutions is not a side effect of green governance; it is the mechanism through which green governance reshapes industrial structure. Firms that anticipate it accumulate capability. Firms that do not accumulate exposure.

A word on where I stand. My earlier work on industrial clusters treated capital intensity as a background variable—a determinant of how easily firms could relocate or upgrade. In the cases examined here, capital intensity becomes the central object of analysis, and the cluster framework has to be stretched to accommodate it. I have relied on published cost curves, company disclosures, and a smaller number of interviews with planners and engineers in Hebei, Shanxi, and Inner Mongolia conducted between 2021 and 2023. Those interviews were not systematic. Where a claim rests on a single conversation rather than a documented source, I have tried to say so.

6.1 Decarbonization Challenges in Heavy Industry

6.1.1 The arithmetic of a heavy plant

Start with a number that organizes everything that follows. A modern integrated steelworks in China emits roughly 1.8 to 2.0 tonnes of carbon dioxide per tonne of crude steel, counting from coke making through rolling and crediting only energy recovered on site. Chinese crude steel output reached 1.03 billion tonnes in 2021, according to the National Bureau of Statistics. That implies sector emissions near 1.8 to 2.0 billion tonnes of CO2—roughly 18 percent of national emissions and close to 5 percent of the global total. Cement is comparable in scale. China produced about 2.4 billion tonnes in 2021, and the calcination of limestone plus fuel combustion in kilns releases roughly 0.6 tonnes of CO2 per tonne of clinker, putting sector emissions near 1.3 to 1.4 billion tonnes. Add aluminum, where primary smelting runs overwhelmingly on coal-fired electricity, and the three sectors together account for well above a third of national emissions.

These figures are familiar to anyone who works on Chinese energy policy. What gets less emphasis is their implication for the shape of the mitigation problem. In a power system, decarbonization is additive: you build zero-carbon generation while the existing fleet keeps running. In heavy industry, the emissions are embedded in the chemistry. A blast furnace cannot be made low-carbon by cleaning up the electricity it consumes, because most of its emissions come from reducing iron ore with coke—a reaction that releases CO2 as a product, not as a byproduct of combustion. Roughly two-thirds of cement kiln emissions come from calcium carbonate decomposing, not from the fuel burned to heat it. This is what "hard to abate" means in practice. The phrase identifies a class of emissions that cannot be eliminated by substituting inputs at the margin.

The mitigation options form a short list. The rest of the chapter turns on their relative costs, so it is worth setting them out plainly.

• Efficiency and scrap. Energy-efficiency retrofits, more scrap in electric-arc furnaces, better yield management. These can shave perhaps 10 to 15 percent off sector emissions at low or negative cost. It is the no-regrets margin, and China's most modern plants have largely exhausted it.
• Fuel switching. Replacing coal with natural gas, biomass, or electricity in kilns and furnaces cuts combustion emissions but does nothing about process emissions. Real, but bounded.
• Process substitution. Replacing the blast furnace–basic oxygen furnace route with hydrogen-based direct reduced iron (DRI) plus electric-arc furnace, or replacing limestone calcination with alternative clinker chemistry. This addresses process emissions directly. It is the only route to deep decarbonization, and it is capital-intensive in a way the other options are not.
• Carbon capture, utilization, and storage (CCUS). Capturing CO2 from flue gas or process streams and sequestering it. Technically mature in several applications, expensive, and dependent on transport and storage infrastructure China is only beginning to build.
• Demand reduction and material efficiency. Less steel and cement per unit of output through better design, longer product life, substitution with timber. Engineers pay this option the least attention; economists pay it the most. It is also the option that most directly threatens incumbents.

The costs are steep. The International Energy Agency's 2020 Iron and Steel Technology Roadmap estimated that near-zero-emission steelmaking would cost 10 to 50 percent more than the conventional route, depending on energy prices and hydrogen availability. For cement, the IEA's Technology Roadmap put the additional cost of near-zero clinker at 40 to 100 percent. The ranges are wide because they depend on the price of green hydrogen, the price of electricity, and the cost of capital—all policy-dependent. A planner committing capital in 2024 to a plant that will operate until 2050 is not solving an engineering problem. She is betting on a set of prices that do not yet exist.

6.1.2 Why the capital stock matters more than the flow

China's heavy-industry capital stock was built overwhelmingly between 2000 and 2015. A blast furnace runs fifteen to twenty years between relines; a cement kiln, twenty to thirty; an aluminum potline's pots last four to six years, but the shell and infrastructure twenty-plus. A large share of the sector's emissions-producing assets will therefore come up for replacement, relining, or major overhaul in the 2030s. The decision made at each of those moments—reinvest in the incumbent technology or switch to a low-carbon process—sets the sector's emissions trajectory for the following two decades.

This is what I mean by capital stock turnover rather than operating efficiency. A plant's emissions are largely fixed at the moment of investment. A firm that relines a blast furnace in 2028 has locked in conventional steelmaking until the mid-2040s. A firm that builds a DRI plant instead has locked in a different cost structure, with different carbon exposure and different input requirements. The transition is lumpy and path-dependent. It does not happen gradually as firms optimize; it happens in discrete steps as assets are replaced, and the timing of those steps is set by the age profile of the existing stock.

China's age profile is unusually young. Because so much capacity went up in the 2000s, a large fraction of the stock is not yet due for replacement. The window for influencing the replacement decision is still open—but the same youth is why the sector's emissions are locked in for longer than a purely economic calculation would suggest. A mature industrial economy with an older capital stock would face more replacement decisions sooner and could, in principle, decarbonize faster. China's young stock is a competitive advantage in the short run and a source of lock-in in the long run. The same assets that made Chinese steel cheap in 2010 will make Chinese steel expensive to decarbonize in 2040.

The age profile is not uniform, and the aggregate conceals real heterogeneity. Hebei's integrated mills are older on average than those in the southern coastal provinces, and the scrap-based electric-arc capacity in Jiangsu and Guangdong runs on an entirely different replacement cycle. A firm-level analysis would show a wider dispersion of decision windows than the sectoral average implies. The point stands as a central tendency, not a universal rule.

6.1.3 The policy context: from output caps to carbon

The policy environment has shifted in stages, and the stages matter because each one acted on a different variable.

Supply-side reform and air-quality regulation dominated from roughly 2016 to 2020. The "2+26" cities plan, winter heating-season production restrictions, and the campaign against "backward" capacity all cut output and, incidentally, emissions. These instruments were blunt and administratively costly, but they raised the cost of operating marginal capacity and accelerated the retirement of the least efficient plants. The Handan furnace was a casualty of this stage.

In 2021 the "dual carbon" goals entered sectoral planning. The Ministry of Industry and Information Technology issued peak-carbon guidance for steel, cement, and aluminum that year. The national emissions trading scheme, initially limited to power, began extending toward industry. Benchmark standards for energy consumption per unit of output arrived with the threat that plants failing to meet them would close. This stage operates through expectations rather than immediate prices, which makes it harder to measure. A plant manager who believes carbon will be priced in 2030 behaves differently in 2024 than one who does not, even at a 2024 carbon price of zero.

Explicit carbon pricing in heavy industry is only beginning. The national ETS completed its first power-sector compliance cycle in 2021 and has been expanding slowly toward the industrial sectors in its original design. The trajectory is uncertain—Chapter 7 takes up the design questions—but the direction is clear enough that firms have started building carbon prices into capital budgeting. In my 2022 and 2023 interviews, planners used internal carbon prices ranging from 50 to 200 yuan per tonne. That range reflects genuine disagreement about policy, not differences in analytical method.

The three stages interact in a distinctive pattern. Output caps cut emissions in the short run without changing technology. Energy benchmarks push the least efficient plants out but leave the best untouched. Carbon pricing, once it carries enough weight, changes the relative economics of technologies and therefore the replacement decision. The first two stages have done most of the work so far. The third is where stranded-asset risk crystallizes, and it is where the rest of this chapter lives.

6.1.4 The cluster dimension

In the labor-intensive sectors examined in Chapter 4, the cluster was the unit of analysis because the relevant capabilities—design, sourcing, quick-response production—were distributed across firms and institutions in a geographic space. Heavy-industry clusters still matter, but for a different reason. They are organized around shared infrastructure: coke ovens, power plants, rail links, ports, and increasingly hydrogen pipelines and CO2 transport. The economics of process substitution depend heavily on whether that shared infrastructure can be repurposed.

A steel cluster built around a coking plant and a coal-fired captive power station faces a different transition cost than one built around a port with access to imported hydrogen or a geological formation suitable for CO2 storage. The cluster determines the cost of the complementary infrastructure that any process substitution requires. This is why the geography of Chinese heavy industry—concentrated in Hebei, Shanxi, Inner Mongolia, Liaoning, and a handful of other provinces—matters so much. Those provinces have abundant coal, limited renewable resources (with the important exception of Inner Mongolia's wind and solar potential), and in most cases no CO2 storage within economic transport distance. The clusters that grew up around coal are the ones that will find process substitution most expensive.

The general claim is enough for now: the cluster context, which in light industry was a source of collective capability, functions in heavy industry as a determinant of transition cost. The same geography that made these clusters competitive in the coal era makes them expensive to convert.

6.2 Process Substitution and Technological Breakthroughs

6.2.1 The hydrogen route to steel

Hydrogen-based direct reduced iron is the most discussed process substitution in heavy industry. The blast furnace reduces iron ore with coke, producing pig iron and CO2. The DRI route reduces iron ore with a hydrogen-rich gas, producing direct reduced iron and—if the hydrogen is green—water vapor instead of CO2. The reduced iron is then melted in an electric-arc furnace that can run on renewable electricity. Green hydrogen plus electric-arc furnace is the only demonstrated route to near-zero primary steelmaking, and it is what the EU's steel decarbonization projects and a growing number of Chinese pilots are pursuing.

The technology is not new. Direct reduction has run commercially since the 1970s, mostly in gas-rich countries—Iran, India, Mexico—where natural gas was cheap. What is new is the prospect of running it on green hydrogen, and the scale required to matter for Chinese steel. Hybrit in northern Sweden delivered its first steel to Volvo in 2021. It is a demonstration, not a template for Chinese conditions: Sweden has cheap hydroelectricity and high-grade iron ore.

China's pilots cluster in the north and along the coast. Baowu, the largest producer, has announced a hydrogen-based DRI project in Zhanjiang, Guangdong, and a smaller demonstration in Xinjiang. Hebei Iron and Steel runs a pilot in Zhangjiakou on hydrogen from renewable electricity. Scale is modest—tens of thousands of tonnes a year against national output of a billion—and the economics lean heavily on subsidies and cheap green hydrogen. The gap between pilot and commercial plant is where the transition will be won or lost.

The cost arithmetic is unforgiving. Green hydrogen made from renewable electricity at 0.2 yuan per kilowatt-hour, with 70 percent electrolyzer efficiency and capital cost of 4,000 yuan per kilowatt, costs roughly 20 to 25 yuan per kilogram before compression and transport. DRI needs about 50 to 60 kilograms of hydrogen per tonne of steel, so the hydrogen input alone runs 1,000 to 1,500 yuan per tonne—against a conventional hot-metal cost of roughly 2,500 to 3,000 yuan per tonne. Add the electric-arc furnace, the renewable electricity for melting, and the capital cost of the DRI shaft, and the green-steel premium in China in the mid-2020s comes to roughly 1,000 to 2,000 yuan per tonne, or 25 to 50 percent above conventional cost. These are estimates, sensitive to every assumption. A halving of electrolyzer capital cost, plausible by 2030, would cut the premium substantially.

The critical variable is the price of green hydrogen, which depends on the price of renewable electricity, which depends on the grid. China's renewable resources sit in the northwest and north, far from the coastal steel clusters that produce most of the sector's output. Moving hydrogen long distances is expensive—as compressed gas, as liquid (cryogenic temperatures, a large share of the energy content consumed), or as ammonia (synthesis and cracking, each with losses). The alternative is to move steelmaking to the hydrogen rather than hydrogen to the steelmaking, which means relocating a significant share of the industry to Inner Mongolia, Gansu, and Xinjiang. That is a spatial reconfiguration of the sector, not a technological fix, and its political and social consequences run well beyond the engineering.

6.2.2 Electrified cement and alternative clinker

Cement is harder, because two-thirds of its emissions come from limestone chemistry rather than fuel. The options are narrower. Electrifying the kiln—plasma or resistance heating in place of a coal flame—eliminates combustion emissions but leaves process emissions intact. Capturing CO2 from kiln off-gas is technically feasible and is being pursued in a handful of European and Chinese demonstrations, but it needs a transport and storage network that does not yet exist at scale in China. The most radical option changes the clinker chemistry itself: substituting calcined clay for part of the calcium carbonate can cut process emissions by 30 to 40 percent, and entirely new binders that avoid limestone altogether are under development.

The Chinese cement industry has moved slowly on these options for a structural reason. The sector is fragmented—thousands of kilns owned by hundreds of companies, with an average plant small by international standards. A carbon-capture retrofit costs a lot relative to the value of a single kiln, and fragmentation makes shared infrastructure hard to finance. The large producers, Anhui Conch and China National Building Material, have the balance sheets to invest, but they face the same uncertainty about carbon prices and storage as everyone else.

The most promising near-term lever in cement is not a breakthrough technology but a change in product mix. Blended cements that replace part of the clinker with fly ash, slag, or calcined clay have been used for decades and can cut emissions per tonne of cement by 20 to 30 percent at little or no cost. The barrier is regulatory and commercial: building codes and standards often specify minimum clinker contents, and contractors are conservative about unfamiliar mixes. China's standards have been revised in recent years to permit higher substitution rates, and the effect on emissions has been measurable. Unglamorous, but one of the few levers that can be pulled at scale in the 2020s.

6.2.3 Aluminum and the electricity question

Primary aluminum is, in emissions terms, a child of the grid. The smelting process itself—electrolytic reduction of alumina in a cryolite bath—releases CO2 from carbon anode consumption, roughly 1.5 to 2.0 tonnes per tonne of aluminum. But the dominant source in China is the coal-fired electricity running the potlines. Chinese aluminum's emissions intensity is roughly three times the global average for this reason, and the single most effective decarbonization lever is shifting smelting to renewable electricity.

That shift is already underway, driven less by climate policy than by renewable power economics. Chinese aluminum capacity has been migrating from the coal provinces of the center and east to the renewable-rich northwest—Yunnan with its hydropower, Inner Mongolia and Xinjiang with wind and solar. Average emissions intensity has fallen meaningfully over the past decade, and a new industrial geography has emerged, more exposed to hydrological risk (Yunnan's dry seasons have periodically forced smelters to curtail) and to transmission constraints on the northwestern grid.

The remaining problem is the anode. Inert anodes, which do not consume carbon and therefore eliminate process emissions, have been pursued for decades and are now in early commercial demonstration. If the technology matures at scale and cost, aluminum becomes one of the few heavy industries that can approach zero emissions with a single technology change. If it does not, anode chemistry sets a floor under the sector's emissions for the foreseeable future.

6.2.4 The common structure of the substitution problem

The three cases differ in engineering but share an economic structure, and making that structure explicit matters because it governs the stranded-asset analysis that follows.

Each substitution requires a large, irreversible capital commitment. Each depends on an input—green hydrogen, renewable electricity, CO2 storage—whose price is uncertain and policy-dependent. Each produces a commodity with a small or nonexistent green premium. And each faces an incumbent technology that is cheap, well-understood, and still improving at the margin. The firm is not choosing between a profitable new technology and an unprofitable old one. It is choosing between a known cost with known carbon exposure and an unknown cost with unknown carbon exposure, under a policy regime that could move either way.

That is a real-options problem, and the rational response is to wait—keep the option alive with pilots and engineering studies while deferring the large irreversible commitment until uncertainty resolves. Which is, in fact, what most Chinese heavy-industry firms are doing. The pilots are real; the commercial commitments are not. The policy question is how to make waiting more expensive than committing. The instruments are well understood in principle and have been deployed in Europe: carbon contracts for difference, guaranteed offtake, subsidized finance, public investment in shared infrastructure. Their application in China is complicated by the fiscal capacity of provincial governments, the fragmentation of the industry, and the absence of a carbon price high enough to make the green premium visible in the first place.

The wait-and-see characterization is a stylization, and it fits some firms better than others. The largest state-owned producers have the balance sheets and political cover to commit early, and several have. Private producers, especially in fragmented cement and aluminum, have neither, and their waiting is less a strategic choice than a constraint. The distinction matters for policy: a carbon contract that works for Baowu may not work for a privately owned cement kiln in Henan.

6.2.5 What counts as a breakthrough

"Breakthrough" gets used loosely in this literature. For the purposes of industrial transition, a technology is a breakthrough when it meets three conditions: it can be deployed at the scale of the sector, it can be financed without permanent subsidy, and it can be operated by the existing workforce with retraining rather than replacement. The pilots described above meet the first condition only at demonstration scale and the second not at all. The third is rarely discussed and may prove decisive. A hydrogen DRI plant needs a different skill mix than a blast furnace, and retraining a workforce that has spent its career tending coke ovens takes years.

None of this argues against pursuing the technologies. It argues against the assumption, common in policy documents, that a technological breakthrough will arrive and solve the problem on its own schedule. A decade of pilot projects suggests the technologies will improve incrementally, costs will fall along learning curves shallower than advocates expect, and the binding constraint will be financing and coordinating deployment rather than inventing the process. The breakthrough, when it comes, will be organizational as much as technical.

6.3 Stranded Asset Risk and Financial Implications

6.3.1 Defining the risk

A stranded asset, in the climate-finance literature, is one that has suffered an unanticipated or premature write-down, devaluation, or conversion to a liability as a result of the transition to a low-carbon economy. The Carbon Tracker Initiative popularized the concept in 2011; it has since been applied to fossil-fuel reserves, coal plants, and, more recently, heavy-industry capital. The definition is deliberately broad, and the breadth is a weakness: almost any asset can be called stranded if the counterfactual is loose enough. I want a narrower definition, one that is analytically useful for the firms in this chapter.

For our purposes, an asset is stranded when its expected cash flows under a plausible carbon-price path are insufficient to recover its remaining book value over its operating life. This is a financial definition, not a moral one, and it is measurable in principle. It also makes clear that stranding is a matter of degree and timing rather than a binary state. A blast furnace relined in 2025 is not stranded in 2025. It becomes stranded if the carbon price in 2035 is high enough that its operating cost exceeds the cost of a replacement technology, and if the firm cannot recover its investment before that point.

The measurement problem is severe. The carbon-price path is uncertain, the cost of the replacement technology is uncertain, and the asset's operating life depends on maintenance decisions not yet made. Any estimate of stranded-asset exposure is a scenario, not a forecast. Published scenarios for Chinese heavy industry—from the Bank for International Settlements, from academic groups, from the China Council for International Cooperation on Environment and Development—differ widely in assumptions and therefore in conclusions. I will present the range rather than a point estimate, and be explicit about which assumptions drive the results.

6.3.2 The exposure by sector

Stranded-asset exposure in Chinese heavy industry concentrates in a few asset classes.

Integrated steelworks. The blast furnace–basic oxygen furnace route accounts for roughly 90 percent of Chinese steel output. The assets at risk are the blast furnaces, coke ovens, sinter plants, and the captive coal-fired power stations supporting them. A blast furnace relined after 2025 has an expected operating life extending into the 2040s, well beyond the point at which a meaningful carbon price is likely to be in force. Exposure is largest for furnaces in the interior provinces, where switching to hydrogen or scrap costs the most, and smallest for coastal furnaces with access to imported scrap and potential hydrogen imports.

Cement kilns. China has roughly 1,500 clinker kilns, many small and old. The assets at risk are the kilns and the quarries supplying them. The exposure is complicated by the fact that many kilns are already near the end of their economic life, so stranding is partly natural retirement rather than policy-driven write-down. The marginal kilns will close anyway. The question is whether carbon policy accelerates the closure and whether owners are compensated.

Aluminum smelters. Potlines are long-lived, but the dominant emissions source is the power supply, which is contracted rather than owned in most cases. The asset at risk is less the smelter than the captive coal plants and long-term power purchase agreements underpinning it. A smelter that signed a twenty-year fixed-price coal power contract is exposed to a carbon price its supplier must pass through.

Petrochemicals and chemicals. Ammonia, methanol, and ethylene are large emitters with long-lived assets. Ammonia is the most exposed: its process emissions are concentrated, and green ammonia is a credible substitute. The Chinese ammonia fleet is relatively young and coal-based, which makes the stranding risk significant.

Aggregate exposure is hard to state with confidence. A 2021 estimate by the China Council for International Cooperation on Environment and Development put potential stranded-asset value in the power and industry sectors at several trillion yuan under a stringent decarbonization scenario, but the number is sensitive to the discount rate and the carbon-price path. A more useful framing is the share of the capital stock likely to be affected: perhaps 20 to 30 percent of the heavy-industry capital stock built before 2015 is at risk of premature retirement or write-down under a path consistent with the 2060 neutrality target. That is a large number, and it is why the financial implications cannot be a footnote.

6.3.3 Who bears the loss

The distribution of stranded-asset losses is a political question as much as an economic one.

The firms. The first bearer is the asset owner. A write-down reduces equity, and for a leveraged firm it can push the balance sheet into distress. Chinese heavy industry is highly leveraged—the average debt-to-asset ratio of listed steel companies has hovered around 60 percent—so a modest write-down can wipe out a large share of equity. State-owned enterprises have implicit government support to absorb the loss. Private firms do not, and while their share of the sector is smaller than in light industry, it is not negligible.

The banks. The second bearer is the banking system holding the debt that financed the assets. Chinese heavy industry is financed overwhelmingly by bank loans, and the largest lenders are the state-owned commercial banks. A wave of stranded-asset write-downs would surface as non-performing loans, absorbable in aggregate but not without cost. The regional banks in Hebei, Shanxi, and Liaoning are more exposed than the national champions, and their capital buffers are thinner.

The local governments. The third bearer is the local government, which depends on the sector for tax revenue and employment. A plant closure cuts both, and the fiscal and social costs land on the locality. This is why local governments have historically resisted closures and often subsidized marginal plants to keep them running. The stranded-asset problem is therefore also a fiscal-federalism problem: the costs of the transition concentrate in jurisdictions with the least capacity to absorb them.

The investors. The fourth bearer is the equity investor, domestic and foreign, holding shares in affected companies. The A-share market has begun to price carbon risk, but the pricing is uneven and disclosure is poor. A 2022 study by the China Financial Information Center found that fewer than half of listed companies in heavy-industry sectors disclosed their Scope 1 emissions, and the disclosures that exist are not comparable across firms. An investor trying to price stranded-asset risk faces a data problem that is resolving only slowly.

The distribution determines the politics. A transition whose costs fall on private firms and their bankers is one kind of political problem; a transition whose costs fall on provincial governments and their employees is another. The Chinese case is a mixture, and the mixture shifts as the private share of the sector changes and provincial fiscal capacity diverges.

6.3.4 The financial system's response

The financial system has begun to respond, though the response is uneven and, in my reading of the evidence, still more about disclosure than about pricing.

The People's Bank of China's green finance framework, introduced in 2021, includes transition finance as a category, and the major banks have begun offering transition-linked loans to heavy-industry borrowers. The terms typically tie the interest rate to the borrower's emissions intensity or to specific reduction targets. Volume is small relative to the sector's financing needs—the green loan balance was roughly 20 trillion yuan at the end of 2022, but the share going to heavy-industry transition is a small fraction—and the pricing differential is modest.

The bond market has been more active. Sustainability-linked bonds, which tie the coupon to the issuer's sustainability targets, have been issued by several steel and cement companies. The structure is promising in principle because it aligns the cost of capital with the transition, but the targets are often weak and the penalties for missing them small. A bond that pays 10 basis points more if the issuer misses a modest emissions target is not a meaningful discipline.

The most consequential financial development may be the gradual introduction of carbon into credit rating. The major Chinese rating agencies have begun incorporating carbon risk, and the state-owned banks have begun stress-testing heavy-industry exposures. The stress tests are not public and their results are not disclosed, which makes it hard to judge how seriously the banks take the risk. The direction is clear, though, and it feeds back on the firms: a borrower whose credit rating is downgraded for carbon exposure faces a higher cost of capital, which makes the transition investment more expensive, which makes the asset more likely to strand. The financial system is not a neutral observer of the transition. It is one of its transmission mechanisms.

6.3.5 The real-options problem revisited

The real-options framing connects the stranded-asset analysis to the process-substitution discussion, and it carries a policy implication that is often missed.

A firm facing an irreversible investment under uncertainty has an incentive to wait. The value of waiting is the option premium—the value of the information that arrives if the firm defers. In heavy industry, that information is the carbon price, the cost of green hydrogen, and the policy support available. The option premium is large, which is why firms wait.

Stranded-asset risk changes the calculus in a subtle way. If the firm waits, it keeps the option alive, but it also keeps the existing asset running, and the existing asset is accumulating carbon exposure. The longer the firm waits, the more the existing asset is worth in a high-carbon world and the less in a low-carbon one. Waiting is not neutral. It is a bet that the low-carbon world arrives later rather than sooner. A firm that waits and is wrong has not only missed the transition; it has invested in the wrong asset.

This is the sense in which stranded-asset risk is the mechanism of green governance. The risk does not force the firm to invest in green technology; it forces the firm to decide, because the cost of not deciding rises as the policy regime tightens. Firms that have built the capability to decide—engineering teams, financing, political relationships—will make the transition. Firms that have not will wait until waiting is no longer possible, and then be acquired or closed.

The capability to decide is what I have elsewhere called green governance capability, the central concept of this book. In heavy industry it has three components: the technical capability to evaluate and execute a process substitution, the financial capability to fund it, and the organizational capability to manage the workforce and community consequences. They are not independent—a firm without technical capability cannot use financial capability effectively—and they are not evenly distributed. The largest state-owned enterprises have all three. Private firms have the first, sometimes, and rarely the second and third. The gap between them is the space in which the transition will happen, and the firms that close the gap will emerge as the sector's leaders.

6.3.6 A note on the limits of the analysis

I have been describing a transition that has not yet happened, using scenarios and estimates rather than observed outcomes. The numbers in this section illustrate a mechanism; they are not forecasts. The mechanism is robust: capital stock turnover plus carbon pricing plus process substitution equals stranded-asset risk, distributed unevenly across firms, sectors, and regions. The magnitudes are not robust. A carbon price rising faster than expected increases the risk; a breakthrough lowering the cost of green hydrogen decreases it; a policy decision to compensate asset owners redistributes it. None of these can be predicted with confidence, and the honest position is to describe the range and identify the variables that matter.

Three variables matter most in my reading. The carbon-price path in the 2030s determines when existing assets become uneconomic. The cost of green hydrogen determines when the replacement technology becomes competitive. The availability of transition finance determines which firms can afford to switch. The first two are policy and technology variables; the third is a financial-system variable. All three are uncertain, and all three are subject to decisions being made now.

6.4 The Cluster Reconfiguration

6.4.1 From coal clusters to hydrogen clusters

The cluster framework developed in my earlier work and applied to light industry in Chapter 4 needs modification for heavy industry, and the modification is spatial. In light industry, the cluster was a fixed geography within which capabilities circulated. In heavy industry, the cluster itself is being reconfigured by the transition, and the reconfiguration is the object of analysis.

The direction is visible in five years of investment decisions. Aluminum capacity has moved toward Yunnan and the northwest. Steel capacity has been capped in Hebei and pushed to consolidate in coastal provinces with port access. Cement capacity has been rationalized through eliminating small kilns and concentrating production in larger, more efficient plants. Each movement responds to a different set of prices—electricity for aluminum, environmental capacity for steel, scale economies for cement—but they share a feature: production is moving away from the coal fields that anchored the original clusters and toward locations with different resource endowments.

A hydrogen economy would accelerate the shift. Green hydrogen will be produced where renewable electricity is cheapest, which in China means the northwest and, to a lesser extent, coastal provinces with offshore wind potential. A steel industry running on hydrogen will be located differently from one running on coal. The coastal clusters—Tangshan, Rizhao, Zhanjiang—have an advantage because they can import hydrogen or reduced iron. The interior clusters do not. The likely result is further concentration in a smaller number of locations, each combining port access, renewable resources, and infrastructure for hydrogen or CO2 transport.

This is a different cluster dynamic from Chapter 4. There, the cluster upgraded in place, adding design and branding capabilities to an existing manufacturing base. Here, the cluster is being selected for by the transition, and the selection is driven by geography rather than capability. A cluster sitting on a good hydrogen site will thrive; one that does not will decline, regardless of how capable its firms are. The implication for regional policy is uncomfortable: the provinces with the most to lose from the transition have the least capacity to attract the new industries.

6.4.2 The infrastructure bottleneck

The reconfiguration depends on infrastructure that does not yet exist at scale, and the timing of infrastructure investment is itself a determinant of stranded-asset risk. A hydrogen pipeline built in 2030 makes a DRI plant viable in 2032; a pipeline delayed to 2035 makes the same plant uneconomic for three more years. The same holds for CO2 transport and storage, for transmission lines carrying renewable electricity to coastal clusters, and for port facilities to import green ammonia or reduced iron.

China has begun to plan. The National Development and Reform Commission's 2022 hydrogen strategy sets production targets and names the northwest as a priority region. The state grid has announced ultra-high-voltage transmission projects connecting northwestern renewables to coastal load centers. Several provinces have commissioned feasibility studies for CO2 storage in depleted oil and gas fields and saline aquifers. The plans are ambitious, and the gap between plan and execution is the usual one in Chinese infrastructure: financing not fully committed, permitting slow, cross-provincial coordination imperfect.

The bottleneck has a distributional consequence worth noting. Firms that can afford to wait for infrastructure are the large state-owned enterprises, which have the balance sheets and political access to position themselves for the post-infrastructure world. Firms that cannot are the smaller private producers, which will exit or be acquired. Infrastructure is not a neutral public good here; it is a mechanism that selects among firms, and the selection favors incumbents with the deepest pockets.

6.4.3 The labor question

Labor bears directly on the feasibility of the transition and the politics of stranded assets, and I have said too little about it.

Chinese heavy industry employs roughly ten million people directly, concentrated in a few provinces and a few large enterprises. The workforce is older than in light manufacturing, more unionized in the formal sense that the All-China Federation of Trade Unions has a presence, and more embedded in the communities where the plants sit. A blast furnace closure does not just strand an asset; it strands a workforce whose skills are specific to the technology being retired.

The retraining challenge is real but manageable in aggregate. A hydrogen DRI plant needs operators, technicians, and engineers, and many skills transfer from the conventional route. The problem is geography and timing. New plants will be in different locations from old ones, and workers who lose their jobs may not be able to move. The hukou system, which ties housing, education, and social services to place of registration, makes relocation costly for workers with families. The likely result is localized unemployment in the coal provinces even as the sector expands in coastal and northwestern regions.

The social cost of this displacement does not appear instranded-asset calculations, which count the value of capital and not the value of livelihoods. A fuller accounting would include the fiscal cost of unemployment support, the social cost of community decline, and the political cost of resistance to further closures. These costs are real, and they are why the transition will likely be slower and more contested than the techno-economic analysis suggests.

6.5 The International Dimension

6.5.1 Carbon border adjustment and the export exposure

Heavy-industry products are traded internationally, and that trade is exposed to the carbon policies of China's partners. The European Union's Carbon Border Adjustment Mechanism (CBAM) is the most concrete case. It entered its transitional phase in October 2023 and will impose a carbon levy on imports of steel, aluminum, cement, and fertilizers from 2026. The mechanism is designed to prevent carbon leakage—the relocation of emissions-intensive production to jurisdictions with weaker climate policy—and its effect on Chinese exporters is to make the carbon content of their products a cost of doing business in Europe.

The exposure is significant but uneven. Chinese steel exports to the EU are modest relative to total output—roughly 5 to 10 percent of exports, which are themselves a small share of production—but the CBAM's effect is not confined to the covered products. It sets a precedent, and other jurisdictions are considering similar mechanisms. The United States has proposed a carbon border fee, and the UK has consulted on one. The direction of travel is toward a world in which carbon content is priced at the border, and Chinese heavy industry will have to adapt.

Adaptation can take two forms. One is to cut the carbon content of exports, which means investing in the process substitutions described above. The other is to redirect exports toward markets without border adjustment, which means accepting a lower price and a smaller market. The choice depends on the cost of decarbonization relative to the price discount in non-adjusting markets, and on how fast the border mechanisms spread. Firms that have already invested in low-carbon capacity—Baowu's hydrogen pilot, for instance—are positioning for the first option. Firms that have not will find their export markets shrinking.

The magnitude deserves care. The CBAM's initial coverage is narrow, and the administrative burden of compliance may matter as much as the levy itself. Its effect on Chinese heavy industry in the 2020s is likely modest. The effect in the 2030s, if the mechanism expands and other jurisdictions follow, could be substantial. The relevant question for firms is not what the CBAM costs today but what it signals about the world in which their assets will operate over the next two decades.

6.5.2 The technology competition

The transition has also become a site of international competition over technology and standards. The EU, Japan, and South Korea have all launched hydrogen and green-steel programs, and the firms that establish dominant designs will capture the rents. China's position is strong in some segments—electrolyzer manufacturing, battery supply chains, solar—and weaker in others, particularly the engineering and integration of DRI plants and the materials for high-temperature processes.

In the short run the competition is not zero-sum, because the global market for green-steel technology is large enough to support multiple suppliers. It becomes more zero-sum as the technology matures and standards consolidate. The country that sets the definition of "green steel" will shape the terms of trade for a generation, and the contest is already visible in competing definitions issued by the EU, the International Energy Agency, and industry consortia. China's participation is complicated by a domestic carbon accounting system that is not yet interoperable with the EU's, and by political constraints on accepting external verification of emissions data.

6.6 What the Evidence Suggests

The furnace in Handan is cold now. The site has been redeveloped, and the company that owned it has announced a hydrogen pilot in another province. Whether the pilot becomes a plant, and the plant a sector, depends on decisions being made in the next few years by firms, banks, provincial governments, and the ministries in Beijing that set the carbon price and allocate transition finance. Three findings from the evidence assembled here bear on those decisions.

The first is that the emissions trajectory of Chinese heavy industry will be set by the replacement decisions of the 2030s, not by the operating efficiency of the 2020s. A plant's emissions are largely fixed at the moment of investment, and the sector's young capital stock means a large wave of relining and replacement is coming. The corollary is that the transition will be lumpy and path-dependent, and that the timing of policy matters enormously. A carbon price that arrives before the replacement wave shapes the decisions; one that arrives after merely raises the cost of assets already locked in.

Process substitution is technically feasible but economically contingent. The hydrogen route to steel, the electrified and captured routes for cement, and the inert-anode route for aluminum are all demonstrated or demonstrable. Deployment depends on the price of green hydrogen, the availability of renewable electricity and CO2 storage, and the financing of large irreversible investments. The rational firm faces a real-options problem, and the rational response is to wait—which is what most firms are doing. The policy challenge is to change the calculus so that waiting becomes more expensive than committing.

And the stranded-asset risk is not a side effect of the transition; it is the mechanism through which the transition reshapes the industry. The risk forces firms to decide, and firms with the capability to decide—technical, financial, organizational—will make the transition. Firms without it will exit or be acquired. How the losses are distributed across firms, banks, local governments, and investors is a political question, and the answer will set the speed and the equity of the transition.

These conclusions rest on assumptions that could be wrong. If green hydrogen costs fall faster than expected, the transition will be cheaper and faster. If carbon prices rise more slowly, stranded-asset risk will be smaller and the transition slower. If the international trading environment fragments, export exposure will matter less. I have tried to flag these contingencies where they arise, and the quantitative claims should not be held with more confidence than the ranges imply.

What I hold with more confidence is the qualitative claim that connects this chapter to the book's larger argument. The green governance capability described in the light-industry cases—the ability to anticipate regulation, build organizational routines for compliance, and convert compliance into competitive advantage—takes a different form in heavy industry. It is not the incremental improvement of existing processes. It is the capacity to make and execute large, irreversible decisions under uncertainty. That capacity is rarer than the incremental kind, and it is concentrated in fewer firms. The transition of Chinese heavy industry will therefore be more concentrated than the transition of its light industry, and the concentration will reshape the sector's structure in ways only beginning to be visible.

The arithmetic of those decisions is the subject of the next chapter, which turns from the industrial firm to the carbon market that will, increasingly, set the price against which every heavy-industry asset is valued.

Note on sources: Cost estimates for hydrogen-based steelmaking draw on the IEA's Iron and Steel Technology Roadmap (2020) and on company disclosures from Baowu, Hebei Iron and Steel, and Hybrit. Cement cost ranges are from the IEA Technology Roadmap: Low-Carbon Transition in the Cement Industry (2018) and from the Getting to Zero Coalition's sectoral analyses. Emissions factors are from the National Bureau of Statistics and from the IPCC's 2006 Guidelines. Stranded-asset estimates draw on the China Council for International Cooperation on Environment and Development's 2021 report and on the BIS's 2021 climate-risk analysis. Interview material from Hebei, Shanxi, and Inner Mongolia (2021–2023) was collected by the author and is used here as illustration rather than as systematic evidence. CBAM details are from EU Regulation 2023/956. Any errors of interpretation are the author's.