NVR, Inc. (NYSE: NVR) — The Homebuilder That Owns No Land: Best Returns in the Industry, a Merely Fair Price, and Washington in Its Crosshairs

An independent equity-research note. All figures USD, IFRS/US-GAAP as filed, unless stated. Price reference: $6,829 close, 2026-06-26. Fiscal year ends 31 December. NVR has never split its stock, so per-share figures are large. Primary sources: FY2025 10-K (filed 2026-02-11), Q1-2026 10-Q (filed 2026-05-06), DEF 14A (filed 2026-03-17), and the 2023–2026 Form-4 corpus.

⚡ Claude’s Take

This block is the author’s own subjective opinion and general information only — not investment advice and not a recommendation. The detailed analysis that follows takes no position, names no price target, and carries no buy/sell; the single directional view in this piece is fenced inside this clearly-labeled block.

Verdict: HOLD at ~$6,829 / accumulate-on-weakness — the gold-standard compounder, but available at a merely fair price, mid-cycle earnings still rolling over. Medium conviction. Directional fair-value zone ~$6,800–8,000 (≈14–16x normalized ~$460–490 EPS / ~5.0–5.5x a compounding ~$1,380 book); accumulate ~$5,500–6,200 (the sector-drawdown discount, ~4.0–4.5x book); gives up the margin of safety above ~$8,000. Not a short.

NVR is, by the numbers, the best business in US homebuilding: a ~33% return on equity earned on an asset-light land-option model that owns almost no land (capex ~0.2% of revenue), a net-cash balance sheet, and a decades-long buyback that has shrunk the share count ~24% in five years with no dividend to subsidize it. The model is proven downside-protective — NVR stayed profitable in every year of the 2008 collapse while land-owning peers took billions in write-downs — because its maximum downturn loss is a forfeited option deposit, not an inventory impairment. But unlike the richest-ever (Howmet, Wabtec) or cheapest-ever (Dynatrace, Otis) names that make valuation the thesis, NVR is doing neither: it sits at the 47th percentile of its own ten-year valuation history — a genuinely fair price, not a screaming bargain and not a bubble.

What keeps this a HOLD rather than a table-pounding BUY is that the ~17x trailing P/E rests on trough-ish earnings that the market is already assuming will recover. Homebuilding gross margin has compressed from a 27.3% peak (2022) to 21.2% (FY2025) to 19.6% (Q1-2026), diluted EPS has rolled from $507 (2024 peak) to a TTM ~$407, and there is an idiosyncratic wrinkle the diversified majors escape: NVR’s Mid-Atlantic segment is 43% of homebuilding revenue with Washington D.C. as its single largest market, and into the 2025–26 federal-workforce reductions that segment’s Q1-2026 profit fell −51%. The entire valuation reduces to one question — is 19–20% the cyclical margin trough that recovers toward 22–24%, or a structural reset? At ~5x book (a premium fully earned by 2–3x the cohort’s ROE) you are paying a fair price for the gold standard, but with limited cushion. Framing: quality-compounder-at-a-fair-price, not deep value and not momentum — a high-beta (1.37 to the homebuilder factor) cyclical bouncing ~+23% off its May-2026 low but not washed out. Conviction flips bullish if HB gross margin re-expands through 21–22% over 2–3 quarters; bearish if it stalls sub-20% with deepening DC/Mid-Atlantic order softness. Tag: “the gold-standard compounder — on sale only when the whole sector is.”

📈 Stock Price Action — Five-Year Event Map

Text-only by design. The price move is a Fact; the attributed cause is Interpretation. No price target, no support/resistance, no chart-pattern reading. NVR has never split.

The arc. NVR is one of the great long-run compounders in US equities — from ~$5 in 1994 to an all-time high of $9,924 on 2024-10-18 — that has now rolled over with the housing cycle. It trades at $6,829, −31% off that ATH, inside a 52-week range of $5,563–$8,543, having bounced ~+23% off the $5,563 low set 2026-05-15. The five-year picture is a full cyclical round-trip: COVID boom, 2022 rate-shock crash, a near-tripling to the 2024 record, and a 2025–26 affordability-driven decline that found a floor in May.

Cycle narrative. (1) The 2020–21 leg was the pure pandemic housing boom — ~3% mortgages and a demand surge nearly tripled the stock to ~$5,959. (2) The 2022 rate shock inverted it violently: as mortgages doubled to ~6%, NVR fell ~38% to $3,670 even as FY22 gross margin peaked at 27.3% — the textbook builder pattern of the multiple collapsing on peak earnings. (3) From mid-2022 the “peak-rates” bet and the industry-wide mortgage-buydown machine restored affordability, orders re-accelerated, and the buyback shrank the float, carrying the stock ~170% to its $9,924 October-2024 all-time high at the apex of rate-cut optimism. (4) Through 2025 into early 2026 the stock ranged and rolled over as “higher-for-longer” rates capped the advance and EPS fell from the $506.76 FY24 peak. (5) The decline to the $5,563 May-2026 low was the fundamentals catching down — sticky ~7% mortgages, an affordability stall, and the NVR-specific overhang of Washington-DC-metro federal-workforce cuts (Mid-Atlantic segment profit −51% in Q1-2026). (6) The June bounce to $6,829 is a sector relief rally off that trough, supported by the Q1-2026 +7% order inflection and the 21st Century ROAD to Housing Act clearing Congress — basing off a cyclical low, not yet a re-established uptrend.

1. Executive Summary

NVR, Inc. is one of the largest US homebuilders, but it is structurally unlike any of its scale peers. It builds and sells homes (almost entirely pre-sold, never on spec) under three price-tiered brands — Ryan Homes (the first-time/first-move-up volume engine), NVHomes and Heartland Homes (move-up/luxury) — across 37 metros in 16 states plus Washington D.C., and it finances those buyers through a captive, servicing-released mortgage arm (NVR Mortgage) plus a title-services unit. In FY2025 it settled 21,915 homes at a $460,600 average price for $10,342M of revenue.

The defining fact about NVR is what it refuses to do: own land. Rather than buying and developing raw land like D.R. Horton, Lennar, or PulteGroup, NVR controls finished, build-ready lots through option contracts with forfeitable deposits — it put ~$925M of deposits behind ~180,100 controlled lots at year-end 2025 while owning only ~2,300 lots ($39.3M of raw land) outright. The third-party developer carries the land, the capital, and the development risk; NVR’s maximum loss on any deal is the forfeited deposit. This converts the most capital-intensive, cyclically-dangerous part of homebuilding into an off-balance-sheet, walk-away option, and it produces the most extreme returns profile in the sector: ~33% ROE in FY2025 (down from ~40% in 2024 and ~49% at the 2022 peak, but still 2–3x land-owning peers), a net-cash balance sheet, ~0.2%-of-revenue capex, and uninterrupted profitability through the 2008 collapse that bankrupted or crippled much of the industry.

The investment tension is not about business quality — it is the best business in homebuilding — but about price and cycle. NVR trades at ~16.8x trailing EPS / ~5.0x book / ~11x EV/EBITDA, which the a multi-year valuation-percentile dataset places at the 47th percentile of its own ten-year history: a genuinely fair price, neither rich nor cheap. The complication is that the ~17x multiple rests on trough-ish earnings: gross margin has compressed from 27.3% (2022) to 21.2% (FY2025) to 19.6% (Q1-2026) on rate-buydown incentives, lot-cost inflation, and affordability at multi-decade lows; diluted EPS has rolled from $506.76 (2024) to a TTM ~$407. Layered on top is an idiosyncratic concentration risk: NVR’s Mid-Atlantic segment is 43% of homebuilding revenue, Washington D.C. is its largest single market, and the 2025–26 federal-workforce reductions drove that segment’s Q1-2026 profit down 51%. The bull owns a fortress-balance-sheet, 33%-ROE compounder at a fair multiple with orders inflecting (+7% in Q1-2026) and a structural housing shortage underneath; the bear sees a no-moat commodity cyclical with falling margins, a deliberately slow unit grower, a price-insensitive buyback, and ~17x sitting on still-too-high earnings. This memo argues the franchise is genuinely the gold standard and the price is fair, but that the cyclical and DC-specific pressures warrant patience over aggression.

2. Business Overview

NVR builds and sells single-family detached homes, townhomes, and condominiums — almost entirely pre-sold (built to a signed customer contract, not on spec) — across 37 metropolitan areas in 16 states plus Washington D.C. It sells under three brands tiered by price point, not by any demand-side differentiation: Ryan Homes, the volume engine, aimed at first-time and first-move-up buyers and the only brand in all 37 metros; and NVHomes and Heartland Homes, smaller move-up/luxury brands confined to the Mid-Atlantic and Pittsburgh. In FY2025 NVR settled 21,915 homes (−4% YoY) at an average settlement price of $460,600 (+2%), with individual closings ranging from $170,000 to $2.3 million.

The company runs two reportable segments. Homebuilding generated $10,094M of FY2025 revenue (98% of total) and is the entire economic engine. Mortgage Banking, through NVR Mortgage Finance (NVRM), originates loans exclusively for NVR’s own homebuyers — a captive attach, not a standalone lender — and earned $152.0M of pretax income in FY2025 (~9% of consolidated pretax profit) at a steady 86% capture rate. Critically, NVRM sells every loan into the secondary market within ~30 days on a servicing-released basis, retaining no mortgage-servicing rights. That is a deliberate, conservative choice: NVR earns origination and gain-on-sale income (and funds the rate-buydowns that close its buyers) while carrying no MSR mark-to-market volatility and no servicing-advance risk — the kind of item that whipsaws other lenders. NVRM’s earnings are entirely derivative of homebuilding volume: when Q1-2026 settlements fell 22%, loan closings fell ~27% in lockstep. A small NVR Settlement Services unit brokers title insurance for fee income.

Geographic mix is the single most important orientation fact about NVR — and a double-edged one. Its four regional homebuilding segments are heavily Eastern/Mid-Atlantic, with no Sun-Belt-scale Texas/Arizona/Nevada presence of the kind that defines DHI and Lennar:

The Mid Atlantic alone is 43% of homebuilding revenue, and within it the Washington D.C. metro is NVR’s single largest market (~22% of total revenue as recently as 2021; NVR breaks out a metro only when it exceeds 10% of homebuilding revenue, which has historically applied only to Washington and Baltimore). NVR has diversified meaningfully over two decades but remains the most Mid-Atlantic-/DC-concentrated of the major builders — a concentration that intersects directly with the 2025–26 federal-workforce story (see the Industry and Changes sections).

Two operational features round out the picture. First, NVR manufactures its own building components: 10 leased + 1 owned panelized-component plants (~1.5M sq ft) running at only ~45% utilization in 2025, fabricating wall panels and roof trusses that compress on-site cycle time and labor cost — a genuine if modest efficiency edge most peers lack. Second, NVR uses independent subcontractors under fixed-price contracts for all on-site construction and employs only ~6,300 people. It pays no dividend and returns 100% of capital via buyback. NVR is also famous for what it does not do: it holds no earnings calls and issues no guidance — the absence of transcripts is a corporate feature, not a data gap.

Verdict. A large, conservatively run, pre-sold production homebuilder with a captive low-risk mortgage attach — but one whose economics are dominated by a single structural choice (the land-option model, discussed next) and whose geography is unusually concentrated in the Mid-Atlantic/DC corridor rather than the high-growth Sun Belt. Recurring revenue is nil; every home is a one-off, cyclical, commodity sale.

3. Industry Dynamics

US homebuilding is a structurally mediocre, deeply cyclical, commodity industry that has gotten incrementally better on the supply side — and NVR sits in it carrying one extra, idiosyncratic concentration risk its peers do not share.

Structure and consolidation. Homebuilding has consolidated hard from a fragmented cottage industry toward a handful of scaled publics — D.R. Horton (#1), Lennar (#2), PulteGroup, NVR, Meritage, KB Home, Toll Brothers, Taylor Morrison — who steadily take share because they out-buy materials, out-source land, and out-finance buyers relative to local competitors. This is a genuine multi-decade share tailwind favoring the scaled survivors. But consolidation of a commodity industry raises returns only modestly — the product is undifferentiated, buyers have zero switching costs, and demand is exogenous and rate-gated. Getting bigger creates cost advantage, not pricing power.

The cycle. The sector ran the canonical boom-bust: the 2020–22 pandemic boom (sub-3% mortgages, peak margins — NVR’s GM hit 27.3% in FY2022), the 2023 rate shock as 30-year mortgages spiked past 7%, a 2024 recovery on rate-cut hopes, and a 2025–26 affordability stall that is the current reality. The late-cycle signatures are everywhere: NVR’s homebuilding gross margin fell 27.3% (FY22) → 23.7% (FY24) → 21.2% (FY25) → 19.6% (Q1-2026), FY25 new orders fell 10%, the cancellation rate rose to 17% (from 13–14%), and Q1-2026 net income fell 34% YoY. Industry-wide, mortgage rates remain ~6.5–7% (NAHB expects 6%+ throughout 2026); affordability is at multi-decade lows; new-home sales printed ~580K SAAR (−7%) with 10.3 months’ supply (well above a balanced ~6); NAHB builder sentiment sits at 35; and 62% of builders are offering incentives — the 15th consecutive month above 60%. The defining margin dynamic is the mortgage-rate-buydown incentive war: builders spend heavily to buy down customers’ rates to close affordability-constrained buyers (Lennar’s incentives hit ~13.8% of revenue), compressing margins industry-wide. NVR is exposed to the same pressure, though its pre-sold model and lack of spec overhang make its compression somewhat milder.

The long-run demand floor. Against the near-term affordability ceiling sits a genuine structural tailwind: the US is chronically underbuilt — estimates of the deficit cluster meaningfully positive (NAHB ~1.5M units, Freddie Mac ~3.7M, Zillow ~4.5M, NAR ~5.5M). A decade of under-construction relative to household formation, plus the “rate-lock” effect freezing existing-home supply, structurally diverts share toward new construction and gives builders a long-run volume floor — though it is gated by affordability in the near term and confers no pricing power.

The DC-metro / federal-employment angle — NVR’s idiosyncratic risk. This is where NVR diverges from the diversified national builders. With ~43% of revenue from the Mid-Atlantic and Washington D.C. as its single largest market, NVR is uniquely exposed to the 2025–26 federal-workforce contraction. The data already deteriorating: DC-area for-sale inventory up ~50% (vs ~30% nationally); roughly 40% of DC-area agents reporting a client transacting specifically because of federal layoffs/buyouts; ~22,100 federal jobs lost across DC/MD/VA; and NVR’s own Q1-2026 Mid Atlantic segment profit down 51% — the worst of any segment. For a homebuilder whose largest market depends on high-income federal and federal-adjacent (contractor, legal, lobbying) employment, a sustained DC income/confidence shock is a company-specific demand headwind the Sun-Belt-diversified peers do not carry. (The DC market data is Fact; the magnitude of the hit to NVR’s orders is an Open Question until regional order data isolates it — though the −51% Q1 segment profit is at least directionally consistent.)

Marathon capital-cycle read. Homebuilding is mid-to-late cycle in a managed, shallow downturn. The crucial nuance: capital is not being violently destroyed. The public builders entered with fortress balance sheets, and the land-light model lets the whole industry throttle starts and walk away from options rather than dump owned land at a loss. That is double-edged — it caps the downside (no 2008-style capitulation low, no distressed-builder wipeout) but also caps the upside, because there are no washed-out competitors for survivors to harvest and the incentive war can persist. NVR’s own behavior is the capital cycle in microcosm: it grew its controlled-lot pipeline +11% into the weakness, deploying cheap deposits to lock supply while developers are capital-starved.

Verdict. A structurally mediocre, deeply cyclical industry — “a good house on an average street.” Demand is exogenous, rate-gated, commodity-priced with zero demand captivity; returns mean-revert hard. The genuine improvements are all supply-side: consolidation to scaled publics, the land-light model dampening the capital cycle, and chronic underbuilding providing a long-run demand floor. Current cycle position: late-bust / affordability-stalled, margins compressing, with no near-term rate relief sufficient to break the incentive war — plus, for NVR specifically, an added idiosyncratic DC-metro/federal-employment overhang the diversified peers escape.

4. Competitive Position

Every homebuilder turns dirt into houses. What makes NVR singular is that it refuses to own the dirt. This is not a financing nuance — it is the entire economic identity of the company, and the source of the only genuine, durable competitive advantage anywhere in US homebuilding.

The mechanics. NVR “generally does not engage in land development.” It controls finished, build-ready lots through fixed-price Lot Purchase Agreements (LPAs) with third-party developers. Under an LPA, NVR puts down a forfeitable deposit — cash or a letter of credit, typically up to 10% of the lot price — for the option (not the obligation) to buy finished lots just-in-time as it sells homes. The developer selects the project, obtains entitlements, and finances the land and horizontal development “with no support or guarantees from us.” NVR’s own language is precise: “We may, at our option, choose for any reason and at any time not to perform under these LPAs… Our sole legal obligation and economic loss for failure to perform… is limited to the amount of the deposit.” NVR carries no specific-performance obligation, no completion guarantee, and the developers’ creditors have no recourse to NVR’s general credit. It converts the most capital-intensive, slowest-turning, most cyclically-dangerous part of homebuilding — land — into an off-balance-sheet, walk-away option.

The scale of it. At 12/31/2025 NVR controlled ~180,100 lots (169,250 under LPAs), backed by only ~$925M of deposits (≈8% of implied lot value). It directly owned a trivial ~$39.3M of land (~2,300 lots). Against ~22,000 settlements a year, 180,100 controlled lots is roughly an eight-year option pipeline controlled for under $1B of deposits — versus the $20–30bn of owned-and-developed land that DHI and Lennar carry on balance sheet to control comparable supply.

The three consequences — quantified.

1. A tiny balance sheet and elite returns. Because NVR neither buys raw land nor funds development, its capital intensity is a fraction of peers’. FY2025 capex was ~$24.5M — ~0.2% of revenue. Inventory was just ~$1.7B on $5.9B of total assets, the company holds ~$1.0B net cash, and it turns its asset base far faster than land-owners. The outcome is the legendary figure: ROE ~33–35% in FY2025, ~40% in FY2024, ~49% at the FY2022 peak — roughly double to triple the through-cycle ROE of land-owning peers. NVR earns these returns not by taking more risk, but by taking less.

2. No land impairments in downturns — the GFC proof. When demand collapses, a land-owner is trapped with billions in raw land and work-in-process bought at peak prices that must be written down. NVR simply walks away from its options and forfeits the deposit — its maximum loss per community is the ≤10% deposit, not the whole lot. In FY2025’s soft market, total contract-land-deposit impairments were only ~$75.9M — a rounding error against $10B of revenue, the asset-light analog of what would have been a far larger inventory write-down for a land-owner. The historical proof: NVR stayed solidly profitable in every year of the 2007–2011 collapse while Pulte, Lennar, Hovnanian, KB Home, and most of the industry took billions in write-downs and several flirted with insolvency. (The impairments are Fact; the causal attribution to the option model is Interpretation — but it is the mechanism management designed.) The model is a permanent embedded put on the land cycle.

3. The trade-off — thinner boom margins and ecosystem dependence. The model is not free. In boom years a land-owner captures the land-appreciation profit between raw-land cost and finished-lot value; NVR, buying finished lots at market, gives up that profit and runs structurally thinner gross margins in a boom. The model’s edge shows up in returns and downside protection, not peak margin. More importantly, NVR is dependent on a healthy third-party land-developer ecosystem. Its 10-K flags directly that there is “no assurance an adequate supply of building lots will continue to be available… on terms similar to those available in the past.” When development capital is scarce and entitled lots are tight, NVR must either pay up for options (lot cost is its cited #1 driver of 2025 margin compression) or accept slower growth.

Is the edge eroding as peers go land-light? This is the key pressure-test, and the honest answer is narrowing at the contractual margin, intact at the execution margin. DHI now options ~77% of its lots via its 62%-owned developer Forestar; Lennar spun its land bank into Millrose Properties in early 2025 and options ~98% of homesites. The contractual form of the option model is being copied across the industry — which is why land-light cannot be called a proprietary moat on its own. But the returns tell the real story: NVR earns 30%+ ROE on the model that DHI (~12–15% ROE) and Lennar (8.6% ROE, 6.8% ROIC — below cost of capital) earn far less on. The peers are imitating the structure without replicating the outcome, because NVR’s edge is the compounded execution layered on the contract — decades of developer relationships, Mid-Atlantic buying density, captive component plants, pre-sold (never spec) discipline, and an absolute refusal to chase volume at bad returns. Lennar even pays a new recurring takedown fee to Millrose that NVR never incurs because it never owned the land. The land-light model is becoming table-stakes; NVR’s 30-year execution of it is not.

Greenwald tests. Name the moat: a supply-side cost advantage + capital-efficiency edge (economies of scale / capital discipline), emphatically not brand or switching costs — a home is the most infrequent considered purchase a household makes, and “Ryan Homes” is a price-tier label, not a demand-captivity asset. ROIC test: NVR crushes it — 30–40%+ through-cycle ROE on a net-cash balance sheet, roughly double-to-triple peers, with uninterrupted GFC profitability. Market-share-stability test: NVR passes via discipline rather than aggression — it deliberately grows slower than DHI/LEN and has shrunk rather than overpay for lots that would dilute returns; its ~22,000 settlements are ~1/5 of DHI’s ~85–90,000 precisely because it optimizes for return on capital, not units. The question that raises — is the growth-discipline a feature or a limit? — answers both ways: a feature for return quality and downside protection (NVR has never had a near-death cycle, and per-share compounding is elite), a limit on absolute scale.

Verdict. NVR has the only genuine, durable competitive advantage in US homebuilding — validated by 30%+ through-cycle ROE and uninterrupted GFC profitability. It is a real but narrow advantage in Greenwald’s terms — economies of scale without demand captivity, in a commodity industry — which caps how wide it can ever be, but the execution moat is wider and more durable than the now-replicated land-light structure suggests, because the peers copying the contract are not copying the returns. The two things to watch: the erosion of the model’s exclusivity as the industry goes land-light, and the DC-metro/federal concentration that makes NVR’s largest market uniquely vulnerable in 2025–27.

5. Growth History and Forward Opportunities

The compounding is real; the unit growth is not. NVR’s headline record is the stuff of legend — diluted EPS from roughly $195 (2018) to a $506.76 peak (2024), a never-split share trading north of $9,000 at its top. But the engine is not what a casual reader assumes. NVR is not a unit-growth story. Settlements have been functionally flat for six years, oscillating inside a 19,800–22,800 band with the cycle and showing no secular trend: ~19,761 (2020) → 21,540 (2021) → 22,732 (2022) → 20,662 (2023) → 22,836 (2024) → 21,915 (2025). Over 2018→2025 settlements grew ~17% total — about 2% a year — while diluted EPS grew ~124% (≈12% a year). The wedge between a 2%-a-year unit business and a 12%-a-year per-share compounder is the whole story.

EPS decomposition, 2020→2025 (+89.7% diluted EPS). Three levers, ascending in importance: units +10.9% (the least important, and cyclical); average selling price +20.6% (the post-COVID price step-up, now stalling — Q1-26 new-order ASP −2%); the cyclical gross margin (which swung net margin and is now reversing); and buyback: shares 3.70M → 2.80M, −24.3%, which mechanically converts a given level of net income into ~32% more EPS — roughly one-third of the entire EPS gain. In plain terms: NVR’s per-share compounding is driven more by buyback, ASP, and the margin cycle than by selling more houses. You are not buying a unit-growth machine; you are buying an exceptionally high-return capital-recycling machine whose share count shrinks ~4–5% a year.

The forward volume lever is community count, not absorption. With per-community sales pace falling on affordability, NVR’s growth lever has shifted to opening more stores. Average active communities rose +8% in Q1-2026 (432 vs 401), which is what turned a roughly −1% absorption rate into +7% new orders — all of the order growth came from more communities, none from selling faster. Behind that sits the controlled-lot pipeline of 180,100 lots (+11%) — NVR leaning counter-cyclically into cheap option deposits. Community growth is the cleanest forward indicator to track; if it stalls, volume stalls.

Forward opportunities, pressure-tested. (1) Southeast / Sun-Belt expansion — the South East is now 26% of homebuilding revenue and NVR’s fastest community-count grower, but it is precisely where margins have collapsed (segment GM 18.3%, profit −48% in FY25) as Florida/Carolinas oversupply meets weak absorption; the geographic diversification is strategically right but being executed into the worst regional pricing environment. (2) Structural housing shortage (1.5M–5.5M units) — a real multi-year demand floor, but it has coexisted with falling sales for two years because the binding constraint is affordability. (3) Scaled-builder share gains — genuine and favors NVR. (4) Buyback — with reinvestment needs trivial, returning ~all FCF via repurchase remains the default per-share growth lever.

Is “slow grower” a bug or a feature? It is deliberate. NVR is an ROE-maximizer, not a volume-maximizer — it will shrink rather than overpay for lots or chase share at sub-target returns. At ~22,000 settlements NVR is a fraction of DHI’s ~90,000 or LEN’s ~80,000, yet it earns ~33% ROE versus their mid-teens. The unit growth is modest; the quality of each incremental unit — capital-light, impairment-proof, ROE-accretive — is unmatched.

Verdict — high-quality, low-quantity growth. Growth quality is best-in-industry (every unit is capital-efficient and the per-share engine compounds through buyback even with flat units). Growth quantity is modest and cyclical — there is no secular unit-growth thesis, and the ASP and margin tailwinds that flattered 2021–24 EPS are now headwinds. Underwrite NVR as a high-return compounder of capital, not a grower of homes.

6. Financial Quality

The model in one line: NVR is a ~$10bn-revenue, ~33%-ROE homebuilder that earns elite returns not from operating leverage but from a capital-light land-option model — controlling finished lots through forfeitable deposits instead of owning raw land, which strips the capital base to the bone and lets a mid-teens operating margin throw off a ~33% return on equity. The entire financial story right now is gross-margin compression.

Revenue and the volume rollover. Revenue was $10,342M in FY2025, −2% from FY2024’s $10,544M. The decline is a settlements story (21,915, −4%) at a $460.6K average price (+2%). The forward problem is in the order book: FY25 New Orders fell 10% to 20,410 units, and backlog ended down 15% in units / 16% in dollars ($4,008M). Because settlements lag orders by 2–4 quarters, that thin entering backlog drove a violent Q1-2026: revenue −22%, settlements −22%, net income −34% to $198.4M ($67.76/share). Encouragingly, the leading indicator turned — Q1-2026 New Orders rose 7% (on +8% communities) and the cancellation rate improved to 14% from 16%. So the order book is inflecting up even as the income statement, reflecting last year’s weak orders, is still falling. The FY25 cancellation rate of 17.0% (vs 14.2% FY24, 12.8% FY23) was the highest in years — a demand-stress tell.

The margin compression — the central number. The headline figure is the homebuilding gross margin: 21.2% FY2025, down from 23.7% FY2024 and 24.3% FY2023 (27.3% peak FY2022), and 19.6% in Q1-2026 (vs 21.9% Q1-2025). Management attributes the erosion to (i) higher lot costs (NVR buys finished lots at market; developer pricing inflated through the boom — the cited #1 driver); (ii) affordability-driven pricing pressure, i.e. rate-buydown incentives and concessions against a ~7% mortgage backdrop; and (iii) ~$75.9M of contract-land-deposit impairments in FY2025 (vs $7.2M FY24) — the cost of walking from options that no longer pencil. The segment data localizes the damage: South East gross margin collapsed to 18.3% from 22.3% (segment profit −48%) on Florida/Carolinas oversupply; Mid Atlantic fell to 23.3% from 25.0%; North East held best at 25.5%.

SG&A and below-the-line. Homebuilding SG&A was flat at $599.7M (~5.9% of HB revenue — slight deleverage on lower volume). D&A is trivial ($24.5M) and SBC modest ($69.2M). Capex was $24.5M (~0.2% of revenue) — what asset-light looks like on the income statement.

Returns — compute them correctly. third-party financial databases’s reported return_com_eqy of 8.5% and pr_to_book of 1.29x are both garbage (treasury-accounting artifacts). Net income $1,339.8M on average equity of ~$4,037M is a 33.2% ROE (34.7% on ending equity), down from ~40% in FY2024 and ~49% peak in FY2022. The level is elite and the cause is the model — high inventory turnover on a tiny capital base — not operating leverage. The decline is overwhelmingly margin-driven, not equity-bloat: in FY2025 the buyback ($1,833M) actually exceeded FCF ($1,097M), so retained earnings did not balloon the equity base; ROE fell because the numerator (margin) compressed. The risk to watch: if volume stays soft and the buyback can no longer outrun FCF, cash builds and the equity base grows, compounding the margin drag on ROE.

Quality of earnings — clean, with three caveats. NVR’s earnings are unusually high-quality: no goodwill, no acquired-intangible amortization, KPMG unqualified, conservative accounting (it expenses deposit impairments quarterly). Three things to normalize: (1) the tax rate is a real EPS swing factor — 23.96% (FY25) vs 20.50% (FY24) vs 17.46% (FY23); the low prior-year rates were stock-option excess tax benefits ($28.3M FY25 vs $95.1M FY24), so FY2024’s “peak” $506.76 EPS was flattered by ~$67M more option-tax windfall than FY2025, and the peak-to-trough EPS decline overstates the operating deterioration somewhat. (2) Reported OCF is noisy — FY2025’s $1,121M was flattered by a $335M inventory drawdown and drained by a $200.7M deposit build; earnings-quality FCF is better proxied at ~$1.0–1.1bn. (3) Diluted dilution is real — basic EPS $462.00 vs diluted $436.55, a ~5.5% wedge; diluted share count (~3.07M) is meaningfully above shares outstanding (2.80M) because of in-the-money options.

Verdict. A genuinely elite, durable, capital-light return engine — but a high-ROE turnover story, not a scale-economics story; economics do not visibly improve with scale (margins are set by lot costs and affordability, not by NVR’s size). The franchise is cyclically, not structurally, impaired: margins are at a multi-year trough (19.6% in Q1-26) while orders inflect up. The open question for valuation is whether 19–21% is the new gross-margin floor or a cyclical bottom that recovers toward 24%+.

Balance sheet. NVR runs one of the most conservative balance sheets in the sector — its entire asset base is ~$5.9bn, against $20–30bn of land inventory at DHI or LEN. Cash $1,916M; inventory $1,724M; net contract-land deposits $851M; mortgage loans held-for-sale $572M; total assets $5,857M. The only debt is $909M of 3.00% senior notes locked until May 2030 (fair value $852.9M, trading below par as rates rose — cheap, long-dated money NVR would never replicate today), leaving ~$1.0bn net cash; current ratio ~5.5x; the $300M revolver and the (non-recourse) NVRM warehouse facility both undrawn. Book value per share is ~$1,381 (equity $3,864.9M ÷ 2,799,387 shares) — use ROIC’s pr_to_tang_bv (~5.5x), not its bogus 1.29x. The off-balance-sheet land-option exposure (~169,250 lots controlled via LPAs, ~$925M at-risk deposits, with a liquidated-damages cap and no recourse to NVR’s general credit) is the model’s defining feature and is precisely what let NVR stay profitable through 2008. A fortress — the strongest balance sheet in the peer group, deliberately under-levered, with the land risk pushed off-balance-sheet and capped.

7. Capital Allocation

NVR’s capital-allocation policy is famously singular: no dividend, ever; 100% of shareholder return via buyback. The record is a share-count compounding machine — shares outstanding fell from 3.70M (2020) to 2.80M (2025), −24% in five years, against cumulative lifetime treasury stock of $15.7bn.

Two observations drive the assessment. First, the buyback routinely exceeds free cash flow — $1,833M vs $1,097M FCF in FY2025; $2,058M vs $1,345M in FY2024 — funded by drawing down the cash pile (cash fell from $2,561M at YE24 to $1,884M at YE25 to ~$1,700M at Q1-26). This is sustainable while the cash hoard and the 3% debt cushion the difference, but it is not a perpetual-motion machine if FCF stays depressed. Second, the program is cash-flow-paced, not price-disciplined. NVR spent its most dollars ($2.06bn) in FY2024 at its highest average price (~$8,011, within ~20% of the $9,924 all-time high), and is now spending fewer dollars at lower prices. A value-disciplined buyer would lean harder into the $5,563 May-2026 low and less into all-time highs; NVR instead spends roughly all excess liquidity every year regardless of valuation. The buyback has been powerfully accretive over the cycle — retiring shares of a >30% ROE business compounds value — but it is momentum-following, and the average price paid simply trails the stock up.

Elsewhere, capital allocation is disciplined to the point of austerity: essentially no M&A (NVR has grown organically for decades, avoiding the value-destructive land-and-builder acquisitions that litter peer histories), no equity issuance except option exercises, and only a small new drift toward JV/land-development spend ($47.6M in FY2025 vs $3.1M in FY2024) to secure lot supply in tight markets — a minor deviation from pure asset-light, worth monitoring but not yet material.

Verdict — intelligent over any multi-decade window. Share-count compounding, a fortress balance sheet, and the avoidance of bad deals are the heart of NVR’s per-share record. The fair critiques are narrow: the buyback is price-insensitive, and the rigid no-dividend/all-buyback policy offers no flexibility. Both are defensible given that the asset-light model genuinely needs almost no reinvestment capital, so returning ~all FCF is the right default — but management leaves alpha on the table by not buying counter-cyclically.

8. Changes and Headwinds — Last Two Years

The DC-federal concentration is the genuinely new, NVR-specific headwind, and the change that most differentiates NVR’s near-term setup from DHI’s and LEN’s. With 43% of homebuilding revenue in the Mid-Atlantic and the single largest market in Washington D.C., NVR is uniquely exposed to the 2025–26 federal-workforce reductions. The Q1-2026 evidence is already stark — Mid Atlantic settlements −31% and segment profit −51% — against a backdrop of ~22,100 federal jobs lost across DC/MD/VA and Mid-Atlantic active listings up ~42% YoY. Geographically diversified, Sun-Belt-weighted peers do not carry this exposure. It is an idiosyncratic bear leg layered on the cyclical margin pressure everyone shares. (One correction to the data feed: the reported HQ “relocation” is unconfirmed — NVR remains in Reston, VA with no relocation 8-K on file; treat as a non-event.)

Verdict — mixed: structurally neutral-to-positive, cyclically and idiosyncratically negative. Strengthening the thesis: management continuity with the asset-light model fully intact, the founder overhang cleanly removed, a board upgraded with exactly the housing-finance and large-cap operating expertise the moment calls for, and external validation of the sector’s long-run value from Berkshire’s $8.5bn Taylor Morrison deal. Weakening it (near term): an earnings/margin rollover with no clear trough yet, an active idiosyncratic DC-federal demand shock in NVR’s core market, a price-insensitive buyback, no rate relief in sight, and housing legislation that is supply-positive in theory but immaterial to 2026 earnings. Net: the structural thesis is intact, but the next 12–24 months are cyclically and idiosyncratically pressured.

9. Risk Analysis

10. Valuation Discussion (Embedded Expectations)

No price target, no recommendation. The figures below characterize what the market is pricing.

The spot setup. At the $6,829 close, ~2.71M shares outstanding (after the Q1-2026 repurchase) carry a market cap of ~$18.5bn, and against ~$0.8–1.0bn of net cash, an enterprise value of ~$17.6bn. On that base:

Own-history is the core framing, and it says “fair.” The a multi-year valuation-percentile dataset places NVR at the 47th percentile of its own ten-year history on a composite basis — squarely mid-range — with P/E at the 49.8th percentile, P/B at the cheaper-end 33.2nd, and P/S at the 57.9th. This is the deliberate contrast to richest-ever names (Howmet, Wabtec at the 90th-plus percentile) and cheapest-ever, left-for-dead names (Dynatrace, Otis in single digits): NVR is doing neither. Its P/E has ranged from ~8.8x at the 2022 recession-fear trough to ~17x in the 2020–21 boom; today’s ~16–17x is upper-middle. EV/EBITDA has swung from ~6x (2022) to ~12.4x; today’s ~11x sits near the upper end. Not cheap on its own history, but nowhere near the extreme that would make valuation the thesis.

The homebuilder valuation paradox makes the trailing multiple a trap. Builders screen cheapest on P/E at cycle peaks (peak margins, peak earnings, low multiple) and dearest at troughs. NVR’s ~16–17x trailing multiple sits on trough-ish earnings — HB gross margin compressed to 19.6% in Q1-2026, and the Street models a seventh down-year in FY26 (~$413, −5.3% vs the $507 FY24 peak). The governing question is the multiple on normalized, not trailing, earnings. Reverse-engineering it: at ~16.8x/$407, the market is not pricing a permanent trough. If normalized HB margin recovers toward 22–24% as the incentive war eases — implying normalized EPS ~$460–500 — today’s price is ~13.5–15x normalized earnings for a 33%-ROE, net-cash compounder shrinking its share count ~4–5%/yr, a reasonable price. If instead 19–20% is the new structural reality, today’s price is ~17–18x a number that does not recover, and the margin of safety is thin. The market is underwriting a cyclical margin recovery plus continued buyback compounding.

Peer comparison — the P/B premium is earned, not a screening error. NVR’s ~5.0x book screens optically extreme against a cohort at 1.1–2.4x:

The 2–4x P/B premium is the mechanical consequence of NVR earning 2–3x the cohort’s ROE (~33% vs 8.6–15%). On the Gordon relation P/B ≈ (ROE − g)/(r − g), a 33%-ROE business should trade at a multiple of book a 9%-ROE business cannot — NVR’s premium is internally consistent with its returns, not a market error. The cleaner cross-sectional lens is EV/EBITDA and normalized P/E, where NVR sits ~11x / ~16–17x — a quality premium to PHM/TOL/MTH but in line with DHI and LEN despite far superior returns and balance sheet. On that lens NVR is arguably the best-quality name in the group not trading at a quality premium.

Normalized-earnings scenarios (anchored on a compounding ~$1,381 book and the gross-margin debate):

Verdict. NVR is a fairly valued best-in-class operator, with the entire valuation case hinging on whether the Q1-2026 19.6% gross-margin print is the cyclical trough or a structural reset. The high P/B is earned by elite ROE and should not be read as expensive; the ~17x trailing P/E is the more honest worry, because it rests on trough-ish earnings the market is already assuming will recover. Limited margin of safety at the current price, but no obvious overvaluation.

11. Variant Perception

Consensus belief. The sell-side rates NVR Hold (roughly 2 buy / 4–6 hold; average targets ~$7,000–8,200, BofA $8,600) and models a seventh consecutive down-EPS year in FY26 (~$413). The shared view: NVR is the highest-quality, highest-ROE homebuilder with a fortress balance sheet, but a slow unit-grower in a rate-pressured cycle, at a full-but-fair multiple, with no catalyst until mortgage rates fall — own it, don’t chase it. The market treats NVR less as a stock-specific story than as a high-beta housing vehicle: a third-party factor model assigns an Industry: Home-Construction beta of 1.37 with R² 0.73 — ~73% of NVR’s return variance is the homebuilder factor, not idiosyncratic alpha; its factor twins are the whole cohort (TOL, PHM, DHI, LEN) plus the ITB/NAIL ETFs.

The strongest bull case. By the numbers the best business in the industry: a ~33% ROE on an asset-light model that owns almost no land (capex ~0.2% of revenue), a net-cash balance sheet, and a relentless buyback that shrank the share count ~24% in five years with no dividend to subsidize it. The model is proven downside-protective — profitable every year of the GFC while peers took multi-billion impairments. It sits atop a structural housing deficit (1.5M–5.5M units) and orders just inflected +7% in Q1-2026 (on +8% communities). Crucially, the valuation is fair, not stretched — 47th percentile on its own history — a rare chance to own the gold standard without paying a peak multiple. If margins are merely cyclically depressed, normalized EPS power of ~$490–525 makes today ~13–14x, cheap for the quality.

The strongest bear case. Strip away the reverence and NVR is still a no-moat commodity cyclical — the buyer has zero switching cost, the product is undifferentiated, and the high returns are an operating-discipline edge that DHI (Forestar) and LEN (Millrose) are now actively copying. Margins are still falling (21.2% FY25 → 19.6% Q1-26) with no rate relief in sight, so ~17x may sit on still-too-high earnings. NVR carries an idiosyncratic concentration the diversified majors lack — Mid-Atlantic 43% of revenue, DC the largest market, Q1-26 Mid-Atlantic profit −51% into federal RIFs. It is a deliberately slow unit grower ceding share, and its buyback is price-insensitive (most dollars spent at the highest prices). At ~5.0x book it screens optically expensive with no valuation cushion if returns re-rate.

The 3–5 assumptions that matter most. (1) Gross-margin trajectory — does HB margin trough at ~19–20% and recover toward 22–24%, or has the buydown war plus rising lot costs reset it lower? (the whole valuation reduces to this). (2) DC-metro/federal-jobs depth — transient air-pocket or multi-year regional drag NVR must wear alone? (3) Mortgage-rate path — a move toward ~6% re-accelerates orders and margins; ~7.5% deepens the trough. (4) Lot-supply discipline — can NVR keep sourcing finished lots without paying up enough to erode the very margin advantage that justifies the premium? (5) Buyback discipline — does management lean into drawdowns or keep spending all liquidity regardless of price?

Falsification. The bull is falsified if HB gross margin stalls sub-20% for two-plus more quarters with deepening DC/Mid-Atlantic order softness — proving the reset is structural and the concentration risk is realizing. The bear is falsified if margin re-expands through 21–22% over 2–3 quarters while orders hold the +7% inflection — confirming a cyclical trough behind a 33%-ROE compounder at a fair multiple.

Positioning read. The tape supports a “high-quality cyclical mid-correction, not a washed-out low” reading. NVR is a lifetime +13.5%/yr and ten-year +14.9%/yr elite compounder, presently in a drawdown: y1 −6.2%, m6 −14.5%, −31% off the October-2024 ATH (rs_peak −31%). The recent m3 +25.6% is the bounce off the May-2026 low — a sector relief rally off a trough, not the capitulation washout (the May low was ~4x book, well above the 2022 ~6x-EBITDA trough) that marks a generational entry. The 1.37 home-construction beta means consensus is expressing a macro/rate bet through this name; if rates ease, NVR’s quality leverage works for the holder, but the stock will move with the group until then. Evidence that consensus may be slightly offsides on durability — treating a 33%-through-cycle-ROE franchise as just another rate-levered builder — but not evidence of a mispriced, abandoned name.

12. Fact vs. Interpretation

13. Open Questions

1. Is 19–20% the gross-margin trough or a structural reset? The single most important unknown — the whole valuation reduces to it.
2. How deep and how long is the DC-metro/federal-jobs drag? Transient air-pocket or multi-year regional impairment of NVR’s largest market?
3. Can NVR source enough finished lots without eroding its margin advantage as development capital stays scarce and it leans counter-cyclically (+11% pipeline)?
4. Does management ever lean the buyback into drawdowns rather than spending all liquidity at all prices?
5. CEO Bredow’s strategic stamp — any deviation from the pure asset-light, ROE-maximizing, no-guidance playbook (e.g., the small new JV/land-development spend)?
6. Does the deliberate slow-growth posture cost NVR relevance as DHI/LEN compound unit scale and purchasing power?
7. Normalized through-cycle settlement level — is ~22,000 the durable base, or can community-count growth lift it structurally?

14. What Must Be True

For the bull case (NVR compounds and the multiple holds/expands):

• Homebuilding gross margin troughs near 19–20% and recovers toward 22–24% as the incentive war eases, lifting normalized EPS toward $460–525. Falsification test: HB gross margin stalling sub-20% for two-plus more quarters.
• The DC-metro/federal-jobs shock proves transient, and Mid-Atlantic orders/profit normalize. Falsification test: Mid-Atlantic segment profit remaining down >30% YoY through 2026.
• NVR keeps sourcing finished lots on acceptable terms (pipeline holds/grows) without paying up enough to erode the margin advantage. Falsification test: lot-cost-driven margin compression continuing even after volumes stabilize.

For the bear case (NVR converges toward a rate-levered commodity builder):

• 19–20% gross margin is a structural reset (permanent buydowns + lot-cost inflation), so ~17x sits on still-too-high earnings. Falsification test: two-to-three quarters of margin re-expansion through 21–22%.
• The DC concentration becomes a multi-year drag the diversified peers escape. Falsification test: Mid-Atlantic orders re-accelerating to flat/positive YoY.
• The 5x-book premium compresses as the land-light model commoditizes and ROE drifts toward the peer pack. Falsification test: ROE holding ≥30% through the cycle while peers stay sub-15%.

The shared falsification clock is the next two-to-three quarters of homebuilding gross margin and Mid-Atlantic order data: margin re-expanding through 21–22% with orders holding the +7% inflection is the bull’s vindication; margin sub-20% with deepening DC softness is the bear’s.

15. Source Appendix

(See the full source list in Appendix B below.) Primary sources: NVR FY2025 10-K (filed 2026-02-11), Q1-2026 10-Q (filed 2026-05-06), DEF 14A proxy (filed 2026-03-17), the 2023–2026 Form-4 insider corpus, and recent 8-Ks — all from SEC EDGAR (CIK 0000906163). NVR holds no earnings calls and issues no guidance, so there are no transcripts. Quantitative data: third-party financial databases (statements/ratios/valuation — reconciled to the 10-K, with the return_com_eqy and pr_to_book labels corrected); a multi-year valuation-percentile dataset (own-history percentiles) and a long-run price history; a third-party factor model. Industry/peer context: NAHB, US Census new-home sales/starts, Fitch 2026 starts forecast, HousingWire/Newsweek (DC-metro data), and the public filings of Lennar (LEN) and D.R. Horton (DHI). Frameworks: Greenwald & Kahn, Competition Demystified; Marathon/Chancellor, Capital Returns.

APPENDIX A — Standard Diligence Questionnaire

NVR, Inc. (NYSE: NVR) · Prepared 2026-06-27 · USD. Supplemental to the main note; Fact / Interpretation / Assumption labels applied where material.

General

What thoughtful questions have other investors asked about this company? The dominant questions: is the asset-light land-option model a durable moat or a now-replicable structure (DHI/Forestar, LEN/Millrose)? Is 19–20% gross margin the cyclical trough or a structural reset? How deep is the DC-metro/federal-workforce concentration risk into 2025–26 RIFs? Is the price-insensitive, all-buyback, no-dividend capital policy optimal? And how do you value a stock that compounds via buyback + ASP rather than unit growth, at ~17x trough-ish earnings?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: low-ish, and still falling. Gross margin has compressed from a 27.3% peak (2022) to 19.6% (Q1-2026); diluted EPS rolled from $506.76 (2024) to a TTM ~$407. TTM EPS is not trough-normalized — it embeds a still-rolling margin. Orders, however, inflected +7% in Q1-2026 (the leading indicator), so volume may be basing while margin troughs.

Driven by the external environment or internal actions? Overwhelmingly external — mortgage rates (~6.5–7%), affordability, and the federal-workforce contraction in NVR’s largest market. Internally, NVR is doing the disciplined things (leaning into cheap lot options, +8% communities, buying back stock).

How stable are revenues? Fact: not stable — homebuilding is deeply cyclical and revenue is volume × price, both rate-gated. There is zero recurring revenue. Settlements have oscillated 19,800–22,800 for six years with no secular trend.

Outlook for products/services. Demand is structurally supported (chronic US housing deficit 1.5M–5.5M units) but near-term gated by affordability. International: none — NVR is US-only, East/Mid-Atlantic-weighted.

How big will this market be? New-home construction takes long-run share from a chronically undersupplied existing-home market, but the near-term market is shrinking (Fitch: 2026 single-family starts −4.5%).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Structurally consolidating toward scaled publics (favoring survivors), but it remains a commodity with no demand captivity. The land-light edge NVR pioneered is now being copied, so the structural differentiation is narrowing even as NVR’s execution edge persists.

How profitable is the business (ROIC, ROE)? Fact: ROE ~33% (FY25), ~40% (FY24), ~49% peak (FY22) — 2–3x land-owning peers, on a net-cash balance sheet. (Ignore third-party financial databases’s 8.5% — a data artifact.) The cause is asset turnover on a tiny capital base, not leverage or operating scale.

How profitable is the industry — competitors, barriers? Bifurcated by quality: NVR ~33% ROE, PHM high-teens, DHI 12–15%, LEN 8.6%. Barriers are low for the commodity end (anyone can build a house) but real for scaled, capital-efficient, well-financed builders; NVR’s specific barrier is its 30-year option-model execution and Mid-Atlantic density.

Can the business be easily understood? Yes, with two caveats — the off-balance-sheet land-option mechanics, and the data-feed ROE/P-B artifacts.

Can it be undermined by foreign low-cost labor? No — homebuilding is local, physical, and non-tradeable.

Do brands matter? Marginally — “Ryan Homes” is a price-tier label, not a demand moat. Homebuyers have no brand loyalty or switching cost.

Nature of competition? Price, location, lot supply, and speed-to-build. NVR competes on capital efficiency and disciplined returns, not on out-building peers on volume.

Customers’ switching costs? Zero — a home is a one-off, infrequent purchase.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The controlled-lot option pipeline (~180,100 lots for ~$925M of deposits) is largely off-balance-sheet optionality — an asset (cheap control of an 8-year lot pipeline) the balance sheet understates. The cheap 3.00% senior notes (below par) are a hidden liability benefit.

Off-balance-sheet liabilities? The $733.9M of future LPA payment obligations are contingent and walk-away-able (liquidated-damages cap = forfeited deposit); not a fixed liability. NVRM warehouse facilities are non-recourse-style and undrawn.

How conservative is the accounting? Very — no goodwill, no acquired-intangible amortization, quarterly expensing of deposit impairments, KPMG unqualified.

How CapEx-hungry? Almost not at all — capex ~$24.5M (~0.2% of revenue). This is the entire point of the model.

Capital Allocation & Management

How much FCF, and how is it used? ~$1.0–1.1bn normalized FCF (FY25), returned 100% via buyback — no dividend, ever. Cumulative treasury $15.7bn; shares −24% in five years.

Significant acquisitions? Essentially none — NVR grows organically and has avoided the value-destructive land/builder M&A that litters peer histories.

Buying back shares? Yes, relentlessly — but price-insensitively (most dollars at the highest prices: ~$8,011 avg in FY24 near the ATH). Accretive over the cycle given 33% ROE, but momentum-paced.

Issuing large amounts of stock to insiders? No — SBC ~$69M (modest); the ~5.5% basic-vs-diluted wedge reflects legacy in-the-money options, but net dilution is contained by the buyback.

Compensation policy. Below-25th-percentile target cash; bonus capped at 100% of salary (80% pre-tax profit + 20% net new orders — no ROE/ROIC metric); LTI delivered via periodic plain stock-option blocks plus 4–8x ownership requirements. Aligned (options worthless unless the stock rises) but an absolute-price beta machine rather than a relative/alpha test.

Motivations of management. Long-tenured, disciplined, ROE-focused, allergic to guidance and earnings calls. Founder Dwight Schar has fully exited (no remaining family-control overhang); Saville (Exec Chairman) is the largest insider at 6.3%.

Valuation & Market Data

ADR, MLP, or K-1 issuer? None — NVR is a US C-corp common stock. No K-1, no ADR. It has never split, so the share price (~$6,829) and per-share figures are unusually large.

Dividend policy? No dividend — 100% of return via buyback.

How profitable? The most profitable homebuilder by return on capital (~33% ROE), the least by absolute unit volume among the majors.

Is net income diverging from cash from operations? Reported OCF is noisy (inventory and deposit timing swings) and should not be read literally year-to-year; normalized FCF ~$1.0–1.1bn ≈ net income. No adverse divergence.

Risks & Downside

What would cause the stock to decline? Sustained 7%+ mortgages; gross margin stalling sub-20%; a deeper DC/Mid-Atlantic federal-jobs shock; a structural-margin-reset re-rate of the ~5x book premium; lot-cost inflation eroding the return advantage.

Risk of catastrophic loss? Very low — net cash ~$1bn, owns only ~2,300 lots, max downturn loss is forfeited deposits, GFC-proven survivor. The realistic downside is a valuation/earnings re-rate, not solvency.

Chance of a total loss? Negligible on any reasonable horizon.

Recent News & Events

Has the business environment changed recently? Yes — margins compressing on the rate-buydown incentive war, and a new, NVR-specific federal-workforce shock in its largest (DC-metro) market.

Significant acquisitions? None by NVR. Sector-relevant: Berkshire’s ~$8.5bn agreement to acquire Taylor Morrison (May 2026), creating the #4 builder.

Change in accounting policies? None material.

Recent changes — markets, facilities, management? CEO transition (Saville → Bredow, 2022); founder Schar’s full exit; board additions (DeVito ex-Freddie Mac, Oliver ex-Johnson Controls, 2025); a small new drift into JV/land-development spend to secure lots. (The reported HQ relocation is unconfirmed — no 8-K on file; treat as a non-event.)

APPENDIX B — Source Appendix

NVR, Inc. (NYSE: NVR) · Compiled 2026-06-27. Primary sources first. SEC EDGAR CIK 0000906163. Quantitative figures reconciled to NVR’s filings; third-party financial databases and the cross-checked, with data-feed artifacts corrected (see notes).

Primary — SEC Filings (mirrored locally from EDGAR to output/NVR/sources/)

• NVR, Inc. FY2025 Form 10-K (filed 2026-02-11) — Business, segment data, MD&A (homebuilding gross margin, lot pipeline, contract-land-deposit footnotes), risk factors.
• NVR, Inc. Q1-2026 Form 10-Q (filed 2026-05-06) — orders/backlog/community-count, segment commentary, Q1 results.
• NVR, Inc. FY2023 Form 10-K (filed 2024-02-14) — multi-year reconciliation.
• NVR, Inc. DEF 14A proxy (filed 2026-03-17) — executive compensation, insider ownership (Vanguard 12.7%, BlackRock 11.2%, Saville 6.3%), board composition.
• NVR, Inc. Form 3/4/5 insider corpus (2023–2026) — the insider-transaction read (option exercise-and-sell pattern; DeVito’s two open-market buys).
• NVR, Inc. 8-K material-event filings (buyback authorizations, board changes).

Quantitative Data Sources

• third-party financial databases — income statement, balance sheet, cash flow, valuation multiples, enterprise value (multi-year). Reconciled to the 10-K; two labels corrected: (i) return_com_eqy (8.5%) is a treasury-accounting artifact — true ROE ~33% = net income ÷ average equity; (ii) pr_to_book_ratio (1.29x) is also artifactual — true P/B ~5.0–5.5x = pr_to_tang_bv_per_sh.
• a multi-year valuation-percentile dataset — own-history percentile ranks (composite 46.97th, P/E 49.8th, P/B 33.2nd, P/S 57.9th = mid-range own-10yr-history) and five-year (since-1986) daily price history. NVR has never split.
• a third-party factor model factor model — Industry: Home-Construction beta 1.37 (R² 0.73); risk-adjusted track record (lifetime +13.5%/yr, y10 +14.9%/yr; y1 −6.2%, m6 −14.5%, m3 +25.6%; rs_peak −31%); related-stock cohort (TOL/PHM/DHI/LEN + ITB/NAIL ETFs).

Industry & Peer Context

• NAHB (builder sentiment, mortgage-rate outlook, incentive prevalence, housing-deficit estimate); US Census Bureau (new-home sales SAAR, months’ supply, median price, starts); Fitch Ratings (2026 single-family starts forecast −4.5%); Freddie Mac / Zillow / NAR (housing-deficit estimates).
• DC-metro / federal-workforce data — HousingWire, Newsweek, Bright MLS (DC-area inventory +50% vs +30% national; ~22,100 federal jobs lost DC/MD/VA; agent-survey data on federal-layoff-driven transactions).
• Sector events — Berkshire Hathaway–Taylor Morrison ~$8.5bn agreement (May 2026); the 21st Century ROAD to Housing Act (Senate 85-5, House 358-32, June 2026); KB Home Q2 results.
• Analyst consensus — public.com, stockanalysis.com, Investing.com (BofA $8,600 target); consensus Hold, average targets ~$7,000–8,200.
• public filings and reporting of Lennar (LEN, 2026-06-27) and D.R. Horton (DHI, 2026-06-19) — peer economics, the shared housing cycle, and land-light comparisons (Forestar / Millrose).

Analytical Frameworks

• Bruce Greenwald & Judd Kahn, Competition Demystified — barriers to entry, the three genuine advantage types, market-share-stability and ROIC tests.
• Edward Chancellor (ed.) / Marathon Asset Management, Capital Returns — supply-side capital-cycle analysis and the high-returns-attract-capital mean-reversion lens.

All multiples and per-share figures recomputed at the 2026-06-26 close of $6,829 and ~2.71M shares outstanding (after the Q1-2026 buyback); ROIC’s stale internal price ($7,292) was not used. Management’s (sparse) public commentary is treated as hypothesis and validated against the filings.