The short version.

• The standard corporate-finance story — that a project's Internal Rate of Return (IRR) converges toward its Cost of Capital (CoC) over time, because capital rushes toward high-IRR opportunities until competition and diminishing returns pull them back down — is directionally correct and describes a real mechanism. It is not, on the empirical evidence, a fast or clean process.
• Research from Morgan Stanley's Counterpoint Global Insights (led by Michael Mauboussin) on the spread between Return on Invested Capital (ROIC, the realised-return analogue to a project's IRR) and Cost of Capital finds an average "persistence factor" of 0.79 across sectors (range 0.70-0.90) — meaning roughly four-fifths of any given year's ROIC-CoC gap is still present the following year, implying "fade rates" of only 0.10-0.30 per year, not the instantaneous margin-arbitrage the textbook story implies.
• Across the full 1970-2024 history of US public companies (excluding financials), the average ROIC-WACC spread was actually -0.8 percentage points — slightly negative, not converged to zero — with enormous dispersion around that average (25th percentile companies at -4.3 points, 75th percentile companies at +5.7 points). That's a genuine correction to the tidy "returns settle near cost of capital" framing: on this evidence, they settle into a wide, persistent distribution instead.
• What sustains a large, durable gap for specific companies is well studied separately: economies of scale, network effects, switching costs, regulatory advantages, and patents all function as barriers to entry that slow the fade rate specifically for the company holding them — the same "exceptions" logic the convergence story usually gestures at, but now with named mechanisms behind it.
• India's own 2026 sector data, covered separately on this blog, shows both halves of this story in the same year: cement and steel margins compressing as capital chases the same government-infrastructure-driven demand (convergence in progress), while toll roads' contractually inflation-indexed revenue functions as exactly the kind of regulatory moat that resists it, and aviation's and fertilisers' returns sitting below cost of capital is precisely the condition the "hurdle rate effect" predicts should trigger supply contraction — a testable prediction, not yet a confirmed outcome.

The textbook mechanism, and where it's accurate

The basic logic connecting a project's IRR to its cost of capital is standard, well-established corporate finance, and worth stating plainly before complicating it. When a type of project offers an IRR well above the going cost of capital for that risk class, capital moves toward it: more firms enter, competition for inputs and customers intensifies, and the combination of rising input costs and falling prices compresses the IRR available on subsequent projects of the same type. Symmetrically, when a project's expected IRR falls below the cost of capital — the hurdle rate a rational firm requires before committing capital — firms decline to invest, supply in that category shrinks, and the reduced competition eventually allows returns to recover back toward (or above) the hurdle rate. Net Present Value itself is defined around this same logic: a project's NPV is exactly zero when its IRR equals its cost of capital, which is the mathematical statement of "this project neither creates nor destroys value relative to what the capital could otherwise earn." None of this is wrong. It's the pace, and the completeness, of the convergence that the simple version overstates.

The capital-allocation/competition/diminishing-returns/hurdle-rate mechanism described here is standard corporate finance theory, consistent across mainstream treatments of NPV and IRR; this piece treats it as established background rather than a claim requiring its own primary-source citation.

What the empirical research actually finds: convergence is real, and slow

The most direct empirical test of how fast this convergence actually happens comes from work on the spread between Return on Invested Capital and cost of capital — ROIC being the realised, ongoing-business analogue of a project's IRR calculated at the outset. Morgan Stanley's Counterpoint Global Insights research group, associated with Michael Mauboussin, has published specifically on what it calls the "fade rate": the rate at which an above- or below-cost-of-capital spread decays back toward zero each year. Their empirical estimate, aggregated across sectors, puts the average persistence factor at 0.79, with a range of 0.70 to 0.90 depending on sector — meaning that on average, roughly 79% of this year's ROIC-cost-of-capital gap is still present next year. That implies fade rates of only 0.10 to 0.30 annually (average around 0.21), a meaningfully slower process than the "capital rushes in and margins collapse" framing suggests on its own. A fade rate of 1.0 would mean full reversion to the mean within a single year; the empirical range found here is nowhere close to that.

A separate piece of evidence from the same research programme, an IPO cohort study covering companies that went public between 1990 and 2022, tracked the ROIC-WACC spread from the IPO date forward and found it wide at listing, narrowing over the following years, and stabilising at a new, lower level around year five post-IPO — a specific, dateable timeframe for how long meaningful convergence actually takes for a typical newly-public company, rather than an assumption that it happens within a single planning cycle. And looking at the full historical distribution rather than any single cohort, the same research puts the average ROIC-WACC spread for all US public companies (excluding financials) across 1970-2024 at -0.8 percentage points — that is, on average, realised returns have sat slightly below cost of capital across more than five decades of data, not neatly converged to meet it, with the 25th percentile of companies at -4.3 points and the 75th percentile at +5.7 points. That spread between the 25th and 75th percentiles — a full 10 percentage points — is the real headline: the empirical distribution is wide and persistent, not a tight band clustered around zero the way a clean, fast-acting convergence mechanism would predict.

The 0.79 average persistence factor, its 0.70-0.90 sectoral range, and the resulting 0.10-0.30 (average ~0.21) fade-rate estimates are drawn from Morgan Stanley Counterpoint Global Insights' research on competitive advantage period and fade rates, reached via search-indexed summaries of the underlying PDF (morganstanley.com); direct fetch of the PDF was blocked in this research environment. The 1990-2022 IPO cohort study and its finding of spread stabilisation around year five post-IPO, and the 1970-2024 historical average spread of -0.8 percentage points (with 25th/75th percentile figures of -4.3 and +5.7 points), are from the same research programme, also reached via search index rather than a direct read of the source PDF. This piece could not independently verify these figures against Morgan Stanley's own document and reports them as found in search-indexed summaries, consistent with treating them as somewhat lower-confidence than a primary document read in full, though multiple independent search queries returned matching figures.

What actually slows the fade: named barriers, not just "exceptions"

The same research programme frames the ROIC-cost-of-capital spread, sustained over time, as the quantitative definition of what's often called a competitive "moat" — and identifies specific mechanisms that slow a company's fade rate rather than leaving "barriers to entry" as an unexplained residual category. Two broad sources of competitive advantage are identified: a consumer advantage (customers preferring a company's product or service independent of price, often via brand or habit) and a production advantage (a company producing at a genuinely lower cost than rivals can match). Concretely, this cashes out into named mechanisms: economies of scale, network effects, switching costs, and regulatory advantages, alongside more familiar items like patents. Each of these functions by making it harder, slower, or more expensive for a competitor to erode a specific company's above-cost-of-capital return, which is exactly why the fade rate for companies holding a strong version of one of these advantages sits at the low end of the 0.10-0.30 empirical range, while companies without any such advantage sit at the high end — converging toward cost of capital within a few years rather than persisting for a decade or more.

The consumer-advantage/production-advantage framing and the named barrier-to-entry mechanisms (economies of scale, network effects, switching costs, regulatory advantages) are drawn from Morgan Stanley Counterpoint Global Insights' "Measuring the Moat" research, reached via search-indexed summaries; direct fetch was blocked in this research environment. This piece's inference that companies with stronger moats sit at the low end of the 0.10-0.30 fade-rate range and weaker-moat companies at the high end is a reasonable reading of how the persistence-factor concept is defined in the sourcing found, but was not itself independently confirmed as an explicit stated finding, and is flagged as this piece's own connective reading rather than a directly quoted result.

Both halves of the story, in India's 2026 sector data

This blog's own recent survey of ICRA's 2025-2026 sector outlook notes across fourteen Indian industries happens to show both the convergence mechanism and its exceptions playing out in the same year, which is worth connecting directly rather than leaving as an abstract illustration. Cement and steel are the clearest case of convergence in motion: both sectors' demand growth leans on the same roughly ₹12.2 lakh crore government infrastructure capital-expenditure allocation, and in both, margins are compressing even as volumes grow — cement's OPBIDTA per tonne is projected to fall from roughly ₹900-950 to ₹820-870 between FY2026 and FY2027, and steel's operating margin is projected to hold roughly flat around 14.8% despite rising input costs, with sector leverage rising from 2.6x to about 3.0x. That's the textbook mechanism operating in real time: capital and capacity chasing an identifiable demand source, with margins giving ground as they do.

Toll roads are closer to the moat side of the story: inflation-linked toll-rate revisions, projected at roughly 3.2% for newer projects and 1.6-2.0% for older ones in FY2027, are a contractual, regulatory mechanism that insulates a specific, known revenue stream from the kind of open competitive erosion that would otherwise apply — not unlike the "regulatory advantage" category of moat described in Section 3, just delivered through a concession agreement rather than a patent. And aviation and fertilisers, the two of three sectors ICRA downgraded to Negative in 2026 for shared, externally-driven reasons (the West Asia conflict and evolving US tariff developments, covered separately on this blog), are sitting in the condition the hurdle-rate side of the convergence story describes directly: returns below cost of capital, which textbook logic predicts should eventually reduce supply (fewer new aircraft orders, reduced urea capacity utilisation, which ICRA's own aviation and fertiliser commentary already shows happening at the margin) until returns recover. Whether that recovery actually plays out on the empirical fade-rate timeline described in Section 2, roughly a handful of years rather than a single planning cycle, is a genuinely open, testable question this piece doesn't resolve — it's a prediction the theory makes, not yet a confirmed outcome in either sector.

The cement, steel, toll-road and aviation/fertiliser figures referenced here are drawn from this blog's own prior, separately-sourced reporting on ICRA's 2025-2026 sector outlooks, where each figure carries its own citation; they are not re-verified independently in this piece and should be checked against that earlier piece's own sourcing notes for their primary basis. The connection drawn here — framing these sector-specific dynamics as instances of the IRR/cost-of-capital convergence mechanism and its moat-driven exceptions — is this piece's own synthesis, not a claim ICRA itself makes in those terms.

What doesn't follow from any of this

None of this should be read as the underlying convergence mechanism being wrong — the empirical evidence in Section 2 shows real fade, real convergence, and a real, if slower-than-assumed, pull back toward cost of capital for most companies most of the time. What doesn't follow is treating that convergence as fast, complete, or uniform: an average persistence factor of 0.79 means a below- or above-cost-of-capital gap can plausibly last the better part of a decade for a specific project or company, particularly one holding any of the named moat mechanisms in Section 3, and the -0.8-percentage-point historical average spread across five decades of US public companies is itself evidence that "returns settle near cost of capital" is a much looser description of reality than a precise one. It also doesn't follow that India's cement, steel, aviation or fertiliser sectors will converge on any specific timeline just because the general mechanism predicts they eventually will; ICRA's own sector-specific forecasts, cited in Section 4 and sourced fully in this blog's earlier ICRA piece, are the better guide to near-term trajectory than the general theory alone. Readers using any specific figure in this piece for an actual valuation, capital-allocation, or hurdle-rate decision should verify it against the underlying Morgan Stanley Counterpoint Global Insights research directly, rather than relying on this summary of search-indexed excerpts.

Sources and caveats