A cost moat exists when the company can deliver equivalent value at structurally lower cost per unit. It can then price at rivals' cost and still earn a spread, or match prices and out-earn them. Fix two things in your head before the sub-types. (1) It's relative. Being big means nothing if a rival is big enough too. Your question is cost per unit versus the marginal competitor. (2) The moat often hides in low margins. A structural cost leader that shares its advantage with customers through lower prices shows up with thin margins, huge turns, and high ROIC. You met this in the DuPont lesson: margin × turnover drives returns, and margin alone doesn't. There are three sub-types:
- Scale: spreading fixed costs (plants, IT, brand spend, R&D) over more units, plus purchasing power. There's also an underrated one, density. A distribution network's cost per drop falls as stops per route rise, so regional density beats national sprawl.
- Process: a hard-won, difficult-to-copy way of doing things: accumulated yield learning, proprietary manufacturing know-how, a culture-embedded operating system. Toyota's TPS is the canonical case. It's been documented for decades and still hasn't been replicated, because it's spread across ten thousand small behaviors. Process advantages are real but erode as knowledge diffuses. Grade their duration conservatively unless the learning curve keeps moving, as it does at TSMC.
- Location/resource: geology and geography. Picture the lowest-quartile-cost mine, the quarry 30 km from the city (aggregates are nearly unshippable), the plant beside the port or the gas field. These last as long as the resource does, so check reserve life. Labor-cost arbitrage ("our engineers cost less") is location-flavored. But every rival hiring from the same pool shares it, which makes it an industry feature and not a firm moat. It's also why Indian IT's moat is switching costs and not wages.
Advertisement
How does DMart's cost advantage work?
Worked example (India): DMart (Avenue Supermarts). DMart claims to be India's structural low-cost grocer. Its mechanism is a designed system, and every piece reinforces the rest. (1) Buy cheap by paying fast. DMart pays vendors in about a week against an industry norm of 30 to 60 days. That converts its balance sheet into purchasing discounts, a use of capital most retailers can't afford. (2) Own the stores. No rent escalations eat margin, and sites get bought where land is cheap per catchment. (3) Cluster density. New stores open next to old ones, so warehouses, logistics, and store-brand awareness amortize across a tight radius. (4) Ruthless assortment. High-velocity everyday SKUs and few slow ones keep inventory turns high. (5) Every Day Low Price. Savings go back to customers, which drives the volume that feeds (1) to (4).
Illustratively, modeled on ~FY25 scale (pull the annual report), revenue is ~₹57,000 to 59,000 crore across ~400+ stores, or ~₹140 crore per store. Gross margin is only ~14 to 15%, against ~25 to 30% for conventional peers. Yet EBITDA margin is ~7.5 to 8%, inventory turns ~11 to 14× a year, and ROCE is ~19 to 20%+ (mid-20s pre-COVID) against the ~12% hurdle.
Look at that pairing again. DMart has the lowest gross margin in the industry and among the highest returns on capital. You can literally see the moat as cheapness plus turns. This is "scale economies shared," the flywheel from your revenue-machine work, where the customer gets most of the advantage. That sharing is what makes the advantage grow. Worked Example 1 quantifies why an entrant matching DMart's prices earns sub-hurdle returns. It's the moat expressed as the attacker's negative NPV.
How does TSMC's cost advantage work?
Worked example (global): TSMC. TSMC's claim is process power plus scale at the frontier of semiconductor manufacturing. Leading-edge chipmaking runs on a yield learning curve. A new node starts with a large share of dies defective, and the manufacturer that runs the most wafers learns fastest, because every defect analyzed feeds the process. TSMC runs the most wafers, so it reaches high yield first. So its effective cost per good die is lowest, so it wins the biggest customers (Apple, Nvidia), so it runs even more wafers. That's a cost advantage compounding through the learning loop. Capex of ~$30bn+ a year pays for it, and only the leader's cash flows can fund that. Rivals stumbled at successive nodes. Think of Intel's delays and Samsung's yield struggles at the leading edge. They didn't lack money. They lacked accumulated learning.
On illustrative ~2024 figures (pull the 20-F), revenue is ~$88 to 90bn. TSMC has ~60%+ of global foundry revenue and ~90% of leading-edge logic capacity. Gross margins run ~53 to 56% and operating margins ~42 to 46% as the price leader's supplier of last resort. ROIC is ~20%+ through a brutally cyclical, capex-devouring industry. Two flags, though. One is customer concentration. The other is geopolitical and location risk that no moat framework prices.
A moat protects returns from competitors. It can't protect them from history.
How do you test for a cost advantage?
Evidence trail for cost advantages (where to look). Compare cost per unit with named peers: cash cost per tonne, CASK for airlines, cost per transaction. Phase 5 gives each sector's unit. Read gross margin with the strategy in mind. A leader pricing at parity shows a higher margin than peers. A leader sharing scale shows a lower margin but higher turns and ROIC, so never read gross margin without the DuPont. Check payables and receivables terms against the industry. DMart's ~7-day payables are visible in the balance sheet. Look at capex per unit of capacity, and at utilization. Look for market-share gains at equal or better margins, since share bought with discounts is a false signal. Resource claims call for the reserve grade and location disclosures. And for a process advantage, ask whether the gap versus the #2 has persisted through at least one technology or capacity cycle.