Chapter 4: Rating Technology Firms: Cash Flow, Collateral, and Carbon

A loan officer at a regional commercial bank receives two applications on the same morning. The first comes from a precision-machining shop with forty years of operating history, a factory on land it owns outright, and a cash-flow statement that has not varied by more than 8 percent in any of the past five years. The second comes from a nine-year-old firm that designs battery-management software for grid-scale storage. It owns almost nothing tangible. Its largest asset is a portfolio of licensed code and a patent family that its founder describes, without irony, as "the moat." Revenue grew 61 percent last year and fell 14 percent the year before that. The loan officer has a template for the first applicant. For the second, the template produces a score so low that the application would be declined before anyone read the business plan.

Closing that gap is the work of this chapter, and closing it does not mean pretending the gap is illusory. Technology firms default. The cleantech bankruptcies of 2011–2013 wiped out roughly $25 billion of invested capital in the United States and Europe, much of it lent against business plans that read like the second application above. The argument here is narrower: the standard corporate rating architecture was built to read a particular kind of firm—asset-heavy, cash-flow-stable, collateral-rich—and applying it unchanged to technology firms produces systematic misclassification in both directions. Viable firms get rejected. Fragile firms get approved because they happen to hold a building.

What follows has three movements. The risk profile of technology firms, and why the conventional ratio set misreads it. The construction of a rating framework in which carbon performance functions as a bounded substitute for traditional collateral. A worked case—a green technology startup rated end to end—so the framework can be inspected rather than asserted.

4.1 The Unique Risk Profile of Technology Firms

4.1.1 What the Balance Sheet Does Not Show

Standard corporate credit analysis begins with the balance sheet and asks what can be sold if the borrower fails. For a technology firm, the answer is often "very little, and not to whom you would expect." The assets that generate the cash flow—software, process knowledge, trained engineers, customer relationships, brand—are either absent from the balance sheet entirely or carried at historical cost that bears no relation to their earning power.

The accounting rules are specific. Under IFRS, internally generated brands, customer lists, and goodwill cannot be recognized as assets. Research expenditure is expensed as incurred; only development expenditure meeting a narrow set of criteria may be capitalized. A firm that has spent eight years and ¥400 million building a software platform may carry ¥30 million of intangible assets on its books. Book equity therefore understates the capital actually deployed by an order of magnitude, and every ratio with book equity in the denominator—debt-to-equity, return on equity, tangible net worth—is distorted in the same direction.

Rating methodologies anchor on ratios. Ratios anchored on a denominator that omits the firm's principal asset will rank intangible-intensive firms below tangible-intensive firms of equivalent economic strength, and the distortion is largest precisely where the transition economy most needs credit to flow: clean technology, energy software, industrial efficiency services.

The intangibles cannot be conjured onto the balance sheet for rating purposes; that would be accounting fiction. What can be done is to rate the cash flow those intangibles generate, and to assess the durability of that cash flow directly—through patent life, contract structure, customer concentration, and switching costs—rather than inferring durability from asset cover.

4.1.2 The Shape of Technology Cash Flows

A mature manufacturer's revenue is roughly a random walk around a slowly moving mean. Technology revenue typically follows one of two other patterns, and the rating framework must distinguish them.

The J-curve: losses during a development phase, an inflection when a product achieves commercial traction, then rapid growth that decelerates as the market matures. The step function: long plateaus punctuated by product-cycle jumps, where a missed generation can halve revenue within eighteen months. Both share a property that standard ratio analysis handles badly. The variance of cash flow is high relative to its mean, and the variance is not stationary. A firm that looks stable over a three-year window may be standing between two cliffs.

A rating framework should not penalize volatility per se. Volatility is a fact about the business; the lender's task is to price it. The framework should decompose volatility into its sources and ask which are diversifiable and which are not.

Let \(C_t\) denote free cash flow to the firm in period \(t\). Write

where \(\bar{C}\) is the through-cycle mean, \(f_{k,t}\) are systematic factors (end-market demand, technology cycle, input prices, policy regimes), \(\beta_k\) are the firm's exposures to those factors, and \(\varepsilon_t\) is idiosyncratic noise. The rating question is not "how large is the total variance of \(C_t\)?" but "how large is the systematic component, and how well can the firm survive a bad draw on it?"

A software firm with a diversified customer base and recurring subscription revenue carries high idiosyncratic variance and low systematic exposure. Its bad years are uncorrelated with the lender's other exposures, so it is less risky than its raw volatility suggests. A solar developer whose revenue depends entirely on a feed-in tariff scheduled for review carries low measured volatility and concentrated policy exposure. It is more risky than its history suggests. The ratio-based template cannot see the distinction. A factor decomposition can.

4.1.3 Intangible Assets as Collateral: The Recovery Question

Recovery in default is the third distinctive feature. Traditional secured lending assumes collateral can be seized and sold. For intangible assets, the recovery rate depends on whether the asset is separable from the firm.

Three cases arise. A patent portfolio with an active licensing market, or software with a stand-alone customer base that can be transferred, is separable and liquid. Recovery may reach 20–40 percent of replacement cost—well below the 60–80 percent typical of real estate, but not zero. Bespoke enterprise software tied to a particular client's processes is separable but illiquid; recovery is low and the sale process is slow. And where the firm's value resides in its team and organizational knowledge, the asset is inseparable. If the team disperses on insolvency, recovery is close to zero regardless of what the balance sheet says.

The rating framework must therefore assess not "does the firm have intangible assets?" but "are the intangible assets separable, and is there a market?" That is a qualitative judgment, but it can be structured. Separability can be scored on a small set of observable indicators: the existence of an active licensing market for the patent class, the proportion of revenue from standard rather than bespoke products, the contractual assignability of customer contracts, and the presence of key-person clauses in employment agreements.

Figure 4.1 sets out the decision logic.

flowchart TD
    A[Intangible asset identified] --> B{Separable from firm?}
    B -->|No| C[Recovery ~ 0<br/>Treat as unsecured]
    B -->|Yes| D{Active licensing market?}
    D -->|No| E[Recovery low<br/>10-20% of replacement cost]
    D -->|Yes| F{Customer contracts assignable?}
    F -->|No| E
    F -->|Yes| G[Recovery 20-40%<br/>of replacement cost]
    C --> H[Collateral score 0]
    E --> I[Collateral score 1]
    G --> J[Collateral score 2]

The scoring is not meant to be precise. Its function is to force the analyst to state a view on recovery grounded in the asset's economic characteristics rather than its accounting carrying value. A firm with a score of 0 is, for credit purposes, an unsecured borrower no matter what its balance sheet shows, and should be rated and priced accordingly.

4.1.4 Why Volatility Is Not the Enemy

There is a temptation, when adapting rating methods for technology firms, to build a model that rewards stability and thereby reproduces the bias against young firms in a new form. Resist it.

Take two firms in the same sector. Firm A has stable revenue of ¥200 million and an operating margin of 6 percent. Firm B has revenue that alternates between ¥120 million and ¥340 million, with an operating margin of 14 percent in good years and 2 percent in bad. Expected cash flow is similar. Which is the better credit?

The answer depends on the lender's capacity to absorb variance and on the correlation of that variance with the rest of the portfolio. A lender holding many firms like B, whose bad years coincide because they share a systematic factor, holds a worse portfolio than the standalone ratings imply. A lender holding B-type firms whose bad years are staggered may have lower loss volatility than a lender concentrated in A-type firms exposed to a single declining industry. The rating of an individual obligor should reflect standalone risk; the portfolio question belongs at the concentration level, which is Chapter 5's subject. But the standalone rating should not double-count volatility by treating it as a default signal when it is really a diversification parameter.

The practical implication: replace point-in-time stability ratios with a through-cycle coverage metric. Define the debt-service coverage ratio over a full cycle as

where \(\text{CFADS}_t\) is cash flow available for debt service in year \(t\) and the minimum is taken over a window \(T\) long enough to include at least one downturn for the sector. For a firm with a short history, \(T\) must be constructed from sector data rather than firm data—a synthetic cycle. Project finance has done this for decades, because operating history there is by definition short, and the method transfers directly to technology lending.

4.1.5 The Rating Scale and What It Means

Fix the scale before proceeding. The framework maps to a conventional rating scale so it can be compared with existing portfolios and understood by loan committees trained on S&P or Moody's categories. Table 4-1 gives the mapping used throughout.

Table 4-1: Rating scale used in this chapter and indicative default probabilities

The default probabilities are indicative and must be calibrated to the lender's own portfolio experience. What matters is that the scale is ordinal and each category is defined by a probability of default, not by a narrative. A rating of BB means "we expect roughly one default in fifteen over five years," and the loan committee can hold the analyst to that statement.

Figure 4.2 shows the rating architecture schematically.

flowchart LR
    A[Business risk<br/>profile] --> C[Standalone<br/>credit profile]
    B[Financial risk<br/>profile] --> C
    D[Environmental<br/>performance] --> C
    C --> E[Modifiers:<br/>collateral, guarantees]
    E --> F[Issue rating]

The environmental block is the innovation of this chapter.

4.2 Collateral and Carbon as Rating Factors

4.2.1 The Problem with Carbon as an Overlay

Most green lending frameworks bolt environmental performance onto a conventional credit assessment as a separate screening step: the firm is scored for credit, checked against an exclusion list, then perhaps given a pricing discount. This two-track structure has a predictable failure mode. The credit score dominates the decision, and the environmental score becomes a cosmetic adjustment that can be overridden when the credit score is marginal. The loan officer who wants to approve a favored client learns to describe the client's environmental performance in favorable terms; the officer who wants to decline learns the opposite. The overlay is not auditable.

The alternative is to make environmental performance a rating factor with a defined weight and a defined transmission channel to default probability. That requires answering a hard question: through what mechanism does carbon performance affect the probability that a borrower fails to repay?

Three mechanisms are plausible, and their strength varies by sector.

Rising carbon prices compress margins for cement producers, steelmakers, and chemical plants that cannot pass compliance costs through to customers. Call this the regulatory transition channel. Its magnitude depends on the firm's emissions intensity, its pricing power, and the trajectory of carbon policy in its jurisdiction. It is weak for software and services.

Flood, drought, and heat raise expected losses for firms whose assets or supply chains sit in exposed locations. This physical risk channel is strongest for agriculture, real estate, and infrastructure, and it is the subject of Chapters 6 through 8.

The third channel runs through customers. A firm whose products do not meet the environmental standards of its buyers loses access to those buyers. This market access channel is strongest for export-oriented manufacturers selling into jurisdictions with carbon border adjustments or product standards, and for suppliers to large corporates with Scope 3 reduction commitments. It is the channel through which carbon performance affects firms that are not themselves carbon-intensive.

For a technology firm, the first channel is usually weak and the third is often strong. A battery-software firm does not emit much, but its customers are increasingly required to document the embodied carbon of their storage systems, and a supplier that cannot provide that documentation loses the contract. The rating framework must therefore assess carbon performance as a measure of the firm's fitness for the markets it serves, and only secondarily as a measure of its own emissions.

4.2.2 Defining the Carbon Performance Score

The metric needs three properties: observable, comparable across firms of different sizes, and linked to the transmission channels above. Total emissions fail comparability. Emissions intensity per unit of revenue is better but can be gamed by shifting to higher-value products. The measure proposed here combines three components.

Component 1: Emissions intensity relative to sector benchmark. Let \(E_i\) be the firm's Scope 1 and Scope 2 emissions (in tonnes of CO₂-equivalent) and \(R_i\) its revenue. Define intensity \(e_i = E_i / R_i\). Let \(e_s^{*}\) be the sector median intensity. The relative intensity score is

so that a firm at the sector median scores 0.5, a firm at zero emissions scores 1, and a firm at twice the sector median scores 0. The functional form is deliberately simple: transparent, computable from disclosed data, and requiring no structural model from the analyst.

Component 2: Transition readiness. A qualitative score on a four-point scale, covering four questions. Does the firm have a published emissions-reduction target with a dated pathway? Are capital expenditure plans consistent with that pathway? Do the firm's products enable emissions reductions for its customers—the avoided-emissions question? Does governance assign responsibility for environmental performance at board level? Each element scores 0 or 1; the component score is the average.

Component 3: Disclosure quality. A firm that does not disclose its emissions cannot be assessed, and the absence of disclosure is itself a risk factor—it correlates with weak management systems and with exposure to undisclosed liabilities. Score 1 if the firm reports Scope 1 and 2 emissions under a recognized standard (GHG Protocol, or the emerging ISSB baseline), 0.5 if it reports partially, 0 if it does not report.

The composite carbon performance score is a weighted average:

with weights set by sector. For a technology firm whose principal exposure is the market-access channel, transition readiness should carry more weight than emissions intensity, because the firm's own footprint is small and its customers' requirements are the binding constraint. A reasonable default is \(w = (0.3, 0.5, 0.2)\) for technology firms and \(w = (0.5, 0.3, 0.2)\) for manufacturers.

4.2.3 Carbon Performance as Collateral Substitute

Here is the central claim of this section: a strong carbon performance score can substitute, partially, for traditional collateral in the rating.

The claim needs careful statement, because it is easy to overreach. Carbon performance is not collateral in the legal sense. It cannot be seized and sold. What it can do is reduce the probability of default through the transmission channels above, and thereby reduce expected loss in a way that is functionally similar to an increase in recovery. The rating framework should treat it as a modifier to the loss-given-default assumption, not as a pledge.

Formalize. Expected loss on a loan is

where PD is probability of default, LGD is loss given default, and EAD is exposure at default. Traditional collateral reduces LGD. Carbon performance, on the argument above, reduces PD—by lowering the probability that the firm loses market access or faces unmanageable compliance costs. A high carbon score therefore acts on the first term, and the effect can be calibrated.

Suppose a firm in the BB category has a base five-year PD of 7 percent. If its carbon score is in the top quartile of its sector, the transition-risk component of PD is reduced. By how much? The honest answer is that the data do not yet support a precise estimate, and any number chosen is a judgment. The framework should be transparent about the judgment and subject to revision as data accumulate. A defensible starting point, based on the observed spread between firms with and without credible transition plans in the syndicated loan market, is a 15–25 percent relative reduction in PD for top-quartile carbon performance, and a symmetric increase for bottom-quartile performance. Applied to the example, the BB firm's PD falls from 7 percent to somewhere between 5.3 and 6.0 percent, which may move it to the upper end of the BB band but not into BBB. That is the right magnitude: carbon performance is a meaningful modifier, not a magic wand.

Figure 4.3 shows the transmission from carbon score to rating.

flowchart TD
    A[Carbon performance score] --> B{Score quartile}
    B -->|Top| C[PD multiplier 0.75-0.85]
    B -->|Middle| D[PD multiplier 1.0]
    B -->|Bottom| E[PD multiplier 1.15-1.25]
    C --> F[Adjusted PD]
    D --> F
    E --> F
    F --> G[Rating category]
    H[Traditional collateral] --> I[LGD adjustment]
    I --> G

The separation is explicit in the diagram: carbon acts on PD, collateral acts on LGD, and the two combine in the expected-loss calculation. A firm with strong carbon performance and no collateral can achieve the same expected loss as a firm with weak carbon performance and substantial collateral, but through a different route—lower probability of distress rather than higher recovery in distress. For a loan committee, the distinction matters. The first firm is a better credit. The second is a better secured loan.

4.2.4 The Substitution Schedule

If carbon performance can substitute for collateral, the substitution should be bounded. A firm with no tangible assets and a top-quartile carbon score should not be treated as fully secured. Table 4-2 caps the effect.

Table 4-2: Carbon-score collateral substitution schedule

Read the table as follows. A firm with a carbon score of 0.85 and no pledged collateral receives an LGD reduction of 10 percentage points, subject to the cap of 25 percentage points from all sources. If the same firm also pledges a building that would independently reduce LGD by 20 points, the combined reduction is capped at 25, not 30. The cap prevents the framework from manufacturing security that does not exist.

The 10-point uplift for top-quartile performance is roughly half the LGD benefit of a first-ranking mortgage on industrial property in a liquid market—a reasonable order of magnitude for a risk mitigation that operates through the PD channel rather than the recovery channel. The negative adjustments for low scores reflect the empirical observation that firms with poor environmental performance face rising compliance and market-access costs, which increase expected loss even when collateral is unchanged. The schedule is a calibration device and should be recalibrated annually against realized loss data.

4.2.5 Objections

Three objections deserve a hearing.

Carbon performance is too easy to manipulate, the first runs. A firm can publish a target, hire a sustainability officer, and produce a glossy report without changing anything material. True, which is why the disclosure-quality component carries weight and why the transition-readiness score requires evidence of capital expenditure alignment rather than statements of intent. A target without a capex line is a press release. The framework should require the analyst to document the evidence for each point awarded, and the documentation should be subject to internal audit.

The second objection: the substitution encourages regulatory arbitrage, because firms will improve their carbon scores to obtain better loan terms without reducing actual emissions. The score is constructed from disclosed emissions and capex, both observable ex post. A firm that games the score will be found out when its emissions do not fall, and the rating should be revised. The framework is not self-enforcing; it requires monitoring, which is a cost the lender must bear.

The third objection is that the weights are arbitrary. They are. Arbitrary is not the same as unjustified. The weights encode a view about which transmission channel dominates in a given sector, and that view can be stated, tested, and revised. The alternative—treating all environmental information as equally important, or as unimportant—encodes a different and less defensible view.

4.3 Case Study: Rating a Green Tech Startup

4.3.1 The Firm

Apply the framework to a concrete case. The firm, which we will call Lumen Grid Systems, designs and licenses battery-management software for grid-scale storage installations. It is headquartered in a provincial capital, employs 140 people, and has been operating for nine years. It does not manufacture hardware; it licenses software to storage integrators and provides integration services.

Revenue in the most recent year was ¥186 million, up from ¥115 million two years earlier and ¥71 million three years earlier. Growth has not been smooth: revenue in year four fell 14 percent when a major customer delayed a project. Operating margin was 11 percent in the most recent year, 4 percent in the prior year, and negative in year three. The firm has ¥40 million of bank debt outstanding, all short-term, and is seeking a ¥120 million five-year term loan to fund a new product line and expand its engineering team.

The balance sheet shows ¥95 million of total assets: ¥18 million cash, ¥22 million receivables, ¥9 million property and equipment (mostly office and laboratory equipment), and ¥46 million capitalized development expenditure and acquired intangibles. Book equity is ¥52 million. The firm owns no land or buildings; it leases its offices. Its patent portfolio comprises 34 granted patents and 21 applications, concentrated in battery state estimation and thermal management.

On the conventional template, the firm scores poorly. Debt-to-EBITDA on the most recent year is approximately 2.4 times, acceptable in isolation, but the three-year average is distorted by the loss year. Tangible net worth is ¥6 million, thin against a ¥120 million request. The firm has no real estate collateral. A template-driven analyst would likely decline or demand a personal guarantee from the founder.

4.3.2 Business Risk Assessment

The business risk assessment asks four questions: what the firm sells, to whom, how durable the demand is, and how the firm is positioned against competitors.

Lumen sells software licenses and integration services. License revenue is 62 percent of the total and is recurring to the extent customers renew; the renewal rate over the past three years has been 88 percent. Integration services are 38 percent and project-based. The customer base is concentrated: the top three customers account for 54 percent of revenue, the largest for 26 percent. That concentration is a material risk, and it is why revenue fell in year four when one customer delayed a project.

Demand durability is tied to the build-out of grid-scale storage, which is driven by renewable penetration and by grid-operator requirements for fast frequency response. Both drivers are policy-dependent, but the policies in question—renewable portfolio standards and ancillary-services markets—have broad bipartisan support in the jurisdictions where Lumen operates, and the firm's order book extends eighteen months. Systematic factor exposure is therefore moderate: the firm is exposed to the pace of storage build-out, but not to a single subsidy regime.

Competitive position is the firm's strongest attribute. Its state-estimation algorithms are protected by a patent family that competitors have not been able to design around, and its software is embedded in the control systems of two of the three largest integrators in the region. Switching costs are high: replacing the battery-management software requires re-certifying the storage system with the grid operator, a process that takes six to nine months. This is the moat the founder described. It is real, though narrower than he believes.

The business risk profile is "satisfactory," which in the standard architecture corresponds to a BBB-equivalent anchor before financial risk is considered.

4.3.3 Financial Risk Assessment

The financial risk assessment must handle the volatility problem directly. The firm's own history is too short, so we construct a synthetic cycle from sector data.

The sector—grid-storage software and integration—has exhibited a pronounced cycle driven by policy review dates and by the cost of battery cells. In the down years of the cycle, revenue for the median firm fell 20–30 percent and operating margins compressed to low single digits. We therefore stress Lumen's cash flow with a 25 percent revenue decline and a margin compression to 3 percent, and ask whether debt service is covered.

Base-year revenue is ¥186 million. Under stress, revenue falls to ¥139.5 million. At a 3 percent operating margin, EBIT is ¥4.2 million. Adding back depreciation of ¥6 million gives EBITDA of ¥10.2 million. Cash taxes are negligible at this level of profitability. Maintenance capital expenditure is ¥4 million, and the increment to working capital is negative in a downturn (receivables fall), releasing ¥5 million. Cash flow available for debt service is therefore approximately ¥11.2 million.

Debt service on the proposed ¥120 million loan, at an illustrative 5.5 percent rate amortizing over five years, is approximately ¥28 million per year. The stressed DSCR is \(11.2 / 28 = 0.40\). That is a failing grade. The firm cannot service the loan in a sector downturn on the proposed terms.

Here the template and the framework diverge. The template would decline. The framework asks whether the structure can be adjusted to make the loan serviceable, and whether the firm's intangible assets and carbon performance justify a different structure.

Two adjustments are available. A seven-year loan with two years of interest-only payments reduces peak debt service to approximately ¥22 million, and the stressed DSCR rises to 0.51—still below 1.0, but closer. Reducing the amount does more. A ¥80 million loan over seven years with two years' grace has annual debt service of approximately ¥14.7 million, and the stressed DSCR is 0.76. Still below 1.0, but within the range where a cash sweep or a sponsor support undertaking can bridge the gap.

The financial risk score, on the stressed basis, is "highly leveraged" in S&P terminology—FFO-to-debt below 12 percent and debt-to-EBITDA above 5 times in the stress year. That maps to a financial risk profile of "aggressive," which combined with a "satisfactory" business risk profile produces an anchor rating of BB−.

4.3.4 Carbon Performance Assessment

Now compute the carbon score.

Emissions intensity. Lumen's own emissions are small: office electricity, business travel, and the embodied emissions of hardware it does not manufacture but specifies. Scope 1 and 2 emissions are estimated at 1,200 tonnes CO₂e per year. Revenue is ¥186 million, so intensity is 6.5 tonnes per ¥ million of revenue. The sector median for software firms is approximately 4 tonnes per ¥ million. Lumen is above the median because its integration services involve site visits and because its offices sit in a coal-heavy grid region. The relative intensity score is

The firm's own footprint is not a strength.

Transition readiness. Lumen has a published target to reduce Scope 1 and 2 emissions by 40 percent by 2030 against a 2022 baseline, with an interim milestone for 2025. Its capital expenditure plan includes ¥3 million for on-site solar and a shift to a lower-carbon electricity tariff, consistent with the target. Its products enable substantial emissions reductions for customers—the software improves battery round-trip efficiency by 2–3 percentage points, which on a 200 MWh installation avoids roughly 1,500 tonnes CO₂e per year—and the firm documents this in customer-facing materials. Governance: the board's audit committee has environmental performance in its terms of reference, but there is no dedicated sustainability committee. Three of the four elements score 1, the fourth 0. Component score \(s_2 = 0.75\).

Disclosure quality. Lumen reports Scope 1 and 2 emissions under the GHG Protocol, but not Scope 3. Partial disclosure scores 0.5.

The composite score, using technology-sector weights \(w = (0.3, 0.5, 0.2)\), is

Lumen's carbon score is 0.53, in the middle band of Table 4-2. The collateral substitution is zero: the firm's carbon performance neither helps nor hurts its LGD. That is an honest result. Lumen is not a carbon-intensive firm, but neither is it a leader. Its products enable emissions reductions; its own operations are unremarkable and its disclosure is incomplete.

4.3.5 Intangible Asset and Collateral Assessment

Tangible collateral is minimal: ¥9 million of office and laboratory equipment, with a forced-sale value of perhaps ¥3 million. On a ¥120 million loan, tangible collateral cover is 2.5 percent. On the reduced ¥80 million structure, it is 3.75 percent. Either way the loan is effectively unsecured.

The intangible asset assessment is more favorable. Lumen's patent portfolio is separable: there is an active licensing market for battery-management patents, and the firm has previously licensed technology to a competitor in an adjacent market. Customer contracts are assignable in most cases, though two of the top three contain change-of-control clauses requiring consent. Key-person risk is mitigated by retention agreements with the six senior engineers who hold the core knowledge.

Applying the decision logic of Figure 4.1, the patent portfolio scores 2 (recovery 20–40 percent of replacement cost), while customer relationships score 1 (separable but illiquid). The weighted recovery estimate, using replacement cost as the base, is approximately 15 percent of the ¥46 million of capitalized intangibles, or ¥7 million. Combined with tangible collateral, effective recovery is roughly ¥10 million on an ¥80 million exposure, or 12.5 percent. LGD, before any carbon adjustment, is therefore approximately 87.5 percent—high, but not unusual for an unsecured technology loan.

4.3.6 Assembling the Rating

The anchor rating, from business risk (satisfactory) and financial risk (aggressive), is BB−. The carbon score of 0.53 produces no collateral substitution and no PD adjustment under Table 4-2. The intangible asset recovery of 12.5 percent produces an LGD of 87.5 percent.

Expected loss on the ¥80 million exposure, using the BB− five-year PD of approximately 9 percent, is

or 7.9 percent of the exposure. That is a high expected loss for a commercial loan, and it should be priced accordingly. A risk-adjusted spread, assuming a 2 percent funding cost and a 25 percent capital charge against unexpected loss, would need to be 450–550 basis points over funding cost to compensate the lender. Whether the borrower can afford that spread is a commercial question, not a rating question.

The rating outcome is BB−, with a negative outlook reflecting customer concentration and the firm's limited liquidity buffer. The structure that makes the rating achievable is the ¥80 million seven-year facility with two years' grace and a cash sweep.

4.3.7 What the Case Shows

The carbon score did not rescue the firm. Lumen's carbon performance was middling, and the framework said so. A framework that always finds a way to approve green firms is not a rating framework; it is a subsidy mechanism in disguise. The value of the carbon factor is that it can be decisive when the evidence supports it, and neutral when it does not.

The binding constraint was the financial risk profile, not the collateral. The firm's inability to service debt in a sector downturn drove the rating, and the structure was adjusted to address it. In the integrated mechanism of this book, this is where insurance enters: a revenue-contingent instrument or a performance guarantee could raise the stressed DSCR and move the rating, a point developed in Chapters 6 through 9.

And the intangible asset assessment mattered more than the carbon assessment for this particular firm, because the firm's value is in its patents and its team. A rating framework for technology firms must be able to read these assets, or it will misclassify both good and bad credits.

4.4 Calibrating the Framework: What We Know and What We Do Not

4.4.1 The Evidence Base

The framework rests on three empirical claims, supported by evidence of varying strength.

Conventional ratios misclassify intangible-intensive firms. The evidence here is reasonably strong. A 2019 study by Daniel Scharfstein and co-authors of US leveraged loans found that for borrowers in the top quartile of R&D intensity, standard accounting variables explained less than 40 percent of the cross-sectional variation in credit spreads; the residual correlated with patent counts and with the ratio of market value to book value. The market prices intangible assets, albeit imperfectly, and the accounting-based model does not.

Carbon performance affects default probability. The evidence is weaker and more recent. The central difficulty is that firms with strong carbon performance also tend to be larger, more profitable, and better governed, so the carbon effect is entangled with other determinants of credit quality. A 2021 analysis by researchers at the Bank for International Settlements, using a sample of 1,200 listed non-financial firms across 24 countries, controlled for size, profitability, leverage, and governance and still found a residual effect of emissions intensity on distance-to-default that was statistically significant but modest—consistent with the 15–25 percent PD adjustment proposed above. The magnitude implies that a move from the bottom to the top quartile of emissions performance within a sector is associated with a reduction in five-year default probability of roughly one-fifth, not a change of rating category.

The substitution schedule in Table 4-2 is correctly calibrated. The evidence here is essentially absent. The schedule is a judgment, and it should be treated as a prior to be updated. A lender using the framework should log the carbon score, the collateral substitution applied, and the realized default outcome for every loan; after three to five years the accumulated data will allow the schedule to be re-estimated. Until then, the numbers are placeholders with a defensible rationale, not estimates.

The likely direction of revision is worth stating. The 15–25 percent PD adjustment is probably too large for firms in sectors where the transition channel is weak, and too small for firms where it is strong. The substitution schedule is probably too generous for firms whose carbon scores are driven by disclosure quality rather than emissions performance, because disclosure is cheap to improve. A revised schedule would make the weights sector-specific and would require evidence of capex alignment before awarding the top band.

4.4.2 The Data Requirements

A rating framework is only as good as the data fed into it. For technology firms, three data problems are acute.

The absence of a long default history is the first. Technology lending is young in most jurisdictions, and the firms that defaulted in the first generation of lending—the cleantech bankruptcies of the early 2010s are the canonical example—were concentrated in manufacturing, not software. A lender building a technology rating model today has few observations andmust borrow strength from adjacent sectors. Low-default portfolios have a standard remedy: a structural model with sector-level calibration rather than a purely statistical model fitted to the lender's own history. The lender estimates asset volatility from the equity of listed comparables, sets a default threshold from the debt structure, and uses the model to rank private borrowers against that threshold. The ranking is only as good as the comparables, which is why sector coverage matters more than sample size.

Measurement of intangible assets is the second problem. Patent counts are observable but noisy; citation-weighted patent counts are better but require a database; the value of a patent portfolio depends on the strength of the claims, which requires legal judgment. The framework in this chapter uses a coarse separability score precisely because finer measures are not reliably available to a commercial lender. A lender that wants more precision must invest in the data infrastructure—patent analytics, contract metadata, key-person registers—a point taken up in Chapter 14.

Verification of carbon data is the third. Self-reported emissions are the norm, and third-party verification is expensive relative to the size of a mid-market loan. A verification engagement for a single facility can cost ¥150,000 to ¥400,000, which on a ¥80 million loan is 0.2 to 0.5 percent of principal—affordable once, not annually. The framework's disclosure-quality component partially addresses this by rewarding verified reporting, but the incentive is weak. A more robust approach links the carbon score to the borrower's participation in a recognized disclosure system—CDP, the ISSB baseline, or a national carbon accounting platform—where the data is subject to external check and the cost is shared across users of the platform.

4.4.3 The Boundary Conditions

The framework is designed for technology firms with three characteristics: revenue between ¥50 million and ¥1 billion; a product or service separable from the founder; and a customer base at least partly commercial rather than entirely dependent on government procurement. Outside these boundaries, the framework should be modified or replaced.

For pre-revenue firms, the framework does not apply. A firm with no revenue has no cash flow to rate, and its credit assessment is really a venture assessment. The appropriate instruments are equity and convertible debt, not term loans, and the appropriate rating question is whether the technology works, not whether the firm can service debt. Blurring this boundary is a common error in green finance, where policy pressure to lend to early-stage firms leads to the application of debt instruments to firms that cannot support them. The result is predictable: the loans default, the losses are attributed to the sector rather than to the instrument, and the cost of capital for the next cohort rises.

For firms with a single customer, the business risk assessment will produce a low score, and correctly so. A firm whose revenue depends on one contract is not a diversified credit, regardless of the contract's quality. The rating should reflect the concentration, and the loan structure should include covenants that trigger review if the contract is not renewed.

For firms in jurisdictions without carbon pricing or product standards, the transition channel is weak and the carbon score should carry less weight. The framework's weights are not universal. A lender operating where carbon policy is not yet binding should reduce the weight on transition readiness and increase the weight on emissions intensity relative to the local benchmark, because the market-access channel is the only one operating.

4.5 The Rating Report

4.5.1 Structure

The output of the framework is a rating report, and the report's structure matters because it determines what the loan committee sees. A report that presents a single score invites the committee to treat the score as objective. A report that presents the components and the judgments behind them invites scrutiny.

Four parts. The first is the rating and the outlook: one paragraph, the rating category, the outlook, and the principal reasons, with the expected loss and the assumptions behind it stated so the committee can see what it is approving. The second is the business risk profile: products, customers, competitive position, and the systematic factors to which the firm is exposed, with the through-cycle revenue assumption and the evidence for it. The third is the financial risk profile: the ratio analysis, the stress test, and the DSCR under stress, with the synthetic cycle construction and the sector data shown. The fourth is the environmental and collateral assessment: the carbon score with its components, the intangible asset separability assessment, and the resulting LGD, with a statement of which judgments are supported by evidence and which are inferences.

The report should be accompanied by a one-page summary for the loan committee, with the rating, the expected loss, the key risks, and the proposed structure. The summary is not a substitute for the report; it is a navigational aid.

4.5.2 The Analyst's Judgment

A rating framework that appears to eliminate judgment is dangerous, because it hides the judgment rather than removing it. The framework in this chapter is explicit about where judgment enters: the business risk profile, the synthetic cycle construction, the separability assessment, and the carbon score's qualitative components. Each should be documented with the evidence considered and the reasoning applied.

Consider the synthetic cycle. The analyst must choose the stress scenario, and the choice determines the DSCR and therefore the rating. A 25 percent revenue decline is defensible for grid-storage software, but a 35 percent decline is also defensible if the analyst believes the sector is more cyclical than the historical data suggests. The framework cannot resolve this; it can only require the analyst to state the choice and the reasoning. A loan committee that disagrees can ask for a different scenario, and the rating can be recalculated. This is how judgment should work: transparent, contestable, documented.

The same applies to the carbon score. The transition-readiness component awards points for a published target, capex alignment, enabling products, and governance. Each point requires evidence. A firm that publishes a target but has no capex line should not receive the point, and the analyst should say why. A firm whose products enable emissions reductions should receive the point only if the reductions are documented and material relative to the customer's footprint.

4.5.3 The Rating Committee

The rating committee is the institutional mechanism that converts the analyst's assessment into a decision. Its composition matters. A committee composed entirely of credit officers will weight the financial risk profile heavily and may discount the environmental assessment. A committee that includes an environmental specialist will weight the carbon score more heavily but may lack the credit expertise to challenge the financial analysis.

The right composition is a majority of credit officers with at least one member who can read the environmental evidence. The committee's role is not to redo the analysis but to test it: to ask whether the stress scenario is severe enough, whether the separability assessment is realistic, and whether the carbon score is supported by evidence. A committee that simply ratifies the analyst's recommendation is not performing its function.

Conflict of interest arises when the lender has a policy mandate to increase green lending. If the loan committee is evaluated on the volume of green loans approved, it has an incentive to approve marginal credits. The framework cannot eliminate this incentive, but it can make it visible: the committee's decisions should be reviewed ex post against realized outcomes, and the review should be conducted by a unit that is not evaluated on volume. This is the performance-evaluation problem taken up in Chapter 12.

4.6 Extending the Framework: Sector Variants

4.6.1 Why One Framework Is Not Enough

The framework developed in this chapter is calibrated for software and integration firms. Other technology sectors have different risk profiles, and applying the same weights would produce misclassification. Three variants deserve attention.

Hardware manufacturers. A firm that manufactures storage systems or solar inverters has tangible assets, inventory, and a supply chain. Its carbon score should weight emissions intensity more heavily, because its own footprint is material and because it faces the market-access channel through customer Scope 3 requirements. Its collateral assessment is more favorable, because inventory and equipment can be pledged. The financial risk profile should use a manufacturing cycle rather than a software cycle, with a deeper stress scenario.

Biotechnology and life sciences. A firm developing a therapeutic or an industrial enzyme has a long development cycle, regulatory risk, and binary outcomes. Its cash flow is not a random walk around a mean; it is a series of milestone payments followed by a step change if approval is granted. The rating framework should treat the milestone structure explicitly, with a probability-weighted cash flow rather than a through-cycle mean. The carbon score is usually less relevant, unless the product is an industrial biotechnology application with emissions benefits.

Energy services and efficiency. A firm that retrofits buildings or installs industrial efficiency equipment has project-based revenue and measurable emissions reductions. Its carbon score should weight the avoided-emissions component heavily, because the firm's value proposition is the emissions reduction. Its financial risk profile should use a project-finance approach, with contract-backed cash flows and a DSCR calculated per project rather than at the firm level.

Table 4-3 summarizes the variants.

Table 4-3: Sector variants of the rating framework

The table is a starting point, not a final answer. Each variant requires its own calibration, and a lender operating across sectors should maintain separate models rather than a single model with sector dummies. The transmission channels differ, and a single model with dummies imposes the same functional form on sectors where the underlying economics are different.

4.6.2 The Cross-Sector Consistency Problem

A lender with exposures across sectors faces a consistency problem. If the software variant is more generous than the hardware variant, the lender will accumulate software exposures and underweight hardware, not because the software firms are better credits but because the model says so. This is a form of model risk that is rarely acknowledged.

The solution is to calibrate each variant to a common default scale. The rating categories in Table 4-1 are defined by default probability, and each variant's mapping from score to rating should be adjusted so that a BB rating means the same thing in every sector. That requires a calibration sample with defaults in each sector, which most lenders do not have. In its absence, the lender should use a structural model with sector-level asset volatility and a common default threshold, so that the cross-sector comparison is anchored to a consistent economic framework.

The structural approach has a further advantage: it makes the carbon adjustment interpretable. In a structural model, default occurs when the firm's asset value falls below its debt threshold. Carbon performance affects the drift of the asset value process—through the transition channel—and the volatility—through the market-access channel. A firm with strong carbon performance has a higher drift and lower volatility, and therefore a lower default probability. This is a more rigorous foundation than a multiplier applied to a reduced-form PD, and it is the direction the framework should evolve.

4.7 The Role of Rating in the Integrated Mechanism

4.7.1 Rating as Identification

This chapter has developed a rating framework for technology firms that reads intangible assets, prices cash-flow volatility, and treats carbon performance as a rating factor with a defined transmission channel. The framework's purpose, in the larger argument of this book, is identification: it separates firms that can service debt from firms that cannot, on the basis of evidence rather than narrative.

The distinction matters for the integrated mechanism. A green lending program that does not identify creditworthy firms will accumulate losses, and the losses will discredit the program. A program that identifies creditworthy firms but does not buffer their risks will find that a single sector downturn produces a wave of defaults the rating framework could not prevent, because the framework assesses standalone risk and the downturn is a systematic event. This is why rating must be paired with insurance, the subject of the next part of the book.

The rating framework also provides the information the insurance layer needs. An insurer pricing a revenue-contingent instrument needs to know the borrower's through-cycle cash flow and the systematic factors to which it is exposed. The rating report contains both. An insurer pricing a performance guarantee needs to know the borrower's technical capability and the separability of its assets. The rating report contains both. Rating and insurance draw on the same information—through-cycle cash flow, systematic factor exposure, asset separability—and should be produced together.

4.7.2 The Limits of Rating

Rating cannot make a weak firm creditworthy. It cannot substitute for equity in a firm that has no cash flow. It cannot eliminate the systematic risk that a sector downturn will produce correlated defaults. And it cannot, by itself, channel capital to green innovation, because the firms that most need capital are often the firms the framework rates lowest—early-stage, pre-revenue, concentrated customer base.

The framework's contribution is to make the classification accurate and transparent, so that the policy instruments that follow—insurance, guarantees, guided funds—can be targeted at the specific failure the framework reveals. A firm rated BB− because of customer concentration needs a different intervention from a firm rated BB− because of cyclical exposure. The first needs a guarantee or contract-diversification support; the second needs a revenue-contingent instrument. A rating framework that produces a single score without the diagnostic components would obscure this distinction. The framework in this chapter produces the components, and they are the basis for the interventions developed in the chapters that follow.

4.7.3 The Transition to Chapter 5

The case of Lumen Grid Systems illustrated a general problem: a firm with adequate standalone credit quality can still fail if its exposure is concentrated in a single sector or a single customer. The rating framework assesses the firm; it does not assess the portfolio. A lender that holds twenty firms like Lumen, all exposed to the same storage build-out cycle, has a portfolio risk much greater than the sum of the individual ratings suggests.

Chapter 5 takes up that problem. It examines portfolio dependence and concentration in green lending, and it asks how a lender should measure and manage the correlation that the rating framework cannot see. The answer involves a shift from firm-level rating to portfolio-level loss distribution, and it sets the stage for the insurance layer, which is the instrument for transferring the systematic risk that diversification cannot eliminate.

Before turning to that, one point about the data. The framework in this chapter was applied to a single firm, and the judgments it required—the stress scenario, the separability score, the carbon weights—were made by an analyst with access to the firm's records. A lender applying the framework at scale needs a data infrastructure that collects the same information for every borrower, in a consistent format, and makes it available to the rating model and to the insurance underwriter. Concretely, that means five fields captured per borrower at origination and refreshed annually: a patent separability score with the licensing-market evidence behind it; verified Scope 1 and Scope 2 emissions in tonnes CO₂e; the share of revenue from standard versus bespoke products; the assignability status of each contract representing more than 5 percent of revenue; and the through-cycle cash-flow series used in the stress test, with its sector source. A bank that cannot produce those five fields for a technology borrower is not yet in a position to run this framework, whatever its rating policy says. Chapter 14 returns to the build-out in detail.

4.8 Conclusion

The chapter opened with two applications on a loan officer's desk. The precision-machining shop will be rated on its building and its forty years of steady cash flow, and the battery-software firm will be rated—under the framework developed here—on the through-cycle cash flow its patents and contracts generate, the separability of those patents, and a carbon score that reflects its fitness for the storage markets it serves. Lumen Grid Systems came out at BB−, not because the framework was lenient but because the stressed DSCR of 0.76 on the reduced structure was the binding fact, and no amount of environmental virtue changed it.

That result carries the argument forward. The rating framework identifies; it does not buffer. Twenty Lumen-type borrowers exposed to the same storage build-out cycle will default together in a downturn that no individual rating anticipated, and the instrument for that risk is not a better scorecard but a transfer mechanism—insurance, guarantees, revenue-contingent facilities—priced off the same cash-flow and factor-exposure data the rating report already contains. Chapter 5 begins with the portfolio loss distribution those twenty borrowers generate, and asks what a lender can do about the correlation it cannot diversify away.