Chapter 9: Sci-Tech Finance and the Catalysis of Green Innovation

The first two pillars of this book deal with identification and buffering. A credit rating that reads environmental performance tells a lender which technology firm is worth backing; an insurance contract that pays out on a rainfall deficit keeps a forestry cooperative solvent through a bad season. Neither instrument, on its own, creates the thing being financed. A rating does not invent a solid-state electrolyte. A weather index does not design a heat pump. The third pillar—sci-tech finance—is where the financial system stops grading and buffering existing activity and starts pulling new green technology into existence. That is a different job, with a different failure mode, and it deserves its own treatment.

I want to be precise about what "catalysis" means here, because the word is doing real work. A catalyst lowers the activation energy of a reaction without being consumed by it. Sci-tech finance, at its best, does something analogous: it lowers the cost and risk of moving a green technology from laboratory demonstration to commercial deployment, and it does so through instruments structured to absorb the early-stage risk that ordinary credit cannot. The analogy has limits. Financial catalysts can be consumed—a government fund that loses its capital is gone—and they can distort the reaction they are meant to accelerate. Both limits matter, and I will return to them.

9.1 Market Failure in Green Innovation Financing

The International Energy Agency's tracking of clean-energy investment puts annual global spending in the vicinity of $1.8 trillion as of the mid-2020s, against an estimated requirement of roughly $4.5 trillion per year by the early 2030s to stay on a net-zero trajectory. The gap is not uniform across technologies. Mature renewables—utility-scale solar, onshore wind—attract private capital readily because their cost curves have flattened and their revenue is contractible. The gap concentrates in what the climate-finance literature calls "hard-to-abate" and "early-stage" categories: green hydrogen electrolyzers, long-duration storage, cement and steel decarbonization, sustainable aviation fuel, carbon capture at scale. These are precisely the technologies where the innovation is not finished.

Why does private capital not close this gap on its own? The standard answer invokes two market failures, and I want to separate them because they call for different instruments. The first is the public-goods character of knowledge. A green R&D program generates two outputs: the private return to the firm that funds it, and a spillover—the knowledge, the trained engineers, the failed experiments that save others from repeating them—that accrues to society. The firm captures the first and not the second. When the gap between private and social return is large, private investment falls below the social optimum. This is the classic argument for public R&D subsidy, and it applies to green technology with unusual force because the social return includes avoided climate damages that no individual firm can monetize.

The second failure is more specific to green technology and less often named precisely. Call it the deployment-risk wedge. Even when a green technology is technically proven, the first commercial-scale plant carries risks that a tenth plant does not: construction cost overruns on an unproven process, performance shortfalls, the absence of a supply chain for specialized components, and—critically—the absence of a track record that a lender or insurer can price. A steel mill considering a hydrogen-based direct-reduction route faces a capital cost per ton that is two to three times that of a conventional blast furnace, and the incremental cost is recoverable only if the green premium for low-carbon steel holds. The first mover bears the learning cost; the followers free-ride. This is a coordination failure as much as an information failure, and it explains why green innovation financing is not simply a matter of making capital cheaper.

Three frictions stand out when I look at how these failures actually block finance in practice. The first is the maturity mismatch. Green technology development runs on a ten-to-fifteen-year horizon from lab to bankable asset. Venture capital funds have seven-to-ten-year lives. Bank credit is shorter still, and secured. The result is a gap in the middle of the development cycle—what practitioners call the "valley of death"—where the technology is too mature for venture capital's risk appetite and too immature for project finance. The second friction is the collateral problem. Banks lend against assets. A green start-up's principal asset is intellectual property and a team, neither of which a credit officer can seize and sell at a predictable price. The third is the information asymmetry that runs through the whole book: the entrepreneur knows more about the technology's prospects than the investor, and the gap is wider for novel technologies because there is no comparable cohort to benchmark against.

It is worth pausing on the maturity mismatch, because it is where the most policy energy has been spent and where the results are most mixed. The valley of death is not a metaphor; it is a specific segment of the financing curve. Figure 9.1 lays out the stylized funding sequence and where the gap opens.

flowchart LR
    A[Basic research<br/>TRL 1-3] --> B[Applied R&D<br/>TRL 4-5]
    B --> C[Pilot & demonstration<br/>TRL 6-7]
    C --> D[First commercial<br/>TRL 8]
    D --> E[Diffusion<br/>TRL 9]
    A -.->|Public grants,<br/>university funds| A
    B -.->|Venture capital,<br/>angel investors| B
    C -.->|GAP: valley of death| C
    D -.->|Project finance,<br/>green bonds| D
    E -.->|Bank credit,<br/>corporate bonds| E

Figure 9.1: The green innovation funding sequence and the valley of death at the pilot-to-commercial transition (TRL = technology readiness level).

The diagram is stylized, and the boundaries are porous—some venture capital does reach into demonstration, some project finance does back first-of-a-kind plants. But the shape is right, and it carries a policy implication that is often missed. The valley of death is not a single problem. It is at least three problems wearing the same coat. There is a risk-capital problem (no equity willing to absorb demonstration-stage losses), a risk-pricing problem (no actuarial base for insuring first-of-a-kind performance), and a risk-sharing problem (no mechanism to distribute the learning cost across the firms that will benefit from it). Each calls for a different instrument. Conflating them is why so many "green finance" programs amount to a subsidy that fills one hole while the others stay open.

There is a fourth friction that deserves separate mention because it is specific to the green transition and does not appear in generic innovation-finance accounts. Green technologies face a demand-side uncertainty that conventional technologies do not. A firm that develops a better semiconductor can rely on a global market that exists. A firm that develops a low-carbon cement process relies on a market that is being created by policy—carbon pricing, public procurement rules, building standards—and the policy can change. This policy risk is not diversifiable in the way technology risk is, because it is correlated across the entire sector. When a government signals that it may weaken its carbon price, every green-technology business case in the country weakens at once. Investors price this correlation as a sector-level discount, which raises the cost of capital for all green innovation regardless of individual quality.

I have seen this play out in the Chinese context, where the policy commitment is strong but the instrument mix has shifted repeatedly. The 2030 peak and 2060 neutrality targets, announced in 2020, gave the sector a long-horizon signal. But the specific support mechanisms—feed-in tariffs, subsidy schedules, green certificate markets—have been revised multiple times, and each revision resets the demand-side uncertainty for anyone whose business case depends on them. Du Xiangwan, honorary director of China's National Expert Committee on Climate Change, has argued that the dual-carbon goal requires the financial system to make a contribution that is "important and irreplaceable"—his phrase—but the contribution only materializes if the policy signal is stable enough for capital to price. A target is not an instrument.

This is the point at which the argument connects back to the first two pillars. The information asymmetry that blocks green innovation finance is not different in kind from the asymmetry that blocks green credit generally; it is more severe because the technology is novel. A rating framework that can read environmental performance—the subject of Chapters 3 and 4—also generates the comparable data that lets an investor distinguish a good green start-up from a bad one. And the insurance instruments of Chapters 6 through 8 provide a template for the risk-sharing that the valley of death requires: an entity willing to absorb a defined risk in exchange for a premium, which converts an unpriced uncertainty into a priced cost. Sci-tech finance does not replace these pillars. It uses them.

Market-failure reasoning justifies public intervention in principle; it does not specify the form. A government that decides to fill the valley of death can do so with grants, with equity, with guarantees, with tax credits, or with procurement, and the choice matters enormously for the outcome. The rest of this chapter is about that choice.

9.2 Instruments of Sci-Tech Finance

The instruments of sci-tech finance are not new. Venture capital, government-guided funds, and technology insurance all predate the green transition by decades. What is new is their application to green technology, and the adaptations that application has required. I will treat each in turn, with attention to the friction it addresses and the failure mode it carries.

9.2.1 Venture Capital and the Risk-Appetite Problem

Venture capital is the instrument best suited to the early-stage end of the valley of death, because its structure—a fund that holds a portfolio of high-risk bets and returns capital through exits—is designed to absorb the losses that come with technological uncertainty. A venture fund expects most of its portfolio to fail and a small number to return many times the investment. That distribution is exactly what green technology development produces, at least in principle.

In practice, green venture capital has underperformed the broader venture market on the dimensions that matter for the transition. The reasons are instructive. First, the exit environment is weaker. A software start-up can reach an acquisition or IPO in five to seven years; a green-technology start-up often needs a decade or more to reach a commercial exit, because the technology must be demonstrated at scale and the market must mature. Longer holding periods reduce the internal rate of return for a given multiple, which makes green funds less attractive to limited partners who benchmark against conventional venture. Second, the capital intensity is higher. A software company can scale on servers; a green-hydrogen start-up needs an electrolyzer plant. The capital required per unit of technological progress is an order of magnitude larger, which means venture capital—which typically writes checks sized for software—can fund the early stage but not the demonstration. Third, the policy dependence described above introduces a risk that venture investors cannot hedge and do not have a comparative advantage in bearing.

The result is a structural gap: green venture capital is concentrated in asset-light segments—software for energy management, marketplace platforms, materials discovery—and thin in the hardware-intensive segments where the emission reductions actually are. This is not a failure of venture capital as such. It is a mismatch between the instrument's return profile and the technology's capital profile. The instrument is doing what it is designed to do; the design is wrong for the job.

The adaptation that has emerged is the corporate venture arm and the strategic investor. An industrial firm—a steelmaker, a utility, an automaker—can invest in a green-technology start-up with a return horizon that matches the technology, because the strategic return includes the option to deploy the technology in its own operations. This changes the calculus. The corporate investor does not need a financial exit in seven years; it needs the technology to work, and it can provide the demonstration site that the start-up cannot afford. I have seen this model work in the electrolyzer space, where established industrial-gas firms have taken minority stakes in electrolyzer start-ups and provided both the balance sheet and the first customer. It is not a substitute for a functioning venture market, but it fills part of the gap that pure financial venture leaves open.

9.2.2 Government-Guided Funds and the Risk-Sharing Problem

Government-guided funds are the instrument most directly aimed at the valley of death, and the one whose design varies most across jurisdictions. The logic is straightforward: if private capital will not bear the demonstration-stage risk because the social return exceeds the private return, the government can bear part of that risk on behalf of the public. The question is how.

There are three broad designs, and they differ in how they allocate the risk between the public and private partners. The first is the direct equity fund: the government capitalizes a fund that invests in green technology, either alone or alongside private investors. The second is the matching or co-investment fund: the government commits capital on the condition that private investors commit alongside, typically at a ratio, so that the public money leverages private money and—crucially—imports private due diligence. The third is the risk-sharing or first-loss guarantee: the government absorbs the first tranche of losses on a portfolio, which changes the risk-return profile for private investors without requiring the government to pick winners directly.

Each design has a characteristic failure mode. Direct equity funds are vulnerable to political capture: the investment decisions become subject to non-financial objectives, and the fund becomes a vehicle for industrial policy rather than a catalyst. Matching funds are vulnerable to crowding out: if the government is willing to co-invest, private investors may simply wait for the government to move first, and the leverage ratio in practice is lower than in design. First-loss guarantees are vulnerable to moral hazard: if the government absorbs the first losses, the private investor has less incentive to monitor the investment, and the losses can exceed what the guarantee was priced for.

The Chinese experience with government-guided funds is the largest natural experiment in this design space, and it is worth examining closely because it exposes the trade-offs. China's government guidance funds—a category that includes both industrial funds and the newer "mother funds" that invest in sub-funds—grew rapidly in the 2010s. My own compilation from provincial fund registries and published fund-of-funds disclosures puts committed capital at its peak above RMB 10 trillion. The model is the matching fund in its most ambitious form: the government provides a portion of the capital and sets policy objectives, and private fund managers raise the rest and make the investment decisions. The intent is to combine public risk-bearing with private selection.

The results have been mixed in a way that is instructive. On the positive side, the funds have channeled capital into sectors—semiconductors, new energy, advanced manufacturing—that private capital alone underfunded, and they have built a venture ecosystem in cities that previously had none. On the negative side, three problems have recurred. The first is the return problem: many funds have struggled to generate financial returns, because the policy objectives and the return objectives pull in different directions, and the fund managers are evaluated on both. The second is the concentration problem: because local governments sponsor funds, the investments tend to cluster geographically, which fragments the market and duplicates capacity. The third is the exit problem: the funds hold illiquid positions in unlisted companies, and the exit channels—IPOs, secondary sales—have not kept pace with the capital deployed, so the funds are sitting on unrealized positions and cannot recycle capital into new investments.

These problems are not unique to China, but the scale makes them visible. The lesson for green technology finance is that a government-guided fund is a powerful instrument for mobilizing capital, and a weak instrument for allocating it, unless the design explicitly separates the two functions. The funds that have worked best are those where the government sets the mandate and the risk-sharing terms but delegates the investment decisions to managers whose compensation is tied to financial returns, and where the exit channels are planned from the outset. The funds that have worked worst are those where the investment decisions are made by committee with multiple objectives and no clear exit path.

The government-guided fund is also the instrument that most directly connects the identification pillar to the catalysis pillar. A fund that invests in green technology needs a way to screen candidates—to distinguish a genuine technological advance from a well-presented one—and that is exactly what the credit rating framework of Chapters 3 and 4 provides, extended to early-stage firms. The funds that have built such screening into their process have done better than those that have not. This is a provisional finding; the data are not yet sufficient for a strong causal claim. But the mechanism is plausible and the pattern is consistent.

9.2.3 Technology Insurance and the Risk-Pricing Problem

Technology insurance is the least developed of the three instruments and, in my view, the most underused. The concept is simple: an insurer underwrites a specific technology risk—the performance of a new electrolyzer, the output of a novel solar cell, the reliability of a storage system—in exchange for a premium, and the insured firm or its lender is protected against the risk that the technology underperforms. This converts an unpriceable uncertainty into a priced cost, which is what a lender needs in order to extend credit.

The logic connects directly to the insurance pillar of this book. Weather index insurance, discussed in Chapter 6, works by defining a measurable trigger—rainfall below a threshold—and paying out when the trigger is met, without requiring the insurer to verify the loss. Technology insurance works on a similar principle: the trigger is a performance metric defined in advance, and the payout is triggered by a measured shortfall. The difference is that the performance metric is technology-specific and must be defined contractually, which is harder than measuring rainfall.

The market for technology insurance is small, and it is concentrated in a few areas: performance warranties for solar panels, which are typically provided by the manufacturer rather than an insurer; construction all-risk policies that cover installation; and a nascent market for "performance insurance" on novel technologies, offered by specialist underwriters. The constraint is actuarial: an insurer cannot price a risk that has no loss history. For a first-of-a-kind technology, there is no loss history, so the insurer either declines or charges a premium that is prohibitive. This is the same problem that the valley of death presents to lenders, restated in insurance terms.

The way around it is public risk-bearing at the reinsurance layer. A government can provide a backstop—a guarantee that absorbs losses above a threshold—which allows the primary insurer to write the policy without bearing the full tail risk. This is the structure that has been used for terrorism insurance and for pandemic business-interruption insurance, and it is the structure that a technology-insurance market would need. The public backstop is not a subsidy in the ordinary sense; it is a way of pricing a risk that the market cannot yet price, and it can be priced to be actuarially fair over time as the loss history accumulates.

There is a second, less obvious role for insurance in green technology finance that deserves mention. An insurance policy is a contract that transfers a defined risk, and the existence of the contract changes the borrower's credit profile. A green-technology firm with a performance insurance policy is a better credit risk than the same firm without one, because the policy caps the downside. This is the linkage that Chapter 8 develops for agricultural insurance, and it applies with equal force to technology insurance. A lender that would not extend credit to an unproven technology firm might extend it to the same firm with a performance guarantee, at a lower rate. The insurance does not make the technology work; it makes the risk financeable.

Figure 9.2 lays out how the three instruments map onto the frictions identified in Section 9.1.

flowchart TD
    F1[Maturity mismatch<br/>valley of death] --> I1[Government-guided funds<br/>risk-sharing at demonstration]
    F2[Collateral problem<br/>intangible assets] --> I2[Venture capital +<br/>corporate strategic investors]
    F3[Information asymmetry<br/>novel technology] --> I3[Rating-based screening<br/>Chapters 3-4]
    F4[Performance uncertainty<br/>first-of-a-kind] --> I4[Technology insurance +<br/>public reinsurance backstop]
    I1 --> O[Bankable green<br/>technology project]
    I2 --> O
    I3 --> O
    I4 --> O

Figure 9.2: Mapping sci-tech finance instruments onto the four frictions in green innovation financing.

The diagram makes a point that the prose can only gesture at: no single instrument covers the whole valley. Government funds address the maturity mismatch and the risk-sharing problem; venture capital and strategic investors address the collateral problem by providing equity that does not need collateral; rating-based screening addresses the information asymmetry; technology insurance addresses the performance uncertainty. A functioning sci-tech finance system uses all four, and the integrated mechanism this book proposes is precisely the coordination of these instruments with the credit and insurance pillars.

9.3 Case Studies of Green Tech Financing

The abstractions of the previous sections become concrete only when they are attached to a specific firm, a specific technology, and a specific financing structure. I have chosen three cases, each of which illustrates a different instrument and a different segment of the valley of death. They are drawn from different jurisdictions and different technology domains, and they are chosen not because they are typical—no case is—but because they expose the seams of the model.

9.3.1 The Electrolyzer Start-Up: Venture Capital and the Strategic Investor

Consider a green-hydrogen electrolyzer start-up, of the kind that has proliferated in Europe and China since 2020. The technology is a proton-exchange-membrane (PEM) stack, which splits water into hydrogen and oxygen using electricity. The core science is mature; the challenge is cost, durability, and manufacturing scale. A start-up in this space typically raises a seed round from venture capital to build a laboratory prototype, a Series A to build a pilot line, and then faces the valley of death: a commercial-scale manufacturing plant costs hundreds of millions, far more than a venture round can provide, and the plant is not bankable without a customer, and the customer will not commit without a demonstrated plant.

The financing structure that has worked, in the cases I have examined, is a three-party arrangement. The venture investors provide the equity for the pilot line. An industrial partner—typically an industrial-gas company, a utility, or a chemical firm—provides an offtake agreement for the hydrogen or the electrolyzer, which converts the technology risk into a contracted revenue stream. And a government instrument—a grant, a co-investment fund, or a first-loss guarantee—absorbs the residual risk that neither the venture investors nor the industrial partner will bear. The three parties are not interchangeable; each bears a different slice of the risk, and the structure works because the slices are matched to the parties' risk appetites.

The case exposes a general principle. Green-technology finance is not a single transaction; it is a stack of transactions, each with a different risk profile, and the art is in matching the risk to the instrument. A venture investor cannot bear construction risk; a bank cannot bear technology risk; a government cannot efficiently bear commercial risk. The stack works when each layer is assigned to the party best able to bear it, and it fails when one party is asked to bear a risk outside its competence. This is the same principle that underlies the insurance pillar: the weather index insurer bears the weather risk, the bank bears the credit risk, and the farmer bears the basis risk. The electrolyzer case is the same logic applied to a different risk.

9.3.2 The Low-Carbon Cement Plant: Government Funds and the First-Mover Problem

The second case is a low-carbon cement plant, using a carbon-capture or a novel clinker process. Cement is one of the hardest sectors to decarbonize, because roughly half of the emissions come from the chemistry of limestone calcination, not from fuel combustion, and because the product is a commodity with a thin margin. A green cement plant costs more to build and more to operate than a conventional plant, and the incremental cost is recoverable only if the market pays a green premium—which, in most jurisdictions, it does not, because public procurement rules do not yet discriminate, and because the carbon price is too low to close the gap.

The financing structure that has emerged is a public-private partnership in which the government bears the first-mover cost. The specific design varies. In the European Union, the Innovation Fund—financed by the EU emissions trading system—provides grants for first-of-a-kind commercial demonstration projects, covering up to 60 percent of the incremental capital cost. In China, the model has been provincial government-guided funds combined with state-owned enterprise balance sheets, where the SOE absorbs the first-mover cost as a strategic investment and the government fund provides the equity. In both cases, the logic is the same: the first plant is a public good, because it generates the learning that reduces the cost of the second plant, and the private investor cannot capture that learning.

The case exposes the boundary condition of the government-fund instrument. The first plant can be publicly financed; the tenth plant cannot, and should not be. The test of a successful first-mover program is whether it generates a self-sustaining market—whether the learning from the first plant reduces the cost of the second enough that private capital will finance it without the subsidy. This is an empirical question, and the evidence is not yet in for most of these programs. The EU Innovation Fund's projects are recent; the Chinese SOE-led plants are recent. What can be said is that the programs that build in a declining subsidy schedule—a higher grant for the first plant, a lower one for the second, none for the third—are more likely to generate a self-sustaining market than those that offer a flat subsidy, because the declining schedule forces the learning to happen.

9.3.3 The Forestry Cooperative: Insurance and the Credit Linkage

The third case is smaller and less glamorous, and it connects most directly to the earlier chapters of this book. A forestry cooperative in a mountainous region—I am drawing on the Chinese cases discussed in Chapter 7—plants a mix of timber and carbon-sequestration species, with a rotation of twenty to thirty years. The cooperative needs credit to finance the planting and the early maintenance, before any revenue arrives. The bank is willing to lend, but the loan is secured against the forest, and the forest is exposed to fire, pest, and drought. Without insurance, the bank either declines or prices the loan at a rate that makes the project uneconomic.

The structure that works is a tripartite agreement among the bank, the insurer, and the cooperative, of the kind that the National Financial Regulatory Administration's rural finance guidance permits. The insurer provides forest insurance, with the premium partly subsidized by the government. The bank lends against the insured forest, at a lower rate because the insurance caps the downside. The cooperative gets the credit it needs, and the insurer gets a portfolio of risks that is diversifiable because the fires and droughts are not perfectly correlated across regions.

This case is the clearest illustration of the linkage between the insurance pillar and the catalysis pillar. The insurance does not finance the planting; the bank does. The insurance makes the bank willing to lend, by converting an uninsured catastrophic risk into a priced premium. The catalysis is indirect: the insurance enables the credit, and the credit enables the green investment. The bank's credit assessment screens the cooperative, the forest policy caps the fire and drought exposure, and the loan finances the planting. Each pillar does what it is good at, and the financing closes.

The case also exposes the limits. The basis risk in forest insurance is significant: a fire that destroys one cooperative's forest may not trigger a regional index, and the cooperative bears the residual loss. The premium subsidy is a fiscal cost that must be budgeted. And the tripartite agreement is administratively complex, requiring coordination among three institutions with different reporting requirements. None of these limits is fatal, but each is a reason why the model has not scaled as fast as its logic suggests it should.

Table 9.1 summarizes the three cases against the frictions and instruments developed earlier.

Table 9.1: Three green-tech financing cases mapped to frictions and instruments.

9.4 The Catalytic Mechanism: Conditions and Limits

The three cases point toward a general statement of how sci-tech finance catalyzes green innovation, and it is worth stating it explicitly before turning to the limits.

The mechanism has four links. First, a screening link: the investor or lender must be able to distinguish a good green technology from a bad one, which requires the rating and data infrastructure of Chapters 3 and 4. Second, a risk-allocation link: the risks of the project—technology, construction, market, policy—must be assigned to parties that can bear them, which requires the insurance and guarantee instruments of Chapters 6 through 8 and the risk-sharing designs of Section 9.2. Third, a capital-stack link: the financing must be assembled from multiple sources with different risk appetites and time horizons, which requires the venture, government, and bank instruments to be combinable. Fourth, an exit link: the capital must be recyclable, so that the returns from successful investments can be redeployed into new ones, which requires functioning exit channels—IPOs, secondary sales, and, for the debt instruments, amortization.

The four links are sequential in logic but simultaneous in practice, and a weakness in any one of them weakens the whole. This is why piecemeal interventions underperform. A government that provides capital but no screening mechanism funds bad projects; a government that provides screening but no risk-sharing cannot close the valley; a government that provides risk-sharing but no exit channels traps its capital in illiquid positions. The integrated mechanism this book proposes is not a luxury; it is the minimum condition for the catalysis to work.

What the mechanism does not claim is important. It does not claim that sci-tech finance can make an uneconomic technology economic. A technology that cannot reach cost parity at scale, even with learning, will not be saved by better financing; the financing can only accelerate the learning, not substitute for it. It does not claim that public risk-bearing is costless. Every guarantee and every first-loss tranche is a contingent liability, and the fiscal cost is real even when it does not appear in the current budget. And it does not claim that the mechanism works uniformly across technologies. The evidence from the three cases suggests that it works best where the technology is modular and the market is contractible—electrolyzers, solar, storage—and least well where the technology is integrated and the market is policy-dependent—cement, steel, aviation fuel.

The last limitation deserves emphasis because it bears on the sequencing question that Chapter 14 will take up. The technologies that are hardest to decarbonize are precisely the ones where the catalytic mechanism is weakest, because the policy dependence is highest and the first-mover cost is largest. This is not an argument for abandoning the mechanism in those sectors; it is an argument for combining it with a stable policy signal—a carbon price, a procurement mandate—that reduces the demand-side uncertainty the mechanism cannot address. Sci-tech finance is a complement to policy, not a substitute for it.

I want to close with a note on the evidence base, in the interest of the candor this book has tried to maintain. The cases in this chapter are drawn from public sources and from my own field observations, and they are illustrative rather than representative. The quantitative literature on the effectiveness of government-guided green funds is small and recent, and the results are sensitive to the specification. The claim that the integrated mechanism outperforms piecemeal interventions is a theoretical inference supported by the cases, not a statistically established finding. I state it as a provisional conclusion, to be revised as the evidence accumulates.

What is not provisional is the direction of the policy problem. The valley of death is real, the frictions are identifiable, and the instruments exist. The question is whether they can be assembled into a mechanism that is more than the sum of its parts. The next chapter takes up one of the most important of those parts—government-guided venture capital and the green premium—and asks whether the public money can be structured to pay for itself.

9.5 From Catalysis to Design

The catalytic mechanism described in this chapter has a weak point that the cases make visible but do not resolve. It is the interaction between the government fund and the green premium. A first-of-a-kind plant financed by public equity can demonstrate that the technology works; it cannot demonstrate that the output will sell at a price that covers the incremental cost. The EU Innovation Fund grant covers up to 60 percent of the incremental capital cost of a demonstration plant, but the plant still has to sell its low-carbon cement into a market where the carbon price is too low to close the cost gap. The Chinese SOE-led plants face the same arithmetic, with the SOE balance sheet absorbing the operating loss instead of a grant absorbing the capital cost. In both designs, the public money buys the learning but not the market.

That gap is where the next chapter begins. Government-guided venture capital, structured well, can do something the grant cannot: it can hold an equity position in the downstream buyer as well as the upstream producer, and it can use that position to organize a procurement commitment—a contract for the green output at a price that covers the incremental cost for a defined period, with the price declining as the learning curve bends. The design question is whether such a commitment can be priced so that the public fund recovers its capital from the sale of the equity once the market stands on its own. The evidence on that question is thinner than the evidence on the valley of death itself, and it is the subject of Chapter 10.