NVR, Inc. NVR

Key data

The business

NVR builds homes across the eastern US under Ryan Homes, NVHomes and Heartland, and runs a mortgage arm that funds its own buyers. What makes it odd among builders is what it refuses to do: it does not own land. It controls lots through options, putting down a cash deposit for the right to buy finished lots on a schedule as it needs them. When demand holds, land becomes a small line item and the cash goes to buybacks. When demand breaks, the most it loses on a bad parcel is the deposit, while builders who own the dirt take writedowns that can erase a year of profit. That single choice is the moat: a balance sheet that cannot be forced to dump homes into a falling market to service land debt. The thing a buyer actually signs is a Ryan Homes contract on a lot NVR does not yet own.

The numbers

Two forces pull against each other here.

Profit stepped down hard from late 2025 into 2026. The compounding math is the whole story: June-quarter net income fell 29% year over year, yet earnings per share fell only 23%, because NVR retired about 8% of its shares over the same span. That six-point gap, erased by buybacks, is the machine working as designed. The last two quarters both missed consensus, by about 7% in June.

Across the cycle the business is flatter than the stock's swings: revenue near $10B for four years, per-share earnings climbing on a shrinking share count, then giving some back in 2025. Consensus has this year down about 17% and next year recovering. Now the operating detail the lens turns on:

Orders are the leading print and they inflected up 9% in units with cancellations falling, the clearest sign demand is not collapsing. But the price line is the warning: the average new order came in below the price NVR is still settling old backlog at, so the margin already reported has further to fall before it stabilizes. The market prices NVR as a builder derating into a slowdown, but the per-share engine and the deposit-only downside make the trough shallower and safer than a levered peer's, and the print that settles it is next quarter's gross margin against the order price already booked.

Management

Insiders were net sellers over the past year, about $9.0M sold against $0.6M bought, the largest a $4.05M officer sale in February; plan status is not disclosed in the filings I have, so I cannot call those routine. Weigh that against capital return: over $1.8B in buybacks in 2025 and another $632M in the first quarter of 2026, done steadily for years rather than in bursts. Contract land deposit impairments rose to $30.6M in the first half from $21.3M, the land-light model taking its small, honest losses as some option lots go underwater. That is the system paying its own insurance premium.

How it fails or surprises you

Margin makes a new low (downside). The order price sits below the settlement price, so homes closing over the next two quarters carry thinner margins than the 19.2% just reported. If gross margin prints below 19% with orders flat, the trough is deeper than consensus and 16x looks dear on falling numbers. Watch Q3 gross margin.

The land-light model cracks (downside). Rising deposit impairments, $30.6M and climbing, tell you optioned lots are going underwater faster than NVR can assign them. A sharp jump means the downside insurance is costing more than usual. Only about $12.8M of the $50M in deposits is refundable.

Orders keep inflecting (right tail). Orders already turned up 9% with cancellations falling. If rates ease and that growth holds into spring while the share count shrinks another 8%, per-share earnings re-accelerate off a low base and the multiple proves cheap. First sign: another quarter of positive order units with cancellations still falling.

Closing thoughts

NVR is selling more homes at lower prices, and next quarter's gross margin decides whether volume recovery outruns price erosion or the trough deepens. Gross margin below 19% with flat orders means the downturn is deeper than priced and 16 times looks expensive on falling numbers. Margin holding while orders stay positive means the turn is underway and the multiple proves cheap off a low base. Orders turned positive at 9% growth even as margin kept sliding, so the fat tail is up and the left tail is a slow bleed rather than a wipeout, because the downside is bounded: deposits walked away from, not book value destroyed.

The bet is still that NVR builds the cheap way, options land instead of owning it, and buys back one share in twelve a year so each remaining share owns more. It breaks if margin makes a new low while orders roll back over, and the one pair to watch is next quarter's gross margin against the order price already booked. It turns if orders keep climbing before the price line stops falling.