KBH: Backlog Grew. Demand per Community Did Not.

The 60-second read

Five things to know

1. The EPS beat is not a sales rebound. $1.05 EPS exceeded FMP consensus, but revenue fell 19.96% and net income fell 40.56%.
2. Backlog rose while selling productivity fell. Backlog grew 1.50% YoY, but orders per community declined from 3.8 to 3.1.
3. Built to Order helped sequential margins. Nearly three-quarters of deliveries were BTO; reported housing margin recovered to 16.5% from 15.2%.
4. Tax and investment items supported profit. The quarter included a $3.5M gain and a 19.6% effective tax rate versus 23.3% last year.
5. The book discount carries inventory and funding risk. Inventory reached $5.981B and defined net debt rose $485.764M from fiscal year-end.

Three scenarios, one probability-weighted number

Analyst assumptions; scenario values are not company guidance.

Signposts: what would change our mind

Upgrade evidence requires better selling productivity and capital discipline. Thresholds are our review framework; period-end dates are not invented earnings announcement dates.

The price is not the reaction

KB Home enters this report with a visible discount to reported book value, but the discount is compensation for a difficult earnings cycle. Our reference price is $48.59, the FMP quote timestamped September 22, 2026 at 20:02:43 UTC. The release carries a later clock label and the SEC acceptance field conflicts with that label. We therefore leave the precise release time unconfirmed and do not describe this quote, its daily change, or the daily bar as the market's reaction to the new results. That distinction matters: a price observed around a release is not automatically evidence of how investors interpreted it.

At the reference price, the shares trade at 0.78 times the company's reported $62.56 book value per share. That sounds inexpensive until the composition of equity is examined. Land and homes under construction are the dominant assets; they cannot be spent like cash and can become less valuable when affordability deteriorates. Our central question is whether the business can earn an acceptable return on that capital without relying on progressively more incentives or borrowing. A below-book multiple alone cannot answer it.

The saved fifteen-month daily-price history provides context, not a causal model. For historical prints with both observations available, the reaction calculation compares the release-date closing price with the following trading-day close. This window can include unrelated macroeconomic news and rate moves. Older releases outside the saved price history remain unavailable. The current print has no subsequent completed daily observation in this dataset, so its reaction is deliberately blank. Neither blanks nor stale observations are filled with estimates.

The reference market capitalization from FMP is about $2.98 billion. Our valuation arithmetic instead consistently uses the release's approximately 60.8 million period-end shares, giving $2.95 billion at the same price. The small difference reflects share-count conventions and rounding, not additional cash or an alternative fair value. Current quote data, quarterly average diluted shares, and period-end common shares serve different purposes and should not be interchanged.

The beat and the business

Third-quarter revenue was $1,297.101 million, down 19.96% from $1,620.474 million. Net income fell 40.56% to $65.280 million. Diluted earnings of $1.05 exceeded FMP's $0.892 consensus by 17.71%, while revenue exceeded its $1,296.429 million estimate by only 0.05%. This was a meaningful earnings beat against a depressed expectation, not a revenue acceleration. FMP actuals match the primary release for this quarter; no such agreement is assumed for every historical feed row.

Deliveries declined to 2,732 homes from 3,393, and average selling price was $473,000 versus $475,700. The sales decline was therefore mainly a unit-volume problem. The business did improve sequentially: second-quarter revenue was $1,112.435 million and diluted EPS was $0.43. Management says the return to a predominantly Built to Order model is now achieved, with nearly three-quarters of deliveries in that format. That shift can support product mix and margins, but one quarter does not establish the sustainable margin through a full demand cycle.

The comparison headers are explicit: the income statement presents three months ended August 31, 2026 and 2025, followed by nine months ended August 31, 2026 and 2025. The current year is first in each pair. We do not mistake year-to-date weighted shares for quarterly weighted shares. The primary quarterly diluted count is 61.759 million, compared with 67.737 million a year earlier. Standardized vendor fields can differ when cumulative reporting has been mapped into a quarterly series.

The maintained full-year housing-revenue range of $4.90 billion to $5.10 billion is consistent with management's expectation that the year will finish within its June ranges. It is not a promise that next year's orders will recover. A delivery is recognition of an earlier sale and construction process; a net order is an input to future deliveries. Separating those two clocks is essential to interpreting a homebuilder's apparent rebound.

Backlog growth is not demand growth

Ending backlog increased to 4,398 homes from 4,333 a year earlier, the first year-over-year increase in four years according to management. The exact calculated unit growth is 1.50%, which the release rounds to 2%. Backlog value rose to $2,053.208 million from $1,988.863 million. Those are constructive comparisons, but the sequential roll-forward tells a more restrained story: 4,526 beginning homes plus 2,604 net orders less 2,732 deliveries equals 4,398 ending homes. Backlog declined 2.83% during the quarter.

Demand productivity is the sharper warning. Net orders fell 11.73% from 2,950 even though average community count grew 8% to 279. Monthly net orders per community fell from 3.8 to 3.1, an 18.42% decline using the reported rounded rates. A larger selling footprint can stabilize the aggregate book while each community sells more slowly. It can also tie up capital for longer. The cancellation rate increased to 18% of gross orders from 17%, so the backlog remains contingent on buyers reaching closing.

The regional pattern is uneven. West Coast deliveries were nearly unchanged at 966 versus 972, while Southwest, Central and Southeast deliveries declined. West Coast net orders increased to 937 from 870, but Central orders fell to 587 from 795 and Southeast to 632 from 826. A national headline therefore conceals both a comparatively resilient region and substantial weakness elsewhere. California exposure also means the most supportive current region is already an important concentration risk.

The release supplies regional deliveries, average selling prices, orders, order values and backlog, but not regional profit. We show the reported measures and do not infer profit from unit mix. Multiplying rounded average selling prices by deliveries can approximate regional housing revenue, but that is not an exact segment statement and is not presented here as one. The missing profit split limits how confidently we can identify the source of sequential margin improvement beyond management's own mix explanation.

A better quarter with helpful below-the-line items

Housing gross profit was $213.760 million, or a reported 16.5% margin. Adding back $2.986 million of inventory-related charges produces adjusted housing gross profit of $216.746 million and a 16.8% margin. The prior-year comparable adjusted margin was 18.9%. Inventory charges were smaller than the prior year's $11.338 million, yet the adjusted margin still declined. The weakness is therefore not explained away by a larger impairment charge in the current quarter.

The sequential picture is better: reported gross margin increased from 15.2% to 16.5%, while adjusted margin rose from 15.7% to 16.8%. However, selling, general and administrative expense consumed 11.3% of housing revenue, versus 10.0% a year earlier. Lower incentive compensation and personnel costs partly offset weaker operating leverage. Homebuilding operating income of $67.134 million was roughly half the prior-year $131.163 million. Better product mix can coexist with a smaller profit pool.

The bridge from operations to EPS includes details worth retaining. Homebuilding interest income and other contributed $4.518 million, including a $3.5 million gain on selling an equity investment. Homebuilding joint ventures added $2.147 million and financial services pretax income added $7.381 million, bringing total pretax income to $81.180 million. Income tax was $15.900 million, a 19.6% effective rate compared with 23.3% last year. Management attributes the lower rate mainly to excess tax benefits from stock compensation.

An illustrative sensitivity removes the $3.5 million investment gain and taxes the remainder at the prior-year 23.3% rate: ($81.180 million minus $3.5 million) times 76.7%, divided by 61.759 million diluted shares, equals approximately $0.96 per share. That is our calculation, not company adjusted EPS or a restatement. It shows why simply annualizing $1.05 would overstate the durability of this quarter's combination of tax, investment and operating results. Buybacks also cushioned EPS: shares fell 8.83% while net income fell more sharply than per-share earnings.

Earnings quality: separate delivery from cash

The nine checks below are a consistent diagnostic, not a mechanical credit rating. Homebuilders recognize revenue as homes close while land acquisition and construction cash can occur much earlier. This makes a single-quarter cash-conversion ratio noisy, but it does not make cash irrelevant. The current earnings exhibit does not contain a cash-flow statement. The latest filed detail is the six months ended May 31, 2026, when operating cash flow was negative $93.470 million, compared with negative $165.883 million a year earlier. The improvement still left a cash outflow.

For that same six-month period, property and equipment purchases, net, were $22.890 million. A simple operating-cash-flow-minus-capex measure is therefore negative $116.360 million. Land investment is already inside operating cash flow; subtracting the disclosed land spending again would double-count it. Joint-venture contributions and distributions complicate comparisons further, so we preserve the company's statement classifications and explain the narrower FCF convention. We do not manufacture a third-quarter cash result from balance-sheet movements.

Receivable and inventory days are shown as simple closing-balance diagnostics. They use ninety days, the current period-end balances and the corresponding quarterly denominators. Receivables include items beyond ordinary customer billing, and inventory includes land that may be held longer than a single selling cycle. Accordingly, these are not directly comparable with software receivable days or retailer inventory turns. Their purpose is to show the amount of capital tied to current throughput, not to allege aggressive accounting from a high absolute number.

The annual record illustrates the cycle. Operating cash flow was $1,082.699 million in fiscal 2023, $362.722 million in 2024 and $335.682 million in 2025. FCF on the stated convention was $1,047.231 million, $323.411 million and $287.282 million. The strong earlier release of cash is not a sensible perpetual run rate when the business is buying land again. The current tax benefit and investment gain receive warning flags because they may not repeat; inventory-charge exclusions receive a warning because those costs are economically relevant in this industry.

Book value needs an inventory haircut

At August 31, homebuilding cash and cash equivalents were $159.018 million and notes payable were $2,109.145 million. Subtracting cash from notes produces net debt of $1,950.127 million on that defined scope. At the fiscal year-end those same figures were $228.614 million and $1,692.977 million, giving net debt of $1,464.363 million. The increase of $485.764 million is material even though management reports ample total liquidity. Our net-debt measure excludes other liabilities and is not a complete liquidation obligation.

Reported liquidity of $942.4 million comprises $159.0 million of cash and $783.4 million of available revolver capacity. It does not mean $942.4 million is sitting in the bank. The company already had $415.0 million of revolver cash borrowings outstanding. Undrawn capacity provides flexibility, but using it creates debt and remains subject to the credit agreement. The debt-to-capital ratio increased to 35.7% from 30.3% at November 30, 2025, compared with 33.2% a year earlier.

Inventory increased to $5,981.182 million from $5,670.802 million, a 5.47% rise while nine-month revenue fell. Inventory now represents 85.82% of total assets. Quarterly land and land-development investment rose to $722.3 million from $514.1 million, while nine-month spending fell to $1.79 billion from $1.95 billion. Both comparisons can be true: the business accelerated investment in the latest quarter against a lower cumulative spending base. Lots owned or controlled totaled 61,581, approximately 60% owned and 40% under contract.

Repurchases of $50 million in the quarter and $175 million year to date support per-share ownership, but they are a use of liquidity. The remaining $725 million authorization is permission, not an obligation or a forecast. A simple stress test writes down inventory by 10% with no tax benefit or offsetting recovery. That subtracts $598.118 million, or about $9.84 per period-end share, from reported book value and leaves roughly $52.72 per share. It is a sensitivity, not a prediction of imminent impairments; it explains how a quoted discount can narrow when the assets themselves change.

Three routes, one cautious call

Our rating is HOLD with conviction three out of five and High uncertainty. Base fair value is $50 per share, with a $32 to $70 scenario range. This is a model of potential normalized earning power and capital value, not management guidance or a short-term price forecast. At $48.59, upside to the base is only 2.90%. The probability-weighted value is $48.60, almost identical to the reference price. The model therefore offers insufficient compensation to take a new position solely because the stock is below book.

The first route values normalized EPS of $5 at ten times earnings, yielding $50. Five dollars is an explicit analyst assumption above the current depressed earnings pace and above FMP's fiscal 2027 EPS estimate of about $4.42. It requires recovery in margins and delivery efficiency; it is not a claim that the next year is already secured. The second route stresses the assets by subtracting 10% of inventory from equity without recognizing a tax benefit. On the company's approximate period-end share count, that leaves $52.72 per share before any further discount for time, execution or other liabilities.

The third route capitalizes a sustainable return on book equity: $62.56 book value times a 10% normalized return on equity divided by a 12.5% required equity return equals $50.05. This deliberately assigns no perpetual growth benefit. It is a return-on-capital cross-check, and shares an equity base with the asset route; the routes are not independent statistical confirmations. Their central readings cluster around $50 to $53, and we round the base to $50 rather than pretending the inputs justify cent-level precision.

The reverse calculation is revealing. At $48.59 and a ten-times earnings multiple, investors pay for about $4.86 of sustainable EPS. On the no-growth book-return route, price/book of 0.777 times a 12.5% required return implies roughly 9.71% sustainable ROE. Those are attainable recovery cases, but stronger than today's weak earnings alone prove. Our High uncertainty rating calls for an illustrative 20% discount to base value, or $40, before considering a price-only upgrade. Stronger operating evidence could also change the model; a lower price is never sufficient if the thesis deteriorates.

Wall Street is mostly cautious too

FMP's saved rating distribution contains one strong buy, nine buys, twenty-eight holds and six sells, with an aggregate Hold label. Its price-target consensus is $58.60, median $57, low $50 and high $66. These vendor aggregates do not all necessarily use the same analyst population or update window. They should be read as descriptive context, not as an independently verified probability distribution. At our reference price, the consensus target implies 20.60% upside, considerably more than our $50 base case.

The most recent identifiable target observations in the saved data precede this earnings print. Truist Financial is recorded at $55 on September 16, and Wells Fargo at $50 on September 11. The corresponding ratings history shows Hold for Truist Securities and Underweight for Wells Fargo. Barclays is recorded at a $57 target on June 25, while its rating entry remains Overweight. These dates matter more than the apparent precision of an average: none is presented as a reaction to the quarter announced on September 22.

The annual estimate snapshot has fiscal 2026 revenue of about $5.103 billion and EPS of $3.31, with eleven EPS analysts; fiscal 2027 has about $5.487 billion revenue and $4.42 EPS, also with eleven EPS analysts. Fiscal 2028 has about $5.889 billion revenue and $5.67 EPS but only three EPS analysts. Coverage thins as the forecast horizon extends. Those figures are a saved snapshot that may still be awaiting post-print revisions. Company housing-revenue guidance excludes the small financial-services revenue contribution, so a direct comparison to consolidated consensus is not perfectly like-for-like.

Our disagreement with the target average is primarily about compensation for uncertainty. We acknowledge room for normalized earnings to recover and explicitly model $5 EPS. We also require a restrained cyclical multiple and recognition that the asset base is leveraged inventory. A $58.60 target can be reached if earnings, the multiple or both exceed our base assumptions. The analyst rows are attributed to FMP as the provider; we do not pretend to have read private broker notes or infer their unpublished models.

Grade the process, not the promise

Management deserves credit for completing the stated transition toward predominantly Built to Order deliveries and for improving gross margin sequentially. That does not erase the company's own warning that housing conditions weakened since the June report. We interpret the unchanged full-year ranges as a near-term execution commitment and the demand language as a separate caution about the order funnel. Reading only one of those statements would distort the balance of the release.

Capital allocation is the harder judgment. Repurchasing below stated book value can increase per-share ownership economically if the assets earn adequate returns. It can also reduce liquidity before an extended downturn. This quarter combined repurchases with a higher revolver balance and renewed land investment. The relevant test is whether the company can fund its communities and distribution policy while stabilizing leverage, not whether an authorization remains large. Board permission to buy stock cannot be treated as a floor under the trading price.

The governance structure is comparatively straightforward: the annual report describes one class of outstanding voting securities with one vote per share, and common stock is listed on the NYSE. Ernst & Young has served as auditor since 1991. The historical 2024 registration-termination and exchange-removal forms concerned rights to purchase preferred shares that expired April 30, 2024; the Form 15 expressly retained common-stock reporting. They do not establish an unresolved delisting of the current common shares. Domestic joint ventures and land-option VIE assessments still matter, and should not be confused with a foreign contractual operating structure.

Questions for the next update follow directly from the financial evidence. How much of margin improvement came from Built to Order mix versus incentives, land mix and costs? Can order absorption recover with a larger community footprint? What level of land spending is needed to preserve future delivery capacity? How much cash can be released without impairing that capacity? And what conditions would cause repurchases to slow? These questions seek measurable tradeoffs; a general statement about long-term shareholder value does not resolve them.

What can break the thesis

Affordability is the first risk because it affects both sales pace and the price or financing incentives needed to produce those sales. Higher rates can push a buyer out of qualification even when the household still wants a new home. Lower absorption can then lengthen the time land and construction capital remain employed. Our base case assumes eventual recovery in normalized returns; if affordability does not improve, both the assumed EPS and the asset multiple may be too high.

Inventory valuation is the second risk. The balance sheet carries land and homes rather than a liquid securities portfolio. Lower attainable selling prices, cost overruns or weaker local demand can produce impairments or contract abandonments. The illustrative ten-percent inventory stress is not a forecast or a maximum loss. It simply makes the exposure visible in per-share terms. Actual effects would depend on the location and stage of each asset, tax treatment, cancellation terms and the company's ability to adjust construction pace.

The third risk is that capital allocation outruns cash generation. Revolver capacity offers flexibility, but increased borrowing and repurchases during weak operating cash flow can reduce options later. We do not assert that KB Home is in immediate financial distress; the release reports substantial liquidity. The concern is a narrowing margin of safety if the earnings recovery takes longer than expected. A fourth risk is that the favorable West Coast order pattern reverses, given the company's California concentration.

The strongest bull case is that Built to Order execution protects pricing, community openings mature, and lower rates restore absorption while repurchases amplify EPS. The strongest bear case is that backlog growth masks slower throughput, incentives remain necessary and inventory absorbs more cash just as returns decline. Our HOLD lands between these positions. It recognizes tangible capital and a credible recovery mechanism, but the reference price already sits close to our probability-weighted value. High uncertainty is a reason to demand evidence or a larger discount, not a reason to ignore either side.

The next checks are measurable

The immediate information event is the earnings call scheduled by the release for September 22 at 5 p.m. Eastern. This packet is grounded in the saved release and filed reports; no call transcript or management answer is assumed to have been reviewed. The future Q3 Form 10-Q should supply the missing operating-cash-flow, stock-compensation and regional-profit detail. Its filing date is not invented here. Those data could change the quality assessment even if the headline EPS remains exactly as reported.

The next operating checkpoint is fiscal Q4, ending November 30, 2026. Management guides to 3,000 to 3,500 deliveries, $1.45 billion to $1.65 billion of housing revenue, housing gross margin of 16.0% to 16.6% assuming no inventory-related charges, and SG&A of 10.3% to 10.9% of revenue. The tax guide is about 26%, and ending community count is 270 to 275. We will grade those actual results when published; November 30 is the period-end date, not an asserted announcement date.

A midpoint illustration multiplies $1.55 billion revenue by a 16.3% housing gross margin less a 10.6% SG&A ratio. That produces $88.35 million of homebuilding operating income before financial-services and other items. Applying a 26% tax rate yields $65.379 million, or about $1.06 per share on the current quarter's 61.759 million diluted shares. This is not an EPS forecast: it omits below-the-line contributions, future share-count changes and the difference between housing and other revenue. It illustrates why the lower tax benefit next quarter matters.

Our signposts focus on absorption, gross margin, leverage and cash. A price move toward $40 could create a margin of safety only if the fundamental case remains intact. A recovery in orders per community above 3.5, maintained gross margin and stable financing would instead support revisiting the operating assumptions. Conversely, a stronger backlog headline without better sales productivity would not by itself earn an upgrade. The next packet should explicitly grade these thresholds rather than quietly replacing them.

Source ledger and methods

This packet uses the September 22, 2026 earnings release furnished as Exhibit 99.1 to the current 8-K, the previous eight quarterly earnings releases, the fiscal 2025 Form 10-K, the latest filed fiscal 2026 second-quarter Form 10-Q and saved FMP datasets. The nine-quarter primary rows retain the full income-statement and supplemental tables, including below-the-line contributions, share counts, balance sheets and housing-margin reconciliations. When a regional-profit or current-quarter cash-flow item is absent from the release, the dataset records that absence instead of filling it with a synthetic estimate.

Dollar figures in financial-statement tables are converted from the source's thousands into millions. Per-share figures remain ordinary dollars. Counts of homes and communities remain counts. Percentage-point changes are distinguished from percent changes. Derived values are computed in the accompanying data preparation file and marked computed in the dataset or chart note. Our ninety-day working-capital diagnostics use closing balances and are explicitly not average-balance turnover measures. Net debt here means homebuilding notes payable less homebuilding cash, excluding lease obligations and other liabilities.

The vendor dataset is supporting evidence, not a substitute for the filing. Its current actuals reconcile to the earnings release, but an older annual-quarter earnings row has an inconsistent date and EPS value. We use the verified December 18, 2025 release and $1.55 diluted EPS for fiscal Q4 2025. The primary source also controls quarterly weighted shares when a standardized feed appears to carry year-to-date counts. These reconciliations prevent a convenient database join from silently changing the economic period being analyzed.

Important terms: absorption means monthly net orders per community; backlog is contracted homes not yet delivered and can be cancelled; Built to Order refers to homes configured for a customer rather than primarily speculative inventory; book value is accounting equity per share and not liquidation proceeds; normalized EPS is our assumed earning power through a less depressed environment; the probability-weighted value is the sum of each scenario price times its assigned probability. Probabilities and valuation assumptions are editorial judgments, not company facts. There is no prior Charged Alpha KBH episode identified in the publication registry, so there is no prior packet signpost record to grade. This is the initial dated framework.

• Q3 FY2026 — SEC earnings release
• Q2 FY2026 — SEC earnings release
• Q1 FY2026 — SEC earnings release
• Q4 FY2025 — SEC earnings release
• Q3 FY2025 — SEC earnings release
• Q2 FY2025 — SEC earnings release
• Q1 FY2025 — SEC earnings release
• Q4 FY2024 — SEC earnings release
• Q3 FY2024 — SEC earnings release
• FY2025 Form 10-K
• Q2 FY2026 Form 10-Q
• FMP quote endpoint
• FMP target aggregate
• FMP annual estimates