NVR — Q3 2026 Financial Report Analysis

Orders turned down again, and margin kept slipping

NVR's third quarter of 2026 (three months ended September 30) undid most of the demand recovery it showed in the spring. Net new orders fell 7% to 4,391 homes, after rising 9% in Q2, and the share of buyers who cancelled rose to 21.0% from 19.4%. The company settled (closed and handed over) 7% fewer homes at a 2% lower average price, so homebuilding revenue fell 8% to $2.35 billion. It also kept less of each sale: homebuilding gross margin fell to 18.3% from 21.0%. Net income fell 27% to $248.6 million. Diluted earnings per share (EPS) fell a smaller 20%, to $89.45, because NVR has about 9% fewer shares outstanding after a year of heavy buybacks.

This analysis is based on NVR's earnings release (8-K Exhibit 99.1, filed October 22, 2026). The full 10-Q, with management's region-by-region explanations and the cash flow statement, isn't filed yet.

Key figures

Consolidated revenue is homebuilding revenue plus mortgage banking fees ($53.1M vs. $49.2M). NVR doesn't report an operating-income line, so the pre-tax margin is shown as the operating-margin measure: income before taxes ($334.7M vs. $444.1M) divided by consolidated revenue. The prior-year backlog value is units times the reported average backlog price ($479,500); the release states the dollar change as 4%.

Why revenue fell: fewer closings from a backlog that converted slowly

For a homebuilder, this quarter's revenue mostly comes from homes sold in earlier quarters and delivered now. NVR started the quarter with 10,998 homes in backlog (sold but not yet settled), 9% more than a year earlier. It still settled 7% fewer homes. It closed about 48% of its starting backlog in the quarter (5,268 of 10,998), compared with about 56% a year ago (5,639 of 10,069). The release doesn't explain the slower pace. In Q2, the 10-Q called the same problem "a lower backlog turnover rate," and that hasn't reversed.

Settlement prices held up better than order prices (-2% vs. -5%). Homes settling now were mostly sold earlier in the year, before the latest price cuts showed up.

Margin: land costs, discounts and walk-away write-downs

Homebuilding gross margin (the share of home-sale revenue left after paying for the lot and building the house) fell 2.8 points, to 18.3%. Gross profit fell 20%, from $538.9 million to $429.2 million. The release names three causes:

• Higher lot costs. NVR buys finished lots from developers under fixed-price contracts, often signed years in advance. Lots priced during the land-cost run-up are now reaching the income statement while home prices are falling.
• Pricing pressure. New-order prices fell 4.9%, and management attributes the pressure to affordability and weak consumer sentiment.
• Contract land deposit impairments of about $24.8 million. NVR controls lots by putting down deposits instead of buying land outright. When it expects to walk away from a contract or renegotiate it, it writes the deposit down. The charge was $21.7 million in Q2, so it rose again. Even with this quarter's charge added back, gross margin would have been about 19.3%, still well below last year's 21.0%. (The release doesn't give the prior-year charge, so this isn't a like-for-like adjustment.)

Over the first nine months, gross margin was 19.0%, down from 21.5%. Selling, general and administrative costs fell 5% to $135.0 million, which offset only a small part of the gross-profit decline.

Where demand weakened

The North East drove most of the decline: orders nearly halved, and its average community count fell to 26 from 34. Its higher order price likely reflects which communities were still selling, not price increases (a mix effect). The release doesn't break this out. In the Mid Atlantic, NVR's largest region, communities rose to 147 from 125, but orders still fell, so sales per community there dropped about 19%.

Mortgage banking: the one segment that grew

NVR also lends to its own buyers. Loan closings fell 10% to $1.38 billion, in line with fewer settlements, and 85% of buyers who took a mortgage used NVR's lender (86% a year ago). Even so, mortgage banking fees rose 8% to $53.1 million and segment pre-tax income rose 7% to $35.0 million. The release doesn't explain the higher fees on lower volume. The segment contributed about 10% of pre-tax income.

What the headline numbers hide

• Buybacks are carrying EPS, and they're draining cash. NVR repurchased 37,088 shares for $237.2 million in Q3 and 181,984 shares for $1.23 billion year to date, compared with $683.4 million of net income. Homebuilding cash fell from $1.88 billion at year-end to $1.09 billion, and shareholders' equity fell from $3.86 billion to $3.43 billion. The share count at period end was 7.6% lower than a year ago. That explains the gap between the -27% net income and -20% EPS. With less cash, the buyback pace may be harder to sustain unless operating cash flow picks up.
• A higher tax rate added to the decline. Pre-tax income fell 24.6%, but net income fell 27.5%. The effective tax rate rose to 25.7% from 22.8%. The release doesn't say why; the 10-Q should.
• Inventory grew, and cash conversion can't be checked yet. Homebuilding inventory rose to $2.37 billion from $1.72 billion at year-end (+38%). Most of the increase was in lots and homes already under contract ($1.94B vs. $1.41B), which fits a larger backlog. Unsold lots and homes also rose, to $384.8 million from $252.0 million. The release has no cash flow statement, so how much cash the business generated is a question for the 10-Q.
• Lower interest income. Homebuilding "other income" fell to $12.2 million from $22.0 million. The release gives no reason, but the drop is consistent with NVR holding less cash.
• Write-downs are a recurring cost now. Deposit impairments have been about $22-25 million a quarter for two quarters, roughly 1 point of margin each time. They look less like one-offs and more like a running cost of the current market.

Did last time's read hold up?

Our Q2 2026 analysis expected more settlements in the second half "if backlog turnover recovers." It didn't: settlements fell 7% and turnover slowed further. We also said gross margin was unlikely to return to the 21%+ range and named deposit impairments as the figure to watch. Both held: margin fell to 18.3%, and impairments rose to $24.8 million from $21.7 million. The part we didn't expect was the demand reversal. Q2's order growth (+9%) and lower cancellations both reversed in Q3.

Takeaway: NVR's larger backlog (10,121 homes, +10%) is worth less per home than a year ago (average $453,000 vs. $479,500, -5.5%), and new orders are now falling again. Revenue is likely to stay under pressure, and per-home profit is likely to keep shrinking for at least the next two quarters. Buybacks keep EPS from falling as fast as profit, but they are paid from a shrinking cash balance.

Outlook

NVR doesn't give numerical guidance, and this release has no management commentary beyond the causes of the margin decline. The trend data points to:

• Volume: The September 30 backlog of 10,121 homes is 10% larger than a year ago, so Q4 settlements could finally rise year over year, but only if NVR turns its backlog into closings faster than it did in Q2 and Q3. A 21% cancellation rate means part of that backlog won't close.
• Price and margin: Backlog homes average 5.5% less than a year ago, and new orders were priced 4.9% lower. Revenue per home and gross margin should stay under pressure into 2027, with lot costs rising at the same time. We'd expect gross margin to stay in the 18-19% range rather than recover.
• Land: NVR controlled 185,300 lots at September 30, up 5.7% from a year earlier, mostly through deposits rather than owned land. That limits its downside if the market weakens further: it can walk away and write off a deposit instead of holding land. The rising impairments show NVR is already doing some of that.

What to watch in the 10-Q and Q4: the reason for the slower backlog turnover, whether North East orders stabilize, operating cash flow against the buyback pace, and whether deposit impairments keep rising.