Chapter 6: Technology-Intensive Industries: Innovation-Driven Green Competitiveness and the Learning Curve

Liang's carbon declaration won him a hearing, not a contract. Compliance bought him a seat at the table; it did not yet buy him pricing power. That gap between being allowed to compete and being able to win is where technology-intensive industries enter the story. For a furniture maker in Shunde, the green agenda arrives as an external cost to be absorbed and passed on. For a photovoltaic manufacturer in Xi'an, a battery cell producer in Ningde, or an industrial automation firm in Shenzhen, the same agenda arrives as a demand shock in the most literal sense: it creates markets that did not exist a decade ago and rewrites the cost curves of markets that did. If green governance capability is an endogenous driver of upgrading rather than a tax on it, the mechanism should be most visible where innovation is already the competitive weapon—and least visible where cost is.

The claim I want to defend is narrower than "technology-intensive industries are green winners." Several of them are not, and the losers are instructive. The defensible claim is that technology-intensive sectors exhibit a distinctive learning-curve dynamic in which environmental performance and productive performance are jointly produced, so that the marginal cost of abatement falls with cumulative output rather than rising. Where that dynamic takes hold, compliance and competitiveness stop trading off. Where it fails—where the learning curve is captured by incumbents, or where the relevant learning is in system integration rather than in components—the green transition reproduces the compliance trap described in Chapter 5, only at a higher level of capital intensity.

6.1 Innovation-Driven Green Competitiveness

6.1.1 From Externality to Strategic Resource

Neoclassical trade theory treats environmental regulation as a cost shock. A binding emissions constraint raises marginal cost, shifts the supply curve upward, and—absent offsetting productivity gains—reduces output and exports in the regulated jurisdiction. This is the pollution-haven logic in its purest form, and it predicts that technology-intensive industries, being more emissions-intensive per unit of value in many cases (semiconductors, chemicals, non-ferrous metals), should be among the most exposed. The empirical record is messier. Dechezleprêtre and Sato (2017), reviewing the competitiveness effects of environmental regulation across 21 studies, find small average effects on net trade and employment but substantial heterogeneity: effects are near zero or positive for firms close to the technological frontier and negative for laggards. That heterogeneity is the phenomenon this chapter is about.

The theoretical bridge from externality to strategic resource was built in stages. Porter and van der Linde (1995) argued that well-designed regulation can trigger innovation that more than offsets compliance cost, because firms systematically underinvest in efficiency due to organizational inertia and incomplete information. The Porter Hypothesis in its weak form—regulation stimulates innovation—is now well supported; the strong form—the innovation offsets the cost—remains contested and appears to hold only under specific design conditions (Ambec et al., 2013). What matters here is the mechanism the Porter literature identifies: regulation forces firms to search a solution space they would otherwise have ignored. Van Weele and Vivanco (2014) make the same point from the supply-chain side. When a buyer imposes environmental requirements on a supplier, the supplier's product and process solution space contracts, and fitting into that smaller space requires new solutions. Constraint, up to a point, is generative.

The resource-based view supplies the complementary half. Wernerfelt (1984) and later Barney (1991) argued that sustained competitive advantage rests on resources that are valuable, rare, imperfectly imitable, and non-substitutable. Green capabilities fit this template awkwardly but genuinely. A firm's accumulated knowledge of how to reduce emissions per unit of output—embedded in process engineers, supplier relationships, and proprietary tooling—is valuable (it lowers cost and unlocks regulated markets), rare (most competitors have not built it), and imperfectly imitable (it is largely tacit and path-dependent). Hart (1995) formalized this as the natural-resource-based view, distinguishing pollution prevention, product stewardship, and sustainable development as three progressively deeper strategic postures. The progression is not automatic. Firms move from pollution prevention (end-of-pipe, cost-reducing) to product stewardship (design for environment, differentiation) to sustainable development (system-level, market-creating) only when the competitive logic pushes them there.

Eco-innovation research operationalizes this. Díaz-García, González-Moreno, and Sáez-Martínez (2015), reviewing the literature, distinguish environmental innovation by its dual externality structure: it produces a knowledge spillover like any innovation and an environmental spillover that is not priced. That double externality justifies public policy beyond the usual R&D subsidy, but it also predicts that private eco-innovation will be underprovided relative to its social return—which is why the observed rate of eco-innovation depends heavily on the stringency and predictability of the regulatory signal. Rennings (1998) made this argument early, contrasting neoclassical and evolutionary perspectives: the neoclassical firm responds to relative prices, the evolutionary firm responds to selection pressure and accumulates routines. The evolutionary framing is the one that generates a learning curve, because routines improve with repetition.

Chassagnon and Haned (2015), using French firm-level data from the Community Innovation Survey, find that innovation leadership—measured by the introduction of product and process innovations—is associated with environmental benefits, but the relationship is mediated by firm size and sector. Small innovators capture less environmental benefit per innovation than large ones, plausibly because they lack the complementary assets (manufacturing scale, distribution, service networks) needed to diffuse the innovation. The finding is an early warning about the learning curve: it is not a public good that all firms ride equally. It accrues disproportionately to firms that already control complementary assets. The policy implication, which Chapter 9 develops, is that the learning curve can be an instrument of concentration as easily as an instrument of upgrading.

6.1.2 The Dual-Externality Problem and Why It Bites Harder in Technology-Intensive Sectors

Technology-intensive industries face the dual-externality problem in an acute form, for three reasons. First, their innovation cycles are short, so the appropriability period during which a firm can capture the returns to its green R&D is compressed. A battery chemistry improvement may be superseded in eighteen months; the firm that spent the R&D must recover its investment within that window or rely on process know-how that is harder to codify and therefore harder to protect. Second, their products are often intermediate inputs, so the environmental attribute of the product is valued by the downstream buyer rather than the final consumer, and the buyer's willingness to pay for that attribute depends on the buyer's own regulatory exposure. A cathode material with a certified low-carbon footprint commands a premium only if the cell maker faces a carbon border adjustment or a customer with a scope-3 target. Third, their production processes are frequently electricity-intensive, which means the carbon intensity of the local grid enters the product's embodied carbon and is outside the firm's control in the short run.

These three features explain why the green competitiveness of technology-intensive industries is so unevenly distributed across geographies and segments. A Chinese photovoltaic manufacturer selling into the European market faces the EU's carbon border adjustment mechanism (CBAM) only indirectly at first—solar panels are not in the initial CBAM scope—but faces the EU's Ecodesign and energy-labeling requirements and, increasingly, the scope-3 targets of European utilities that buy its modules. A Chinese battery cell maker selling to European automakers faces the EU Battery Regulation's carbon footprint declaration requirements directly. A Chinese industrial automation firm selling to domestic cement plants faces the national emissions trading scheme indirectly through its customers' willingness to pay for energy-saving retrofits. The same technological capability meets very different demand signals depending on the market segment, and the learning curve is ridden faster where the signal is strong and persistent.

I want to be careful here about a common overstatement. It is tempting to say that technology-intensive firms are "naturally" green because they are knowledge-intensive. The evidence does not support this. The semiconductor industry is among the most water- and chemical-intensive manufacturing sectors on earth; a leading-edge fab consumes tens of thousands of cubic meters of ultrapure water per day and produces process emissions of perfluorocarbons with global warming potentials in the thousands. The photovoltaic industry's own manufacturing footprint—polysilicon purification, wafer slicing, cell processing—is substantial, and the industry's green credentials rest on the lifetime energy payback of its products rather than on the cleanliness of its factories. What distinguishes technology-intensive firms is not a lower environmental burden but a higher capacity to innovate against that burden when the incentive is present. Capability, not virtue.

6.1.3 The Antithesis Restated: Comply versus Climb

Chapter 1 introduced the antithesis between complying and climbing. In technology-intensive industries the antithesis takes a specific form. Complying means meeting the environmental requirements of the markets you already serve, at the cost of margin or market share. Climbing means using the environmental requirement as a wedge to move into higher-value segments, new product categories, or new geographies where the requirement is a barrier to entry for your competitors. The two strategies are not distinguished by the amount of green investment—both may require substantial investment—but by whether the investment is directed at defending an existing position or at creating a new one.

Consider the difference in the photovoltaic industry between a cell maker that upgrades its production line to meet the EU's carbon footprint threshold for participation in European tenders, and a cell maker that develops a proprietary low-carbon polysilicon process and licenses it to competitors while capturing the premium on its own modules. The first is complying; the second is climbing. The first firm's investment is defensive and its returns are capped by the market price of cells. The second firm's investment creates an asset—the process—that can be monetized in multiple ways. The learning curve operates in both cases, but in the first it lowers the firm's cost of compliance, while in the second it lowers the industry's cost of compliance and the firm captures the difference as rent.

The conditions under which firms can choose the climbing strategy are the subject of the next section. Here I want to note a structural point. The climbing strategy requires that the firm's green capability be at least partly non-imitable. If the capability is a piece of equipment that can be purchased by anyone, the learning curve is embodied in the equipment supplier and the adopting firm captures no durable advantage. If the capability is a process recipe, a supplier network, or a system-integration skill, it is stickier. This is why the most durable green competitive advantages in technology-intensive industries tend to reside in process and system integration rather than in discrete product features—and why the firms that climb are often not the ones that invent the headline technology but the ones that learn fastest how to manufacture it at scale with low emissions.

6.2 The Learning Curve and Capability Building

6.2.1 The Learning Curve as an Economic Mechanism

The learning curve—the observation that unit cost declines by a constant percentage with each doubling of cumulative output—was formalized in aircraft manufacturing in the 1930s and has since been documented in semiconductors, solar modules, batteries, and wind turbines. The solar case is the most studied. Module prices have fallen from roughly $76 per watt in 1976 to under $0.30 per watt in the early 2020s, a learning rate of about 20 percent per doubling of cumulative capacity. Battery pack prices have fallen from over $1,000 per kilowatt-hour in 2010 to under $150 by the early 2020s, a learning rate of roughly 18 percent. These are extraordinary numbers, and they have been used to argue that the green transition is self-financing: deploy today's technology, ride the curve down, and tomorrow's technology becomes affordable.

The learning-curve literature, however, is careful to distinguish learning-by-doing from learning-by-searching and from scale economies, and the distinction matters for our argument. Learning-by-doing is the improvement in yield, throughput, and reliability that comes from repetitive production. Learning-by-searching is the improvement that comes from deliberate R&D. Scale economies are the reduction in unit fixed cost that comes from larger plants. In practice all three operate simultaneously, and decomposing them is difficult. The standard approach, following Wright's original formulation and its later refinements, regresses unit cost on cumulative output and a time trend, with the time trend absorbing autonomous technical change. The residual learning rate is then attributed to experience. This is a fragile identification, and the estimates should be treated as indicative rather than precise.

For the argument of this book, the precise decomposition matters less than the qualitative implication. If green performance improves with cumulative output in the same way that cost does, then the firms and countries that deploy early and at scale will accumulate a green capability advantage that latecomers cannot simply buy. The learning curve becomes a mechanism of path dependence in green competitiveness. This is the optimistic reading, and it underpins much of the policy enthusiasm for early deployment subsidies. The pessimistic reading is that the learning curve is global rather than national: knowledge spills across borders, latecomers free-ride on the experience of pioneers, and the pioneer's investment is socialized while the returns are competed away. Both readings have support. The solar industry's history contains elements of both—Chinese manufacturers scaled rapidly and captured the manufacturing value added, while the pioneering European and Japanese firms that developed the early technology captured much less of the eventual market.

The resolution turns on where the learning resides. If learning is embodied in capital equipment, it travels with the equipment and the equipment maker captures the rent. If learning is embodied in tacit process knowledge and supplier relationships, it stays with the firm and the cluster. Keeble and Wilkinson (1999), studying high-technology SMEs in European regions, found that collective learning—the shared development of knowledge through supplier and customer interaction—was a significant source of regional innovative capacity, and that it was strongly localized. Lawson and Lorenz (1999) made a similar argument about tacit knowledge and regional innovative capacity: the knowledge that matters for innovation is often not codified and therefore not easily transferred across distance. This is the theoretical basis for expecting green capability to cluster geographically even in a globalized industry.

6.2.2 The Green Learning Curve: What Exactly Is Being Learned

The phrase "green learning curve" is used loosely in policy discourse. It is worth specifying what is being learned, because different objects of learning have different implications for competitiveness.

The first object is abatement cost. A firm learning to reduce emissions per unit of output is learning a production function. The learning is embodied in process modifications, operating procedures, and maintenance routines. It is highly firm-specific and largely tacit. The returns to this learning are captured by the firm as lower compliance cost or higher margin. This is the learning that underlies the Porter Hypothesis in its weak form.

The second object is product performance. A firm learning to make a product with a lower environmental footprint—a module with lower embodied carbon, a battery with higher recycled content, a building material with lower thermal conductivity—is learning a product specification. The learning is partly codified (design rules, material specifications) and partly tacit (manufacturing tolerances, quality control). The returns are captured by the firm if the product attribute is valued by the market and if competitors cannot easily replicate it. This is the learning that underlies differentiation strategies.

The third object is system integration. A firm learning to integrate green technologies into a functioning system—a zero-carbon factory, a smart grid, a circular supply chain—is learning an architectural skill. The learning is largely tacit and resides in the firm's engineering organization and its relationships with suppliers and customers. The returns are captured by the firm as the ability to sell solutions rather than components, which typically commands higher margins and is stickier. This is the learning that underlies the climbing strategy.

The three objects of learning are hierarchically related. Abatement-cost learning is the entry ticket; product-performance learning is the differentiator; system-integration learning is the durable advantage. Firms typically progress through them in sequence, but the progression is not automatic and many firms stall at the first or second stage. The firms that stall at abatement-cost learning end up in the compliance trap: they meet the standard, they bear the cost, and they compete on price against other compliant firms. The firms that reach system-integration learning escape the trap because they are selling something that is not easily commoditized.

This hierarchy maps onto the empirical literature on eco-innovation adoption. Ryszko (2016), studying 292 Polish firms, finds that proactive environmental strategy is associated with technological eco-innovation and with firm performance, but the relationship is mediated by the firm's absorptive capacity—its ability to identify, assimilate, and exploit external knowledge. Firms with low absorptive capacity adopt end-of-pipe technologies (abatement-cost learning) but do not progress to product or system innovation. Przychodzen and Przychodzen (2015), studying publicly traded firms in Poland and Hungary, find a positive relationship between eco-innovation and financial performance, but the effect is stronger for product and organizational eco-innovations than for process eco-innovations. The process innovations reduce cost; the product and organizational innovations create value. The distinction is the same one I am drawing between abatement-cost learning and the higher stages.

6.2.3 The Conditions for Riding the Curve

The learning curve does not operate automatically. Four conditions determine whether a firm or a cluster can ride it.

The first is the persistence and predictability of demand. Learning requires repetition, and repetition requires a stable market. A firm that faces stop-go demand—a burst of orders when a subsidy is announced, then a collapse when it is withdrawn—cannot accumulate experience smoothly. The solar industry's history is a case study in this. The German feed-in tariff created a demand surge that supported early learning; the subsequent contraction and the Chinese scale-up shifted the learning to a different set of firms. The firms that rode the curve most successfully were those that could plan capacity expansion against a predictable trajectory of demand, which in practice meant those with access to policy-supported markets or to long-term offtake agreements.

The second is the presence of complementary assets. Teece (1986) showed that innovators capture value from innovation only when they control the complementary assets needed to commercialize it—manufacturing, distribution, service, brand. In green technology, the complementary assets are often manufacturing scale and supply-chain relationships. A firm with a superior green technology but no manufacturing scale will license the technology or be acquired; a firm with manufacturing scale but no proprietary technology will compete on cost. The firms that climb are those that combine both. This is why the Chinese solar and battery industries, which built manufacturing scale rapidly, were able to capture value even when the underlying cell and chemistry innovations originated elsewhere.

The third is the presence of a local demand signal strong enough to justify investment. The Chinese "dual carbon" goals, announced in September 2020, created such a signal for domestic firms. The national emissions trading scheme, the green building standards, the renewable portfolio standards, and the local government targets all generated demand for green technologies and for the capability to deploy them. A firm serving only export markets faces a weaker and more distant signal; a firm serving a large domestic market with binding green targets faces a strong and proximate one. This is a point Chapter 3 developed at the policy level; here it matters at the firm level, because the learning curve is ridden faster when the feedback from the market is rapid.

The fourth is the presence of a cluster with collective learning. Keeble, Lawson, Moore, and Wilkinson (1999), studying the Cambridge region, found that collective learning processes and institutional thickness—the density of supporting organizations and shared norms—were significant determinants of innovative performance among high-technology SMEs. The same logic applies to green capability. A firm embedded in a cluster with shared testing facilities, a common labor pool trained in green process skills, and suppliers who understand the firm's environmental requirements will learn faster than an isolated firm. The cluster provides the "buzz" of tacit knowledge that cannot be acquired through formal channels. This is the microfoundation for the cluster-level argument in Chapter 9.

6.2.4 The Learning Curve and the Risk of Lock-In

There is a less discussed risk in the learning-curve framework: lock-in to a suboptimal technology. If learning-by-doing reduces the cost of technology A faster than the cost of technology B, the market may adopt A even if B has a higher long-run potential, because A's early cost advantage forecloses the investment needed to move B down its own curve. This is the standard path-dependence argument (David, 1985; Arthur, 1989), and it applies with force to green technologies, where the relevant comparison is often between an incumbent fossil technology and a challenger renewable one, or between two competing low-carbon pathways.

The risk is particularly acute in technology-intensive industries because the learning curve is steep and the window for switching is narrow. A firm that commits to a particular battery chemistry, a particular photovoltaic cell architecture, or a particular green hydrogen pathway accumulates learning specific to that choice. Switching later means writing off the accumulated learning and starting a new curve. The result can be a market that settles on a technology that is good enough rather than best, and a set of firms that are locked into a capability that becomes obsolete when the technology frontier moves.

This is not a hypothetical. The photovoltaic industry's shift from multicrystalline to monocrystalline silicon, and then to passivated emitter and rear cell (PERC) architectures and now to tunnel oxide passivated contact (TOPCon) and heterojunction (HJT) architectures, has repeatedly stranded the learning investments of firms that committed too early to a particular design. The battery industry's shift from lithium iron phosphate (LFP) to nickel manganese cobalt (NMC) and back toward LFP variants illustrates the same dynamic. The firms that have ridden the curve most successfully are those that built learning capability that is transferable across designs—process engineering, quality control, supplier management—rather than learning tied to a specific design. The distinction between learning that is specific to a technology and learning that is specific to a firm's ability to develop technologies is the distinction between a capability that depreciates and one that compounds.

6.3 Case Studies of Technology-Intensive Firms

The three cases that follow illustrate the mechanism in different configurations. The first is an established industrial automation firm that used the green agenda to move from selling components to selling system solutions—the climbing strategy in its clearest form. The second is a battery materials firm that rode the learning curve but faces the risk of lock-in as the chemistry frontier moves. The third is a photovoltaic manufacturer that achieved scale but has struggled to convert scale into durable green capability, illustrating the limits of the learning curve when complementary assets are missing. The cases are composites, drawn from firm-level interviews and public reporting, and the names are anonymized where the firms did not consent to attribution.

6.3.1 Case 1: Industrial Automation and the System-Integration Climb

Henggong Automation is headquartered in Shenzhen and employs about 4,000 people. It began in the 1990s as a distributor of imported programmable logic controllers (PLCs) for the domestic manufacturing market, moved into proprietary PLC design in the 2000s, and by the 2010s had become a significant supplier of factory automation systems to Chinese cement, steel, and chemical plants. Its traditional business was selling hardware and the software licenses that ran it. The green agenda changed the sales conversation.

The change began with the cement industry. Cement production accounts for roughly 7 percent of global carbon dioxide emissions, and the Chinese cement industry is the world's largest. The sector's abatement options are well understood: substitute alternative fuels for coal in the kiln, improve thermal efficiency, blend clinker with supplementary cementitious materials, and capture and store the residual carbon. The first three are process changes that require precise control of kiln temperature, fuel feed rates, and raw meal composition. That control is an automation problem.

Henggong's engineers had been selling kiln control systems to cement plants for years, but the systems were sold as hardware plus a service contract, priced per point of control. When the dual carbon goals were announced in 2020 and the cement industry came under pressure to reduce coal consumption, the sales conversation shifted. Plant managers were no longer asking how many control points they needed; they were asking how much coal they could save per ton of clinker, and whether the savings would pay back the investment within the two-year window their parent company imposed on capital projects.

Henggong's response was to build a system-level offering. The company developed a kiln optimization package that combined model-predictive control of the kiln with a real-time coal-consumption model and an alternative-fuel feed-rate controller. The package was sold not as hardware but as a performance contract: Henggong would install the system, operate it for a trial period, and be paid a share of the verified coal savings. The first installation, at a cement plant in Anhui province in 2021, reduced coal consumption by about 4 percent and increased the alternative-fuel substitution rate from 2 percent to 9 percent over eight months. The plant's parent company ordered the package for four more lines.

The economics of the shift are instructive. The hardware margin on a conventional kiln control system was perhaps 25 percent; the performance-contract margin on the optimization package was higher, but the more important difference was in the stickiness of the relationship. A hardware sale is a transaction; a performance contract is a partnership that generates continuous data and continuous improvement. Henggong's engineers learned the plant's operating characteristics in detail, and that learning fed back into the product. By 2023 the company had installed the package at 37 cement lines and had begun adapting it to steel and chemical plants. The system-integration capability it built in cement—the ability to model a complex thermal process, instrument it, control it, and verify the savings—was transferable to other process industries.

Henggong did not invent the model-predictive control algorithms; those are available from vendors and in the literature. It did not invent the alternative-fuel burner; those are supplied by equipment makers. What it built was the capability to integrate these components into a system that delivered verified savings to a specific customer in a specific regulatory context. That capability is tacit, embedded in the firm's engineering organization and its accumulated knowledge of cement kilns, and it is not easily replicated by a hardware vendor or a software startup. The learning curve here is not the cost curve of a manufactured product; it is the performance curve of a service capability. Each installation makes the next one cheaper and more effective, because the firm's model of the kiln improves and its library of operating scenarios grows.

The green outcome is real but should not be overstated. A 4 percent reduction in coal consumption per ton of clinker, at a plant producing 3 million tons per year, saves roughly 40,000 tons of coal and about 100,000 tons of carbon dioxide annually. Multiply by 37 lines and the aggregate is meaningful but small relative to the industry's total emissions. The significance of the case is not the tonnage; it is the demonstration that a Chinese automation firm can sell a green performance contract to a domestic heavy-industry customer and build a transferable capability in the process. That capability is the raw material for the export strategy that Chapter 8 discusses: the same system, adapted to local fuel prices and regulatory requirements, could be sold to cement plants in Southeast Asia, the Middle East, and Africa, where the abatement pressure is rising but the local automation capability is thin.

The case also illustrates the role of the domestic demand signal. Henggong's optimization package would not have found buyers in 2018, when coal was cheap and carbon was unpriced. The dual carbon goals, the national emissions trading scheme's inclusion of the cement sector (announced in 2024), and the local government's energy-intensity targets created a demand pull that made the performance contract viable. The firm's engineers had the capability to build the system before the demand existed; the policy signal converted latent capability into a market. This is the mechanism Chapter 3 described at the macro level, operating at the level of a single firm's sales pipeline.

There is a counterfactual worth stating. If Henggong had responded to the green agenda by installing end-of-pipe abatement equipment—dust filters, denitrification units—it would have captured a one-time equipment margin and no durable capability. The choice to build a system-level offering was a strategic choice, not a technological necessity. The firm's management made it because the performance-contract model aligned the firm's incentives with the customer's savings, and because the firm's engineering culture rewarded deep process knowledge over hardware sales. Not every automation firm made the same choice. Several of Henggong's domestic competitors continued to sell hardware and lost share as the performance-contract model spread. By 2023, two of the three largest domestic kiln-control vendors had introduced their own performance-based offerings, but Henggong had a two-year lead in installed lines and a library of operating data that the followers could not quickly replicate. The learning curve rewarded the firms that climbed it first.

6.3.2 Case 2: Battery Materials and the Chemistry Frontier

Rongke Materials is a battery cathode material producer based in Hunan province. It was founded in 2010, grew rapidly as a supplier of lithium iron phosphate (LFP) cathode material to Chinese battery cell makers, and by 2020 had become one of the largest LFP producers in the country. Its growth tracked the learning curve of the LFP chemistry: as cell makers scaled production and drove down the cost per kilowatt-hour, Rongke scaled its cathode production and drove down its own cost per ton. The company's engineers became expert in the synthesis of LFP powder, the control of particle size and morphology, and the doping and coating steps that improve conductivity. That expertise is the firm's core asset.

The green dimension of Rongke's business is real but indirect. LFP cells are used in electric vehicles and grid storage, and the environmental benefit of the end product is the reason the market exists. At the production level, cathode material manufacturing is energy-intensive and generates process emissions, and the firm's customers—particularly European automakers—have begun to require carbon footprint declarations for the materials they buy. The EU Battery Regulation, adopted in 2023, requires a carbon footprint declaration for batteries placed on the European market from 2025, with maximum thresholds to follow. Rongke's European customers began asking for verified footprint data in 2022.

Rongke's response was to build a carbon accounting capability and to invest in process improvements that reduced the footprint of its cathode material. The main levers were the carbon intensity of the electricity used in synthesis (the firm signed a green power purchase agreement with a provincial wind farm), the energy efficiency of the sintering furnaces (a retrofit that reduced natural gas consumption per ton by about 12 percent), and the recycled content of the raw materials (the firm began sourcing recycled lithium and iron from a partner). By 2023 the firm could provide a third-party-verified footprint declaration for its LFP material, and it had won a supply contract with a European cell maker on the strength of that declaration.

So far this is a compliance success story, not a climbing story. The footprint declaration is a defensive asset: it allows Rongke to sell into a regulated market, but it does not command a premium, because the European cell maker can source from other compliant suppliers. The firm's green capability is real—it knows how to measure and reduce its footprint—but it is not yet a source of differentiation. The question is whether Rongke can convert that capability into a climbing strategy.

The answer depends on the chemistry frontier. LFP has displaced NMC in many vehicle segments because of its lower cost and improving energy density, but the frontier is moving toward higher-manganese LFP variants, sodium-ion chemistries, and solid-state designs. Each of these requires different cathode materials and different synthesis processes. Rongke's accumulated learning in LFP synthesis is specific to the LFP chemistry; it does not transfer automatically to sodium-ion or solid-state. The firm faces the lock-in risk described earlier. If it invests heavily in scaling the current LFP process, it rides the learning curve further down, but it may strand that learning when the frontier moves. If it diversifies into new chemistries, it sacrifices some of the scale economies that made it competitive.

Rongke's management chose a middle path. The firm continued to scale LFP production but allocated about 15 percent of its R&D budget to a sodium-ion cathode program and a further 10 percent to a recycling process that would allow it to recover lithium and iron from end-of-life cells. The recycling program is the more interesting bet, because it builds a capability that is transferable across chemistries: the ability to recover and purify battery materials is a process skill that applies regardless of the cathode chemistry. If the recycling capability matures, Rongke could become a supplier of recycled feedstock to the entire industry, which is a higher-value position than being one of many LFP cathode makers.

The case illustrates the distinction between learning that is specific to a technology and learning that is specific to the firm's ability to develop technologies. Rongke's LFP synthesis learning is technology-specific and will depreciate as the chemistry frontier moves. Its recycling learning is capability-specific and will compound. The firm's green competitiveness in 2030 will depend more on the second than on the first. This is a general lesson for technology-intensive firms facing rapid technological change: the durable green advantage is not in any particular product's environmental attribute but in the organizational capability to develop and manufacture low-footprint products across a changing technology portfolio.

The case also illustrates the limits of the compliance strategy. Rongke's footprint declaration was necessary to maintain access to the European market, but it did not change the firm's competitive position. The firms that will climb in the battery materials industry are those that use the compliance requirement as a forcing device to build capabilities that are valuable beyond compliance—recycling, low-carbon process engineering, supply-chain traceability. Rongke has begun that transition, but it is not yet complete, and the outcome is genuinely uncertain. I would not predict it.

6.3.3 Case 3: Photovoltaics and the Scale-without-Capability Trap

The third case is the most cautionary. Sunshine Solar is a photovoltaic module manufacturer based in Jiangsu province. It was founded in 2008, survived the industry's consolidation in the early 2010s, and by 2020 had annual module capacity of about 5 gigawatts. It is a mid-tier player in an industry dominated by a handful of giants with capacity above 30 gigawatts. Its cost per watt is competitive but not leading; its technology is a generation behind the leaders; and its margins are thin.

Sunshine's green story is the industry's green story. Its product enables the displacement of fossil electricity, and the lifetime carbon payback of a module is measured in months. The firm's manufacturing footprint, however, is substantial: polysilicon purification and wafer slicing are energy-intensive, and the firm's Jiangsu facilities draw on a grid that is still roughly 60 percent coal-fired. The embodied carbon of Sunshine's modules is therefore higher than that of competitors manufacturing in regions with cleaner grids or with more efficient processes.

This matters because the European market, which accounts for about a quarter of Sunshine's sales, has begun to discriminate on embodied carbon. The EU's Ecodesign requirements for photovoltaic modules, under development, are expected to include a carbon footprint declaration and possibly a maximum threshold. Several European utilities have announced procurement preferences for low-carbon modules. Sunshine's response has been to purchase green certificates for a portion of its electricity consumption and to commission a carbon footprint study, but it has not invested in the process upgrades that would structurally reduce its footprint—the energy-efficient polysilicon process, the on-site renewable generation, the closed-loop wafer slicing.

The reason is financial. Sunshine's margins are too thin to fund the capital investment, and its scale is too small to spread the fixed cost of the upgrade across enough output to bring the unit cost down. The learning curve that would make the upgrade affordable is precisely the curve the firm cannot get onto because it cannot afford the first step. The firm has enough scale to be a commodity producer but not enough to fund the transition to a differentiated position—the scale-without-capability trap. It is stuck in the compliance trap at a higher level of capital intensity than the Shunde furniture factory, but the logic is the same.

The contrast with the industry leaders is stark. The leading Chinese module manufacturers have invested in low-carbon polysilicon, in energy-efficient cell processes, and in on-site renewable generation. They have done so because their scale allows them to spread the cost, and because their access to capital markets is better. The result is that the green transition in the photovoltaic industry is reinforcing the concentration of the industry, not democratizing it. The learning curve, in this configuration, is an instrument of concentration. This is the risk I flagged earlier, and Sunshine is the evidence for it.

There is a policy dimension. The Chinese government's green manufacturing programs, which designate "green factories" and provide subsidies for energy-efficient equipment, are available to firms of Sunshine's size, but the transaction costs of applying are high relative to the benefit. The provincial government's green credit window offers below-market loans for energy-efficiency upgrades, but Sunshine's balance sheet does not support the debt service. The firm is caught in a gap between the scale needed to access the programs and the scale it has. This gap is a recurring feature of industrial policy for the green transition, and Chapter 9 returns to it.

Sunshine's prospects are not hopeless. The firm could specialize in a niche—building-integrated photovoltaics, or modules for high-temperature or high-humidity environments—where the environmental attribute matters less than the performance attribute and where the firm's engineering can differentiate. It could partner with a larger firm as a contract manufacturer and ride the partner's learning curve. It could exit the module business and redeploy its manufacturing capability into a segment where its cost position is more defensible. Each of these is a strategic choice, and the firm's management is weighing them. The point of the case is not that Sunshine is doomed; it is that the learning curve does not lift all firms, and the firms it leaves behind are not necessarily less capable—they are less able to bear the fixed cost of the first step onto the curve.

6.3.4 Cross-Case Observations

Three patterns emerge from the cases.

First, the climbing strategy is a choice, not a technological inevitability. Henggong chose to build a system-integration capability; Rongke chose to invest in recycling; Sunshine did not choose to invest in process upgrades. The choices were shaped by the firms' existing capabilities, their access to capital, and the demand signals they faced, but they were choices. Henggong's vice president of engineering, who joined the firm in 2005 from a state-owned cement equipment institute, told me the decision to move to performance contracts was made in a single meeting in late 2020, against the objections of the sales director, who feared the revenue volatility. The firm's chairman overruled him. That is a strategic decision made under uncertainty, not a response determined by the technology.

Second, the learning curve operates atthe level of the firm and at the level of the cluster, and the two levels interact. Henggong's system-integration capability was built on the Shenzhen automation cluster's deep pool of control engineers and its dense supplier network. Rongke's cathode capability depended on the Hunan battery-materials cluster's shared testing facilities and its access to skilled process engineers. Sunshine's isolation from the leading-edge cluster in Jiangsu—it is in the same province as the industry leaders but not embedded in their supplier networks—limited its access to the tacit knowledge that would have made the process upgrade feasible. The cluster is not a backdrop; it is a component of the learning mechanism.

Third, the green attribute of the product and the green attribute of the process are distinct sources of competitiveness, and firms often confuse them. Sunshine's product is green in use but brown in production. Henggong's product is not green in itself—a kiln control system has no environmental attribute—but the process capability it embodies delivers green outcomes for the customer. Rongke's product is green in use and increasingly green in production. The firms that climb are those that can align the two attributes, or that can build a capability that is valuable regardless of which attribute the market rewards. The confusion between product-green and process-green is a recurring source of strategic error in technology-intensive industries, and it is worth stating plainly: a firm that sells a green product made in a brown factory is exposed to the same carbon border adjustment as a firm that sells a brown product, and it has less margin to absorb the cost.

6.4 The Learning Curve and the Boundaries of the Argument

6.4.1 What the Chapter Has and Has Not Shown

The evidence assembled here supports a conditional claim. Technology-intensive industries can convert the green agenda from a compliance cost into a competitive advantage when three conditions hold: the firm has or can build a capability that is at least partly tacit and non-imitable; the demand signal is strong and persistent enough to support cumulative learning; and the firm has access to the complementary assets—manufacturing scale, supplier networks, capital—needed to commercialize the capability. When these conditions hold, the learning curve operates in green performance as it does in cost, and the firm can climb rather than merely comply. When they fail, the firm faces the compliance trap at a higher level of capital intensity, and the learning curve reinforces the advantage of the firms that are already ahead.

The chapter has not shown that technology-intensive industries as a category are green winners. The photovoltaic case shows that scale without capability can leave a firm stranded. The battery case shows that capability specific to a technology can depreciate when the frontier moves. The automation case shows that the climbing strategy is available but requires a deliberate organizational choice. The heterogeneity within the category is as large as the heterogeneity between categories, and the policy implication is that sector-level classifications are a poor guide to firm-level outcomes.

Nor has the chapter resolved the question of whether the green learning curve is national or global. The Chinese solar and battery industries rode curves that were partly built on knowledge developed elsewhere, and the European and Japanese pioneers captured less of the value than their early investment might have warranted. The automation case suggests that system-integration capability is stickier and more localized than component manufacturing capability, which is a reason to expect the climbing strategy to be more defensible than the scale strategy. But this is a hypothesis, not a finding, and it would require firm-level data across multiple countries to test properly.

6.4.2 The Compliance Trap in a Higher Key

The compliance trap described in Chapter 5 has a technology-intensive variant. In labor-intensive industries, the trap is that compliance costs are absorbed at thin margins, leaving no surplus for the investment needed to climb. In technology-intensive industries, the trap is that compliance costs are absorbed at adequate margins, but the investment needed to climb is large and lumpy, and the firm's scale is insufficient to spread it. The first trap is a margin problem; the second is a scale problem. Both leave the firm meeting the standard without capturing the rent.

The scale problem is more tractable than the margin problem, because it can be addressed by policy instruments that reduce the fixed cost of the first step—subsidized green credit, shared testing facilities, public procurement of low-carbon products, cluster-level infrastructure. The margin problem is harder, because it requires either a redistribution of value along the chain or an exit from the segment. This is a reason to be more optimistic about the green transition in technology-intensive industries than in labor-intensive ones, but the optimism should be tempered by the recognition that the scale problem can persist for a long time in mid-tier firms, and that the firms caught in it are not necessarily less efficient than the leaders—they are less able to bear the lumpy investment.

There is a further twist. The scale problem in technology-intensive industries is often a financing problem, and the financing problem is often a function of the firm's position in the value chain. A mid-tier module manufacturer sells to project developers and utilities, which have bargaining power and pay late. A mid-tier cathode material producer sells to cell makers, which have even more bargaining power and pay later. The working capital burden of the green upgrade falls on the firm with the least bargaining power, which is precisely the firm that can least afford it. The governance of the value chain, in other words, shapes the distribution of the capability to climb. This is a point that the GVC literature has established for other forms of upgrading (Gereffi, Humphrey, and Sturgeon, 2005), and it applies with force to green upgrading. The lead firms that set the environmental requirements also set the payment terms, the volume commitments, and the technical specifications, and these terms determine whether the supplier has the resources to invest in the capability the lead firm is demanding.

6.4.3 The Coined Term: Green Capability Rent

I want to introduce a term that the rest of the book will use. Green capability rent is the surplus a firm earns from a green capability that competitors cannot easily replicate. It is distinct from compliance cost savings, which are available to any compliant firm, and from green premium, which is the price premium a product commands in the market. Green capability rent arises when the firm's accumulated learning—in process, product, or system integration—lowers its cost of meeting environmental requirements below the industry average, or allows it to offer an environmental performance that competitors cannot match. It is a rent in the classical sense: a return to a scarce resource, where the resource is the firm's accumulated capability.

The concept is useful because it clarifies what the climbing strategy is trying to capture. A firm that invests in green capability and earns only compliance cost savings has not climbed; it has reduced its cost of complying. A firm that earns a green premium has differentiated its product, which is a form of climbing, but the premium is vulnerable to imitation and to the erosion of consumer willingness to pay. A firm that earns green capability rent has built something that is durable because it is embedded in the firm's organization and history. The three cases illustrate the three outcomes: Henggong is earning green capability rent from its system-integration skill; Rongke is earning a green premium that may or may not become a rent depending on whether its recycling capability matures; Sunshine is earning neither, and is absorbing compliance cost.

The concept also clarifies the policy problem. Green capability rent is a private return to a capability that has public benefits—lower emissions, faster diffusion of green technology. The firm that builds the capability captures only part of the social return, because the knowledge spills over to competitors and customers. The underinvestment that results is the dual-externality problem in a different guise. The policy response is not to subsidize the rent directly but to lower the cost of building the capability—through shared infrastructure, training, and the coordination of demand—so that the private return is closer to the social return. This is the argument Chapter 9 develops in detail.

6.4.4 A Rhetorical Question, and an Answer

If the learning curve is real, and if it rewards the firms that climb it, why do so many technology-intensive firms remain in the compliance trap? The question is not rhetorical in the sense of having no answer. It has an answer, and the answer is that the learning curve is a conditional mechanism, not a law of nature. It operates when the conditions described above hold, and it fails when they do not. The firms in the trap are not irrational; they are responding to the incentives they face, and those incentives—thin margins, weak demand signals, limited access to finance, unfavorable value-chain positions—do not support the investment. The policy problem is to change the incentives, not to exhort the firms.

There is a second answer, which is more uncomfortable. Some firms in the compliance trap are there because their managers have not recognized the strategic opportunity. The Henggong case suggests that the climbing strategy required a deliberate reframing of the firm's value proposition, from selling hardware to selling verified savings. Not every manager makes that reframing. The literature on managerial cognition and strategic change (e.g., Tripsas and Gavetti, 2000) shows that incumbent firms often fail to adapt to technological change not because they lack the resources but because their managers' mental models are anchored in the old business. The green transition is a technological change of this kind, and the cognitive barrier is real. This is a reason for policy to include demonstration projects and peer learning, not just subsidies.

6.5 Implications for the Book's Argument

6.5.1 Green Governance Capability as an Endogenous Driver

The chapter's contribution to the book's central argument is to show the mechanism by which green governance capability becomes an endogenous driver of upgrading. In the technology-intensive case, the mechanism is the learning curve: the firm that invests in green capability accumulates learning that lowers its future cost of compliance and opens new market positions. The capability is not a passive product of regulation; it is an asset that the firm builds, and the building is a strategic choice. The regulation creates the demand signal, but the firm's response determines whether the signal is converted into capability or absorbed as cost.

This is the sense in which the book's argument differs from the Porter Hypothesis in its strong form. Porter argued that regulation triggers innovation that offsets compliance cost. The argument here is that regulation triggers a learning process whose outcome depends on the firm's absorptive capacity, its complementary assets, and its value-chain position. The innovation offset is conditional, not automatic. The firms that climb are those that can convert the regulatory signal into a capability; the firms that do not are those that cannot. The policy implication is that the design of the regulation matters less than the design of the support system that determines whether firms can respond.

6.5.2 The Link to the Furniture Case

The furniture industry, which Chapter 8 examines in detail, is a labor-intensive industry, and the contrast with the technology-intensive cases is instructive. The Shunde factory's carbon declaration was a compliance investment that did not generate a capability rent, because the firm's scale and its position in the value chain did not allow it to convert the declaration into a differentiated position. The technology-intensive firms that climbed did so because they had the scale, the complementary assets, and the value-chain position to convert their green investments into capabilities. The difference is not in the firms' willingness to invest; it is in their capacity to convert the investment into a durable asset.

This contrast has a policy implication that Chapter 9 develops. The support system for green capability building should be differentiated by the firm's position in the value chain and by the nature of the capability it is trying to build. A labor-intensive furniture maker needs help with the fixed cost of compliance and with the collective infrastructure—shared testing, common carbon accounting—that reduces the cost of the first step. A technology-intensive automation firm needs help with the demand signal and with the demonstration of system-level solutions. A mid-tier module manufacturer needs help with the financing of lumpy process upgrades. A single instrument will not serve all three.

6.5.3 The Boundaries of the Chapter's Claim

I want to be explicit about what the chapter does not claim. It does not claim that technology-intensive industries will solve the green transition. It does not claim that the learning curve will lift all firms, or that the firms left behind are less deserving. It does not claim that the Chinese experience is generalizable to other countries, although the mechanisms are likely to operate wherever the conditions hold. And it does not claim that the climbing strategy is available to every firm in every technology-intensive sector; the automation case suggests that it is available where the firm can build a system-integration capability, which is not universal.

What the chapter does claim is that the learning curve is a real mechanism, that it operates in green performance as well as in cost, and that its operation is conditional on firm-level and cluster-level conditions that policy can influence. The claim is modest, but it is enough to support the book's broader argument that green governance capability is an endogenous driver of upgrading rather than a tax on it. The next chapter turns to capital-intensive industries, where the mechanism operates differently—through the governance of large-scale process substitution and the management of stranded assets—and where the stakes are higher because the capital at risk is larger.

6.6 Coda: A Return to the Factory Floor

The Henggong engineers I interviewed in Shenzhen worked in an open-plan office above a testing floor where a scaled-down cement kiln control system ran continuously. The system was being used to train the firm's service engineers and to test new control algorithms against recorded plant data. One of the engineers, a woman in her thirties who had joined the firm from a cement plant in Anhui, explained that the hardest part of the job was not the control algorithm but the translation. "The plant manager does not care about carbon," she said. "He cares about coal cost per ton. If I can show him the coal saving, he will listen to the carbon story later."

That translation—from the environmental metric to the operational metric that the customer actually manages—is the practical form of the capability the chapter has described. It is not a technology; it is a skill of framing, embedded in the engineer's understanding of both the plant and the policy. The firms that climb are those that have engineers who can make that translation, and organizations that reward them for doing so. The learning curve is, in the end, a curve of organizational learning, and the unit of learning is the person who can connect the environmental requirement to the operational reality. The next chapter moves from the factory floor to the capital project, where the translation is harder and the stakes are larger.