Chapter 10: Government-Guided Venture Capital and the Green Premium

A government fund can pay for a first-of-a-kind plant. The plant can prove the chemistry works. Neither can prove that anyone will buy the output at a price that recovers the incremental cost of making it clean. That gap between demonstrated technology and bankable demand is where the green premium lives, and it is where government-guided venture capital either earns its keep or becomes an expensive way to subsidize learning curves that never bend.

The chapter ahead works through that gap. It defines the green premium precisely enough to be measured, since a concept that cannot be measured cannot be targeted. It sets out the channels through which government-guided venture capital (GVC) can act on the premium. And it examines what the empirical record actually shows, which is less than the promotional literature claims and more than the skeptics allow.

My argument, stated up front: GVC is necessary for the capital-intensive, first-of-a-kind segments of the transition, and it is close to useless—possibly harmful—when deployed as a general subsidy for commercially viable technologies that private capital would fund anyway. The distinguishing test is whether the investment changes the slope of the cost curve for a technology that has not yet reached self-sustaining scale, or merely subsidizes a firm that has. Most evaluations do not make this distinction, which is why the literature appears to contradict itself.

10.1 The Green Premium: Definition and Measurement

10.1.1 What the premium is, and what it is not

Bill Gates popularized the term in How to Avoid a Climate Disaster (2021), but the underlying economics predates the book by at least a decade. The green premium is the difference between the cost of a zero-carbon substitute and the cost of the fossil incumbent it replaces, expressed per unit of output. For cement, it is dollars per tonne. For steel, dollars per tonne of crude steel. For aviation fuel, dollars per tonne of fuel, or per megajoule if you prefer. For electricity, dollars per megawatt-hour.

The definition sounds simple. Three complications make it hard to operationalize.

The first is the boundary problem. The premium depends on what you count. A steel mill that switches from blast furnace to hydrogen-based direct reduced iron (DRI) faces a higher cost per tonne of steel, but if you credit it for the avoided carbon and value that carbon at a social cost, the premium shrinks or vanishes. Whether you do that crediting is a policy choice, not an accounting fact. I will treat the unsubsidized premium as the baseline—the gap a firm faces before any carbon price, subsidy, or offset—because that is the gap that determines whether private capital will invest without public support.

The second is the reference-case problem. The premium is measured against a fossil incumbent whose cost is itself volatile. When natural gas prices spiked in Europe in 2022, the green premium for hydrogen-based steel collapsed, in some estimates to near zero, not because green steel got cheaper but because grey steel got more expensive. A premium that moves with fossil fuel prices is a poor target for a fixed subsidy; it argues for instruments that float with the reference price, which is one reason the contract-for-difference designs I discuss later matter.

The third is the learning-curve problem. The premium is not a constant. It declines as cumulative production rises, and the rate of decline differs enormously across technologies. Solar photovoltaic modules fell roughly 90 percent in cost between 2010 and 2020. Green hydrogen has fallen far less, and the electrolyzer stack—the core capital cost—has shown stubbornly slow cost reduction. A policy that targets today's premium without accounting for the expected slope of the cost curve will overpay for technologies that would have gotten cheap anyway and underpay for those that need a push.

Formally, if \(c_0(t)\) is the unit cost of the zero-carbon technology at time \(t\) and \(c_f(t)\) is the unit cost of the fossil incumbent, the green premium is

where \(p_c(t)\) is the effective carbon price (explicit or implicit) and \(\epsilon(t)\) is the emissions intensity of the fossil incumbent per unit of output. The term \(p_c \epsilon\) is what policy can move directly; the term \(c_0 - c_f\) is what innovation and scale can move. GVC acts primarily on \(c_0\) through learning and scale, and secondarily on the credibility of \(p_c\) through the signal it sends about future policy.

10.1.2 Measurement: the CICC estimates and their limits

The most widely cited sectoral estimates for China come from CICC Research. Their 2021 work put the green premium for non-passenger transport and for building materials (glass, cement) at 141 percent and 138 percent respectively—meaning the clean substitute costs more than twice the incumbent. For paper, non-ferrous metals, steel, passenger vehicles, and power, the premium ranged from 3 to 17 percent.

Table 10-1 reproduces the broad pattern.

Table 10-1: Approximate green premiums by sector, after CICC Research (2021). Ranges are indicative; the underlying estimates are sensitive to fossil fuel prices and grid emission factors.

These numbers should be read with caution. They are point estimates for a particular year, a particular region, and a particular technology pathway. The 3–17 percent band is wide enough to contain both "investable with a small carbon price" and "not investable without a large one." And the high-premium sectors—cement, long-haul transport—are precisely those where the technology is least mature, so the premium is measured against a demonstration cost rather than a commercial one.

The measurement problem has a deeper layer. A green premium computed at the plant gate ignores the cost of integrating intermittent supply, the cost of building hydrogen pipelines and storage, and the cost of retiring stranded fossil assets early. These system costs can exceed the plant-level premium. A steel mill that switches to hydrogen needs a hydrogen supply chain that does not yet exist at scale; the premium it faces is not the difference in feedstock cost but the difference in the cost of an entire reconfigured supply chain. I have seen estimates for European green steel that put the system-level premium at two to three times the plant-gate figure.

For the purposes of this chapter, the practical implication is this: the green premium is a range, not a number, and it is a range that shifts with fossil prices, learning rates, and system build-out. Policy that treats it as a fixed target will misallocate capital. Policy that treats it as a moving target, and designs instruments that track the movement, has a chance.

10.1.3 Why the premium is the right target for GVC

There is a case for directing public equity at the premium rather than at the technology. The technology may work; the question is whether it can be sold. A venture fund that invests in a green hydrogen start-up is betting on the start-up's ability to produce at a cost that closes the premium. If the premium is 80 percent and the start-up's roadmap gets it to 30 percent in five years, the investment is a bet on the roadmap. If the roadmap stalls at 60 percent, no amount of venture capital saves the company; it needs a carbon price or a mandate.

This is why GVC and carbon pricing are complements, not substitutes. GVC pushes down \(c_0\). Carbon pricing raises \(c_f\) (effectively) and makes the premium smaller and more predictable. A portfolio of GVC investments without a credible carbon price is a portfolio of bets on cost curves that may never cross. A carbon price without GVC is a price signal that capital markets may not be able to act on, because the technologies that would respond are not yet at scale.

The interaction is the heart of the matter. I will return to it in Section 10.3.

10.2 Mechanisms of Government-Guided Venture Capital

10.2.1 What "government-guided" means

The term covers a wide range of institutional forms. At one end, a sovereign wealth fund making direct equity investments in green technology companies. At the other, a government that provides matching capital to private venture funds on the condition that they invest in a defined sector. In between: fund-of-funds structures, co-investment vehicles, guarantees, and hybrid instruments that blend concessional and commercial capital.

The Chinese landscape is instructive because it has experimented with most of these forms. The National Integrated Circuit Industry Investment Fund (the "Big Fund"), established in 2014, is the largest and most studied example, though it targets semiconductors rather than green technology. Its green analogues include provincial-level guidance funds (yindao jijin, 引导基金), which by some counts number in the thousands and hold assets in the hundreds of billions of renminbi. The scale is real; the governance is uneven.

For this chapter I will use a functional definition. GVC is any equity or quasi-equity investment in which a government entity (a) provides capital, (b) exercises some influence over investment selection, and (c) accepts a risk-return profile that private capital would not accept on its own. The third condition is the one that matters. A government fund that invests on commercial terms in commercially viable companies is not doing GVC in any meaningful sense; it is doing asset management with a public balance sheet.

10.2.2 The signaling channel

The first mechanism is the simplest and the most documented. A government investment signals to private investors that a technology or a firm has passed a screening threshold they cannot easily replicate. This is the "certification" effect.

The logic runs through information asymmetry. A private venture capitalist evaluating a first-of-a-kind electrolyzer company faces a technology risk she cannot fully diligence: the chemistry is novel, the failure modes are unknown, and the relevant expertise is scarce. A government fund with access to national laboratories, technical agencies, and a mandate to develop the sector can, in principle, do that diligence better. When it invests, it reveals information. Private investors update.

The empirical literature on this channel is mixed but leans positive for early-stage investments. Brander, Du, and Hellmann (2015) found that companies receiving government venture capital in Canada attracted more private financing subsequently, though the effect was concentrated in early-stage deals and weakened for later-stage ones. Cumming and Johan (2009) found similar certification effects in a sample of pre-seed government funds. The mechanism appears to work best when the government fund has genuine technical capacity and a reputation for picking winners—which is to say, when the signal is credible.

Credibility is the binding constraint. A government fund that invests for political reasons—to hit a sectoral target, to support a favored region, to rescue a failing champion—destroys its own signaling value. Private investors learn to discount the signal, and the certification effect disappears. I have seen this in provincial guidance funds in China, where the correlation between government investment and subsequent private investment is weak or negative in sectors that have been subject to repeated policy campaigns.

The signaling channel also has a timing problem. The signal is most valuable at the earliest stage, when uncertainty is highest. But early-stage investments are the riskiest and the least likely to produce the financial returns that government funds are often judged on. A fund that is evaluated on its internal rate of return will drift toward later-stage, safer deals, and the signaling value will erode. This is a governance problem, not a market problem, and it is the subject of Chapter 11.

10.2.3 The de-risking channel

The second mechanism is more direct. Government capital can absorb a portion of the risk that makes private investors unwilling to commit, changing the risk-return profile of a deal from unacceptable to acceptable.

The standard structure is a co-investment in which the government takes a subordinated position—first loss, or a capped return—while private investors take a senior or preferred position. The government's concession is the price of mobilizing private capital. The private investor's return is protected on the downside; the government absorbs the tail.

This works, but the arithmetic is unforgiving. If the government absorbs the first 20 percent of losses on a portfolio, and the portfolio's expected loss rate is 30 percent, the government has absorbed two-thirds of the expected loss and the private investor is effectively investing in a riskless asset with an equity upside. That is a transfer, not a market. The design question is how much subordination is enough to mobilize capital without turning the vehicle into a subsidy with extra steps.

The European Investment Fund's risk-sharing facilities and the UK's British Business Bank have both grappled with this. The general finding is that a modest first-loss layer—10 to 15 percent—is enough to move private capital in sectors where the underlying technology is proven but the market is thin. In sectors where the technology itself is unproven, no amount of subordination mobilizes private equity, because the risk is not financial, it is epistemic. The investor does not know what the loss distribution looks like.

This distinction matters for green technology. For mature renewables—utility-scale solar, onshore wind—the technology risk is low and the premium is small; de-risking instruments are barely needed. For first-of-a-kind green hydrogen, carbon capture, or sustainable aviation fuel, the technology risk dominates and financial subordination does little. The instrument that works in the second case is not equity subordination but completion guarantees and offtake contracts—commitments that the plant will be built and that the output will be bought. These are closer to insurance than to venture capital, which is why the boundary between this chapter and Chapter 8 is porous.

10.2.4 The market-building channel

The third mechanism is the least discussed and probably the most important. Government-guided venture capital can build the demand-side infrastructure that makes the premium disappear—standards, certification, offtake aggregation, and the shared infrastructure that no single firm will finance.

Consider green hydrogen. A single electrolyzer plant faces a premium that is partly the cost of the electrolyzer and partly the cost of the absence of a hydrogen market: no pipelines, no storage, no standardized quality specifications, no insurance products, no trained workforce. A government fund that invests only in the electrolyzer does not solve the market problem. A government fund that invests in the electrolyzer and in the certification regime and in the first storage facility and in the offtake aggregation mechanism is doing something different—it is building an ecosystem.

This is the argument that Mariana Mazzucato makes in The Entrepreneurial State (2013), and it is broadly right, though the book overstates the case by treating all public investment as mission-driven. The empirical support comes from the history of the technologies themselves. Solar photovoltaic was pushed by government procurement in Germany and Japan before it was competitive. Lithium-ion batteries were pushed by government R&D and early electric vehicle procurement. The public sector did not just fund the technology; it created the first market.

The mechanism is not venture capital in the conventional sense. It is closer to what the development literature calls "market creation"—the public provision of the complementary assets and institutions that a new market needs. GVC can be a vehicle for this, but only if the fund's mandate includes the ecosystem investments and not just the company investments. Most government venture funds do not have that mandate, which is a design failure I will return to.

flowchart TD
    A[Green premium at plant gate] --> B[Technology cost c0]
    A --> C[Fossil cost cf]
    A --> D[Carbon price pc]
    B --> E[GVC: learning and scale]
    C --> F[Fossil price volatility]
    D --> G[Carbon pricing and mandates]
    E --> H[Premium declines]
    F --> H
    G --> H
    H --> I[Private capital enters]
    I --> J[Self-sustaining market]
    K[Ecosystem investment: standards, infrastructure, offtake] --> H
    K --> I

Figure 10-1: The transmission from GVC to premium reduction. The direct channel (learning and scale) is slow; the ecosystem channel is faster but requires a mandate most funds lack.

10.2.5 The crowding-out objection

The standard objection to government venture capital is that it crowds out private investment. If the government is willing to fund a deal, private investors will step back and let the public balance sheet take the risk. The result is a smaller private market, not a larger one.

The evidence is genuinely mixed, and the mix depends on the design. Leleux and Surlemont (2003), studying European public venture capital, found evidence of crowding out in later-stage investments. Cumming and MacIntosh (2006), studying Canadian labor-sponsored funds, found similar effects. But Brander, Du, and Hellmann (2015) found crowding in for early-stage deals, and Da Rin, Nicodano, and Sembenelli (2006) found that public venture capital was associated with the creation of active venture markets in Europe, particularly when the public investment was small relative to the private market.

The reconciliation is straightforward. Crowding out happens when the government competes with private capital on the same terms for the same deals. Crowding in happens when the government takes a position that private capital cannot or will not take—earlier stage, higher technology risk, longer horizon—and thereby expands the set of investable deals. The design question is whether the fund's mandate forces it into the second mode. A fund that is judged on financial returns will drift into the first. A fund that is judged on the additionality of its investments—the deals it makes that would not have happened otherwise—can stay in the second.

This is not a new insight. It is the core of the "additionality" criterion that the European Commission has applied to state aid since the 2000s. The problem is measurement. Additionality is counterfactual, and counterfactuals are hard to observe. I will take up the evaluation problem in Section 10.3.

10.3 Empirical Evidence and Case Studies

10.3.1 What the evaluations can and cannot tell us

The evaluation literature on government venture capital is large and methodologically uneven. Most studies use one of three designs: comparisons of recipient and non-recipient firms matched on observables, instrumental variables that exploit eligibility thresholds or program rollouts, or natural experiments from program discontinuities. Each design has a characteristic failure mode.

Matching studies suffer from selection bias: the government fund selects firms that are already more likely to succeed, so the measured effect of the investment is confounded with the selection. Instrumental variable studies require an instrument that affects selection but not outcomes, which is hard to find and harder to defend. Natural experiments are the gold standard but rare, because government funds are rarely rolled out in a way that creates a clean comparison group.

The result is that the literature can tell us whether recipients did better, but it struggles to tell us whether they did better because of the investment. For green technology, the problem is compounded by the fact that the outcome of interest—premium reduction—is a sector-level variable, not a firm-level one. A fund that invests in ten electrolyzer companies and produces one commercial success may have reduced the premium even if nine of its investments failed. Firm-level evaluations will call that a failure; sector-level evaluations may call it a success.

I want to be candid about this. The evidence base for GVC in green technology is thinner than the policy confidence behind it. The strongest evidence comes from a handful of sectors—solar, batteries, and to a lesser extent wind—where the learning curves are well measured and the public role is well documented. For hydrogen, carbon capture, and sustainable aviation fuel, the evidence is mostly prospective.

10.3.2 The solar case: learning curves and public procurement

The solar photovoltaic case is the best-documented instance of a public push that reduced a green premium. The story is familiar but worth retelling with the numbers.

In 2000, the cost of a solar module was roughly $5 per watt. By 2020, it was roughly $0.20 per watt. The learning rate—the percentage cost decline per doubling of cumulative production—was about 20 percent over the period. No single policy produced this; the decline was driven by a combination of manufacturing scale in China, R&D in Germany, Japan, and the United States, and procurement subsidies that created the first markets.

The German feed-in tariff, established under the Renewable Energy Sources Act (EEG) of 2000, guaranteed above-market prices for solar electricity. At its peak, the tariff was roughly four times the wholesale electricity price. That is a green premium of about 300 percent, paid by electricity consumers. The tariff was phased down over two decades as the premium fell, and by the time it was substantially reduced in the 2010s, solar was approaching grid parity in sunny regions.

The Chinese role came later and was decisive for manufacturing scale. The government's Golden Sun program (2009) and subsequent provincial programs subsidized domestic installations, and the resulting demand supported a manufacturing base that drove down module costs globally. By some estimates, Chinese manufacturing scale accounted for more than half of the global cost decline between 2010 and 2020.

What does this tell us about GVC specifically? Less than the general story suggests. The solar cost decline was driven primarily by procurement subsidies and manufacturing scale, not by venture capital. Venture capital played a role in the early technology companies—Solyndra, the notorious failure, was a venture-backed company—but the cost decline came from the manufacturing side, which was financed by debt and equity in China, not by government venture funds in the conventional sense.

The lesson is that GVC is one instrument among several, and in solar it was not the binding one. The binding instruments were the feed-in tariff (which created demand) and the manufacturing scale-up (which reduced cost). A government venture fund that had invested in solar module companies in 2005 would have made money, but it would not have been the reason the premium fell.

10.3.3 The battery case: the limits of venture capital

Lithium-ion batteries tell a different story, and a more cautionary one.

The technology was invented in the 1970s and 1980s, with major contributions from Exxon, Bell Labs, and later Sony. The first commercial lithium-ion battery was Sony's, in 1991. The cost decline since then has been steady but slower than solar: roughly 10 to 15 percent per doubling of production, depending on the chemistry and the period.

Government support came through R&D funding (the US Advanced Battery Consortium, founded in 1991) and, later, through electric vehicle procurement and manufacturing subsidies. The US Department of Energy's battery R&D programs are estimated to have contributed a meaningful share of the early cost reductions, though the attribution is contested.

Venture capital's role was concentrated in the 2000s and early 2010s, when a wave of battery start-ups—A123 Systems, Ener1, and others—raised substantial private and public money. Most of them failed. A123 Systems, which received a $249 million DOE grant and raised over $1 billion in private and public capital, filed for bankruptcy in 2012. The failure was not a failure of the technology; it was a failure of manufacturing scale-up in the face of Chinese cost competition.

The battery case illustrates a hard limit on GVC. Venture capital is good at funding technology development and bad at funding capital-intensive manufacturing scale-up. The scale-up requires debt, strategic partnerships, and patient capital with a longer horizon than a ten-year venture fund. A government venture fund that tries to do manufacturing scale-up through equity is using the wrong instrument.

This is the "valley of death" problem that the science and technology finance literature has documented for decades. The valley is between the demonstration plant and the commercial-scale plant, and it is a financing gap, not a technology gap. GVC can bridge the earlier gap—between the lab and the demonstration—but the later gap requires different instruments: loan guarantees, offtake contracts, and the kind of risk-sharing that Chapter 8 described for insurance.

10.3.4 The hydrogen case: a prospective assessment

Green hydrogen is the current test case for GVC, and the evidence is still being generated.

The green premium for hydrogen is large and uncertain. Estimates for 2023 put the cost of green hydrogen at $4 to $7 per kilogram in most markets, against $1 to $2 for grey hydrogen produced from natural gas. That is a premium of 100 to 400 percent, depending on the price of gas and the capacity factor of the electrolyzer. The premium is dominated by the cost of electricity and the capital cost of the electrolyzer.

Government-guided venture capital has been active in the sector. The European Union's Innovation Fund has committed billions to hydrogen demonstration projects. The US Department of Energy's Hydrogen Shot program aims to reduce the cost of clean hydrogen to $1 per kilogram by 2031. China's provincial guidance funds have invested in electrolyzer manufacturers and hydrogen production projects.

The early evidence is mixed. Electrolyzer costs have fallen, but slower than the roadmaps projected. The learning rate for electrolyzers is estimated at 10 to 15 percent per doubling, well below solar. The manufacturing base is small and fragmented. The demand side is thin: there are few offtake agreements at scale, and the industries that would use green hydrogen—steel, chemicals, refining—have been slow to commit.

This is where the interaction between GVC and the green premium becomes visible. A government fund that invests in an electrolyzer manufacturer is betting that the manufacturer can reduce costs fast enough to close the premium. But the manufacturer's cost reduction depends on scale, and scale depends on demand, and demand depends on the premium closing. The circularity is the problem. GVC alone cannot break it; it needs a demand-side commitment—a carbon contract for difference, a procurement mandate, a hydrogen blending requirement—to create the first market.

The EU's Carbon Contract for Difference (CCfD) mechanism, piloted through the Innovation Fund, is an attempt to break the circularity. The contract pays the difference between the strike price of green hydrogen and the market price of grey hydrogen, guaranteeing the producer a fixed price for a defined period. The producer can then finance the plant on the strength of the contract. The mechanism is closer to the insurance instruments of Chapter 8 than to venture capital, which is the point: the instrument that closes the premium is not always equity.

10.3.5 A note on the Chinese provincial guidance funds

The Chinese experience deserves separate treatment because of its scale and its distinct governance problems.

Provincial and municipal guidance funds (yindao jijin, 引导基金) proliferated after 2015, when the central government encouraged local governments to use fiscal resources to leverage private capital. By some estimates, there were more than 1,500 such funds by 2020, with committed capital in the trillions of renminbi. Many have green technology mandates.

The evaluation literature on these funds is limited and mostly Chinese-language. The findings that have emerged are consistent with the international evidence: the funds that operate with professional management and clear additionality criteria have positive effects; the funds that operate as extensions of local industrial policy, with investment decisions driven by political targets, have weak or negative effects. A recurring problem is "government guidance" that becomes government direction—the fund is used to bail out local champions or to hit GDP targets, and the signaling value collapses.

The governance lesson is that the distance between the fund and the political authority matters. Funds with independent investment committees, professional staff, and published additionality criteria perform better than funds that report directly to a provincial development commission. This is not a surprising finding, but it is an important one for the design of green GVC, because the political pressure to use green funds for industrial policy is strong.

10.4 The Interaction Between GVC and the Green Premium

10.4.1 The circularity problem

I have referred to the circularity several times; it deserves a formal statement.

A green technology's cost depends on its scale. Its scale depends on demand. Its demand depends on its cost relative to the fossil incumbent—that is, on the premium. The premium depends on the cost. The loop is closed.

where \(Q\) is cumulative production. If \(c_0\) declines with \(Q\) at rate \(\lambda\) (the learning rate), and demand responds to the premium with elasticity \(\eta\), the system has a stable low-scale equilibrium (high premium, low demand) and a potentially stable high-scale equilibrium (low premium, high demand). The transition between them requires a shock—a subsidy, a mandate, a procurement commitment—large enough to push the system over the hump.

GVC is one such shock, but it is a small one. A venture fund investing in a single company moves \(Q\) by a small amount. The shock that matters is the one that moves the system, and that requires coordination across many investors and a demand-side commitment. This is why the most successful green industrial policies have combined supply-side support (R&D, demonstration, manufacturing subsidies) with demand-side support (procurement, mandates, carbon pricing). The supply-side support alone does not break the circularity.

10.4.2 When GVC is the right instrument

The circularity argument suggests a division of labor. GVC is the right instrument when:

The technology is at an early stage, where the risk is epistemic and private capital cannot price it. This is the classic venture capital case, and it is where the certification and de-risking channels work best.

The investment is in the innovation rather than the scale-up. GVC can fund the demonstration plant, the pilot line, the first commercial unit. It cannot fund the tenth plant; that requires debt and strategic capital.

The fund has the technical capacity to diligence the technology. A government fund without in-house engineering expertise is a fund that will be led by its co-investors or by political priorities. The certification signal is only credible if the fund knows what it is certifying.

The fund's mandate includes ecosystem investments. Standards, testing facilities, shared infrastructure, and workforce development are all part of the premium, and a fund that invests only in companies will miss them.

10.4.3 When GVC is the wrong instrument

Conversely, GVC is the wrong instrument when:

The technology is commercially viable and private capital is available. Investing in a utility-scale solar farm or an onshore wind project is not GVC; it is asset management. The additionality is zero, and the crowding-out risk is high.

The binding constraint is manufacturing scale-up. As the battery case showed, equity is a poor instrument for capital-intensive manufacturing. Loan guarantees, offtake contracts, and strategic partnerships are better.

The binding constraint is demand. If the premium is high because no one will buy the output, the instrument is a contract for difference or a procurement mandate, not an equity investment in the producer.

The fund lacks the governance to resist political direction. A fund that is used to hit industrial targets will destroy its signaling value and misallocate capital. Better to not have the fund than to have a fund that operates this way.

10.4.4 The green premium as a portfolio signal

There is a use of the premium that I have not yet discussed: as a portfolio construction tool.

A government fund with a mandate to reduce the green premium can use the premium estimates to allocate capital across sectors. The logic is to invest where the premium is high but the learning rate is also high—where a dollar of public capital buys the most premium reduction per unit of time. This is a different criterion from the conventional venture capital criterion of expected financial return, and it is the right criterion for a public fund.

The metric is the expected decline in the premium per dollar invested, or more precisely, the marginal reduction in \(\Delta\) per dollar of public capital, accounting for the probability of success and the time to impact. Formally, for a portfolio of investments \(i\),

where \(p_i\) is the probability of success, \(\partial \Delta_i / \partial K_i\) is the marginal premium reduction per dollar, and \(K_i\) is the public capital allocated. The fund should allocate to the investments with the highest expected impact per dollar, subject to a portfolio risk constraint.

This is easier to state than to implement. The marginal premium reduction is hard to estimate ex ante, and the probability of success is exactly what the fund is trying to learn. But the framework is useful because it forces the fund to state its theory of change: how will this investment reduce the premium, and by how much? A fund that cannot answer that question for a given deal is a fund that is investing for reasons other than the premium.

10.5 Design Principles for Green GVC

I want to close with a set of design principles, drawn from the evidence and the theory. These are provisional—the evidence base is thin, and the principles should be revised as it grows—but they are the best I can offer.

10.5.1 Separate the innovation fund from the scale-up fund

The first principle is institutional. The skills, the risk tolerance, and the time horizon required for early-stage innovation are different from those required for manufacturing scale-up. A single fund that tries to do both will do one of them badly. The innovation fund should be small, technically staffed, and tolerant of failure. The scale-up fund should be large, financially staffed, and structured as a co-investment with strategic partners.

The Chinese guidance fund model has often conflated the two, with the result that funds drift toward later-stage deals (where the financial returns are safer) and the early-stage gap remains. The separation is not a panacea, but it clarifies the mandate and makes evaluation possible.

10.5.2 Pay for additionality, not for returns

The second principle is about compensation. A government fund that is judged on its financial returns will behave like a private fund, which means it will avoid the deals where it is most needed. The fund's managers should be compensated on additionality—the deals they make that would not have happened otherwise—and on the premium reduction their portfolio achieves, not on the internal rate of return.

Measuring additionality is hard. The practical approach is to require each investment to be accompanied by a written additionality assessment: what would have happened without the public capital, and why? The assessment is a prediction, and it can be checked ex post. Over time, the fund builds a track record that can be evaluated.

10.5.3 Pair equity with demand-side commitments

The third principle is about instrument design. Equity alone does not close the premium, because the premium is a demand-side problem as much as a supply-side one. Every GVC investment in a first-of-a-kind technology should be paired with a demand-side commitment—an offtake agreement, a contract for difference, a procurement mandate—that gives the producer a market.

The pairing is what distinguishes a serious green industrial policy from a subsidy. A subsidy without a market is a grant; a market commitment without a subsidy is a mandate that may not be met. The two together create the conditions for private capital to enter.

10.5.4 Build the ecosystem, not just the company

The fourth principle is about scope. The premium includes the cost of the missing ecosystem—the standards, the infrastructure, the workforce. A fund that invests only in companies will leave the ecosystem gap unaddressed, and the companies will fail for reasons that have nothing to do with their technology.

The practical implication is that the fund's mandate should include ecosystem investments: testing facilities, certification bodies, shared infrastructure, and training programs. These investments have low financial returns and high social returns, which is exactly why they need public capital.

10.5.5 Accept failure, and measure it

The fifth principle is about culture. Early-stage green technology investing will produce failures. The failures are information: they tell you which pathways do not work, and they reduce the cost of the next attempt. A fund that punishes failure will avoid risk, and a fund that avoids risk will not do the job it was created for.

The measurement implication is that the fund should report its failure rate and its learning, not just its successes. A portfolio with a 30 percent failure rate and a clear record of what was learned is a better portfolio than one with a 5 percent failure rate and no innovation.

10.6 What the Chapter Does Not Claim

I want to be explicit about the limits of the argument.

It does not claim that GVC is the main instrument for closing the green premium. The evidence suggests that carbon pricing, procurement, and manufacturing scale are larger levers. GVC is a complement, most useful at the early stage where the other instruments are weak.

It does not claim that the empirical literature supports a strong causal effect. The literature is mixed, the designs are imperfect, and the sector-level outcomes are hard to attribute. The strongest claim I can defend is that GVC can work under specific design conditions, and that the conditions are identifiable.

It does not claim that the Chinese guidance fund model is a success. The scale is impressive; the governance is uneven; the evaluation is incomplete. The model is an experiment, and it is too early to declare it a failure or a success.

It does not claim that the green premium is a sufficient statistic for policy. The premium is one input among many, and it is measured with error. Policy that targets the premium without accounting for the measurement problems will misallocate capital.

10.7 The Unresolved Question

The chapter opened with the tension between demonstrated technology and bankable demand. The analysis has sharpened the tension but not resolved it. Let me state the unresolved question plainly, because it is the hinge on which the rest of this book turns.

A government fund can finance the plant that proves the chemistry. ItIt can even, with enough subordination and enough patience, finance the plant that proves the cost curve. What it cannot do on its own is prove that a buyer will pay the green price when a grey substitute sits on the same shelf at a lower sticker. That proof requires a commitment from the demand side—a utility willing to sign a fifteen-year offtake at a fixed premium, a steel customer willing to pay the green markup on a car, a government willing to write a contract for difference that survives a change of administration. None of these commitments is a venture capital instrument. All of them determine whether the venture capital instrument pays off.

The temptation, when a fund faces this gap, is to close it with more equity. Invest in the offtaker as well as the producer. Build the storage facility, the pipeline, the certification body. This is the ecosystem logic of Section 10.2.4, and it is right as far as it goes. But an equity fund that tries to build an entire market is a fund that has abandoned the discipline of the venture model—staged financing, milestone-based commitment, the option to stop. It becomes a development bank with a venture label, and it will be judged by the standards of a development bank, which are political rather than financial.

The alternative is to accept the boundary and design the fund accordingly. The fund does the early-stage work: it takes the technology risk that private capital cannot price, it certifies the technology to the market, and it exits—or writes off—once the technology is demonstrated. The demand-side commitment is a separate instrument, held by a separate institution, with a separate budget and a separate political sponsor. The two instruments are coordinated but not merged. This is the architecture I will argue for in Chapter 14, and it is the reason the "three-pillar" framing of this book treats rating, insurance, and catalysis as distinct mechanisms rather than as a single blended finance vehicle.

There is a cost to this separation. Coordination is harder when the instruments sit in different institutions. The producer who needs both equity and an offtake contract has to negotiate with two counterparties, on two timelines, under two sets of conditions. In practice, the coordination failure is one of the reasons first-of-a-kind projects take so long to reach financial close. The EU Innovation Fund's grant process, which covers up to 60 percent of the incremental capital cost of a demonstration plant, is a case in point: the grant is generous, but it does not come with an offtake, and projects that win the grant still spend years looking for a buyer. The grant de-risks the capital cost; it does nothing for the revenue risk. That is the gap the contract for difference is meant to fill, and the fact that the two instruments are administered by different parts of the Commission is a friction that shows up in project timelines.

I do not have a clean solution to this coordination problem. The best I can offer is a design principle: co-locate the decision even if you separate the balance sheet. A single investment committee that sees both the equity case and the offtake case, and that can condition one on the other, will coordinate better than two committees that never meet. The balance sheets can remain separate—the equity fund bears technology risk, the offtake facility bears volume and price risk—but the decision to proceed should be joint. This is closer to how project finance works than to how venture capital works, which is appropriate, because the problem at this stage is a project finance problem.

10.8 The Green Premium and the Cost of Capital

One more mechanism deserves attention before I close, because it is often overlooked in the venture capital literature and it matters enormously for the transition.

The green premium is usually measured as a cost gap at the plant gate. But the cost a project actually faces includes the cost of capital, and the cost of capital for green projects is systematically higher than for comparable fossil projects—not because the technology is riskier in an engineering sense, but because the revenue is less certain, the policy support is less durable, and the lenders have less experience with the asset class. A green steel plant may face a cost of debt two to three hundred basis points above a blast furnace, and that spread can add 10 to 20 percent to the levelized cost of the output. In some sectors, the cost-of-capital premium is larger than the technology premium.

This changes the calculus for GVC. A government fund that takes an early equity position can reduce the cost of capital for subsequent investors in two ways. It can absorb first-loss risk, which lowers the required return for senior capital. And it can lengthen the tenor of the capital stack, which matters because green projects often have long payback periods and private equity has a ten-year horizon. A government fund with a fifteen- or twenty-year horizon can hold the asset through the period when the premium is closing, and exit when the asset is bankable. That is a genuine additionality that a private fund cannot replicate.

The empirical support for this is thinner than I would like. The cost-of-capital premium for green projects is documented in the renewable energy literature—Hirth and Steckel (2016) is the standard reference for the effect of revenue risk on the cost of capital for wind and solar—but the evidence for industrial decarbonization projects is mostly anecdotal. What I can say is that the mechanism is plausible, the magnitudes are large enough to matter, and the policy implication is clear: a government fund that holds a long-duration position in a first-of-a-kind project is doing something that a short-duration private fund cannot do, and that something is valuable.

There is a boundary condition. The long-duration position only helps if the fund can actually hold. A government fund that is subject to annual budget reviews, or that is expected to return capital within a political cycle, will be forced to exit early, and the duration benefit will evaporate. The governance of the fund—its capital structure, its reporting lines, its evaluation horizon—determines whether it can play the role the theory assigns it. This is the subject of the next chapter, and it is the reason I have treated governance as a first-order design variable rather than an afterthought.

10.9 Summary of the Argument

Let me consolidate what the chapter has established.

The green premium is the cost gap between a zero-carbon substitute and the fossil incumbent, measured per unit of output. It is a range, not a number, and it moves with fossil prices, learning rates, and system build-out. It is the right target for public policy because it is the variable that determines whether private capital will invest without support.

Government-guided venture capital can act on the premium through three channels: signaling (certification of technology quality), de-risking (subordination and guarantees that change the risk-return profile), and market-building (ecosystem investments that reduce the system-level premium). The signaling and de-risking channels work best at the early stage, where the risk is epistemic. The market-building channel works best when the fund's mandate includes ecosystem investments, which most funds lack.

The empirical evidence is mixed and thinner than the policy confidence behind it. The solar case shows that public support can reduce a premium, but the binding instruments were procurement and manufacturing scale, not venture capital. The battery case shows the limits of equity for manufacturing scale-up. The hydrogen case is still being written, and the early evidence suggests that the circularity between cost, scale, and demand cannot be broken by equity alone.

The design principles that follow from this are: separate the innovation fund from the scale-up fund; pay for additionality rather than returns; pair equity with demand-side commitments; build the ecosystem, not just the company; accept failure and measure it. The unresolved question is the coordination between the equity instrument and the demand-side instrument, and the best I can offer is to co-locate the decision while separating the balance sheet.

The cost-of-capital premium is a second, often overlooked channel. A government fund with a long horizon can reduce the cost of capital for subsequent investors by absorbing first-loss risk and holding duration. Whether it can do so depends on its governance, which is where the next chapter begins.

10.10 A Note on Method and Evidence

I have drawn in this chapter on three kinds of evidence: the sectoral premium estimates from CICC Research and comparable sources; the evaluation literature on government venture capital, which is mostly about non-green sectors but whose design lessons transfer; and a set of case studies—solar, batteries, hydrogen—that are chosen for their illustrative value rather than their statistical representativeness.

The case studies have a selection problem. I have chosen cases where the public role is visible and the cost data are available. There are surely cases where GVC failed quietly and no one wrote a paper about it. The survivorship bias runs in the direction of overstating the effectiveness of the instrument, which is why I have been at pains to state the boundary conditions and the null results.

The premium estimates have a measurement problem that I have flagged but not solved. They are point estimates for a particular year and region, and they are sensitive to assumptions about fossil prices and grid emission factors. A reader who wants to use these numbers for policy should treat them as order-of-magnitude guides, not as precise targets.

The evaluation literature has an identification problem that I have described but not resolved. The best studies use natural experiments, and natural experiments in this area are rare. The result is that the causal effect of GVC on green innovation is known with less precision than the policy debate assumes. I have tried to be honest about this, and I would rather state an uncertain conclusion than a confident one that the evidence does not support.

What I am confident about is the direction of the mechanisms. The signaling channel is real when the signal is credible. The de-risking channel is real when the risk is financial rather than epistemic. The market-building channel is real when the fund has the mandate to pursue it. The cost-of-capital channel is real when the fund has the duration to hold. These are conditional claims, and the conditions are the substance of the design problem.

The next chapter takes up the governance question directly. A fund with the right mandate and the wrong incentives will fail; a fund with the right incentives and the wrong evaluation will drift. The design of performance evaluation and incentive contracts for policy-guided green finance is not a technical afterthought. It is the mechanism that determines whether the instruments described in this chapter do what they are meant to do.