Initiated 2026-09-27 · Price $89.10 (as of 2026-09-25 close, stockanalysis.com) · Mkt cap $7.35bn · Industrials / building products (commercial HVAC and data-center cooling) · Model: verified

Rating: Sell — Conviction: Low

AAON's sales doubled in Q2 on data-center cooling, but margins fell as they did. Free cash flow has been negative for two years because of plant building and project working capital. At ~26x trailing EBITDA the price already assumes the AI build-out keeps compounding and margins return to their 2023–24 peak: the model's bull case sits only 21% above the price, and the base case is well below it. Conviction is Low because the $2.0bn backlog is real, and the data-center cycle could run longer than the base case assumes.

Model value range vs price
Bear $11.06Base $50.11Bull $108.15Price $89.10

Business overview

AAON makes semi-custom commercial HVAC equipment. It has two brands. AAON-branded rooftop units and Alpha Class heat pumps sell through independent reps to commercial buildings ($282.2mm in Q2 2026, +39.3%). BASX-branded custom cooling (air handlers, chillers, coils and liquid-cooling CDUs) goes mostly to data centers ($345.0mm, +216.2%, or 55% of sales). Q2 revenue was $627.0mm (+101.2%). FY2026 guidance is sales growth of 55–60%, gross margin of 25–26% (cut from 27–28%) and SG&A of 13–14% of sales. Earnings turn on three things: hyperscale and AI data-center cooling orders, utilization of the new Memphis, Longview and Redmond capacity (Memphis overhead was $18.1mm in Q2 against $3.0mm a year earlier), and price/cost timing on a fixed-price backlog.

Competition

In data-center cooling, the players that matter are Vertiv, Schneider Electric and Trane, plus a long tail of specialists. Vertiv led liquid cooling with about 11.3% share in 2025, and the top five held only ~35% (GMI Insights, 2025 data). The market is fragmented and new capacity is still arriving. If pressure shows up, it will show first in price and in BASX's bookings. BASX bookings already fell in Q2 from "unusually elevated" levels, although book-to-bill over the prior four quarters was close to 3. In rooftop units AAON gained share in what management called a "soft commercial HVAC market" against Trane, Carrier and Lennox. On one definition (stockanalysis TTM EV/EBITDA, 2026-09-25), Lennox trades at 12.7x, Carrier 18.6x, Trane 24.0x and Vertiv 36.5x. On tariffs, the current Section 232 regime taxes commodity steel and aluminum at 50%, derivative goods at 25% and copper derivatives at 25% (cut in April 2026). That raises AAON's input costs but shields it from imports; Q2's margin miss was partly raw-material and freight price/cost lag in coils.

Bull case

  1. Data-center cooling keeps compounding. BASX backlog is $1.43bn (+185%), and management calls the pipeline "the strongest that it's ever been." It spans liquid cooling, airside and chillers with new customers. Plays out if hyperscaler and neocloud capex keeps growing through 2028. Model: rev_growth, exit_ev_ebitda
  2. Heat pumps take rooftop share. Alpha Class orders rose 54% year to date, and AAON-branded bookings rose 16% in a soft market. Plays out if electrification codes and refrigerant changes favor AAON's semi-custom heat pumps. Model: rev_growth
  3. Margins return to peak. EBITDA margin was 24.1% in FY2023 and 22.8% in FY2024, against 14.8% in Q2 2026. Better-priced backlog ships in Q4, and utilization of the new plants rises. Plays out if gross margin returns to ~30% in 2027–28. Model: ebitda_margin

Bear case

  1. A data-center capex digestion year. More than half of sales now depends on a handful of large, lumpy data-center programs. Plays out if hyperscalers pause, leaving 2027 flat and 2028 down 10% once the backlog ships. Model: rev_growth, wacc
  2. Custom cooling turns price competitive. Vertiv, Schneider, Trane and new entrants are all adding capacity. BASX coil gross margin was already 16%. Plays out if margins stay near 13–15% and the enlarged plant base is under-absorbed. Model: ebitda_margin
  3. The multiple goes back to HVAC. At 26x TTM EBITDA, AAON is priced closer to Vertiv than to Lennox or Carrier. A slowdown would re-rate it to a cyclical HVAC multiple. Model: exit_ev_ebitda

Valuation & balance sheet

Metric (definition) Current Own history (range or 5y avg) Peers Source, as-of
EV/EBITDA, aggregator (TTM, lease-inclusive) 26.0x n/a (unverified) LII 12.7x · CARR 18.6x · TT 24.0x · VRT 36.5x stockanalysis.com, 2026-09-25
EV/EBITDA, model basis (EV $7.92bn, no leases) 34.9x FY2025 · 26.4x TTM · ~21.4x FY2026 guide (est.) n/a n/a (peers not restated) model-inputs.json
Forward P/E (stockanalysis consensus) 32.8x n/a (unverified) LII 15.1x · CARR 18.3x · TT 27.5x · VRT 32.2x stockanalysis.com, 2026-09-25
FCF yield (TTM) −1.6% (FCF −$118.8mm; capex $205.3mm) FCF −$3.1mm FY2024, −$190.0mm FY2025 n/a stockanalysis.com, 2026-09-25

Model-implied value range (from model-summary.json; generic module, Gordon-growth DCF and exit-EV/EBITDA DCF, midpoints): Bear $11.06 · Base $50.11 · Bull $108.15 per share, i.e. implied returns of −87.6% / −43.8% / +21.4% vs $89.10. These ranges show how the bull and bear drivers translate into value; they are not price targets. The price sits much closer to the bull case than the base. The market is paying for data-center growth beyond the backlog plus a near-peak margin recovery.

Re-rating. The base exit multiple of 16x is a 54% de-rating from 34.9x FY2025 EBITDA, and 39% from 26.4x TTM, on the model's own basis. The two methods disagree sharply. The exit method gives $72.96 and the Gordon method $27.26, because an 11.7% WACC with 3.5% terminal growth, applied to FCF still weighed down by capex and working capital, embeds a much lower terminal multiple. Even the more generous exit method sits 18% below the price. The bull exit method ($155.79, at 21x) is the price's best argument.

Tail. BASX is 55% of Q2 sales, so a failure sensitivity was run on a scratch copy. Revenue +55% / −15% / −20% / 0% / +3%, EBITDA margin sliding to 10–12%, NWC 36% of revenue, a 9x exit multiple and a 13% WACC give $3.81 (−95.7%).

Balance sheet: net leverage 2.0x FY2025 EBITDA (model Credit; ~1.5x TTM; covenant ratio 1.46x vs a 3.0x limit), coverage 9.8x, liquidity $163.7mm of undrawn revolver and essentially no cash ($13 thousand plus $12.7mm restricted). Nearest material maturity: the $600mm revolver ($435.0mm drawn, ~5.2%) in May 2030; NMTC obligations $28.9mm. Ratings n/a (unrated).

Model note:

Relative value vs peers
MetricAAONLIICARRTTVRTPeer medianvs median
P/E (TTM)45.5x15.8x40.2x35.0x55.6x37.6x+21%
P/E (forward)31.3x14.4x18.3x28.5x31.1x23.4x+34%
PEG1.1x1.4x1.7x1.9x0.8x1.5x−26%
EV/EBITDA (TTM)25.1x12.2x18.4x24.8x35.1x21.6x+16%
EV/Sales (TTM)3.9x2.7x2.6x4.8x8.2x3.7x+4%
P/B7.0x9.6x3.5x12.0x19.7x10.8x−35%
FCF yield-1.7%6.0%4.2%3.4%3.1%3.8%−5.5pp
P/E (TTM)VRT55.6xAAON45.5xCARR40.2xTT35.0xLII15.8x
P/E (forward)AAON31.3xVRT31.1xTT28.5xCARR18.3xLII14.4x
PEGTT1.9xCARR1.7xLII1.4xAAON1.1xVRT0.8x
EV/EBITDA (TTM)VRT35.1xAAON25.1xTT24.8xCARR18.4xLII12.2x
EV/Sales (TTM)VRT8.2xTT4.8xAAON3.9xLII2.7xCARR2.6x
P/BVRT19.7xTT12.0xLII9.6xAAON7.0xCARR3.5x
FCF yieldLII6.0%CARR4.2%TT3.4%VRT3.1%AAON-1.7%

As of 2026-10-09. One source and one definition for every company: aggregator TTM and consensus-forward multiples, lease-inclusive EV. Not the model's own EV basis (see the report's valuation table). Quotes taken 11:30-11:47 ET with the market open, so multiples reflect intraday prices. Dashed line = peer median. Source: stockanalysis.com /stocks/<ticker>/statistics/, fetched 2026-10-09.

Notes. PEG is the source's figure and rests on analyst growth estimates; leave null when no credible multi-year growth estimate exists.

Scenario stress test

Reasoned from the bull/bear drivers above. The model column comes from the scenario overlays (Base case + shock).

Scenario Effect Mechanism Magnitude Model Δ value vs Base ($/sh)
S1 Fast equity crash − High-beta (1.41) AI-adjacent name de-rates; backlog unaffected Med −$7.32
S2 Slow bear / recession − Commercial construction and data-center capex slow; operating leverage on the new plants. Milder than Base − Bear ($39.05) on purpose: transitory, while the bear case is a permanent end to the data-center cycle High −$15.49
S3 Rapid rate shock − Higher discount rate; costlier financing slows developer-led data-center and commercial projects Med −$7.45
S4 Slow rate grind − Same channel, grinding: construction starts ease Med −$6.37
S5 Soft-landing cuts + Lower rates re-rate growth and help commercial construction High +$9.04
S6 Recession-driven cuts − Demand and capex cuts outweigh the lower discount rate High −$9.04
S7a Credit liquidity shock − Forced selling of high-beta names; revolver leverage modest Low −$3.91
S7b Slow default cycle − Tighter credit for neocloud and developer data-center builders; floating revolver cost rises Low −$4.50
S8 Stagflation − Steel, copper and freight inflation against a fixed-price backlog (the Q2 lag), plus a higher discount rate Low −$4.57
S9a Dollar spike 0 No material effect, not modeled: US-centric production and customers — $0.00
S9b Dollar slide 0 No material effect, not modeled — $0.00
S10 Melt-up + AI-infrastructure momentum extends; high-beta name re-rates Med +$7.32
S11 Energy supply shock − Freight and energy input costs; no direct demand channel Low −$0.09
S12 Mega-cap/AI derating − Hyperscaler capex pullback hits BASX orders directly, and the AI-adjacent premium unwinds High −$12.16

Currently active/on watch per the playbook: state.md lists S3 as partially active; the macro log records all three legs crossed from 2026-09-24 (10Y 5.17% on 2026-09-25). S8, S10 and S11 are on watch. Here S3 and S10 are Med, S8 and S11 are Low.

Model value change vs Base, by scenario
S2 Slow bear / recession−$15.49S12 Mega-cap/AI derating−$12.16S6 Recession-driven cuts−$9.04S3 Rapid rate shock−$7.45S1 Fast equity crash−$7.32S4 Slow rate grind−$6.37S8 Stagflation−$4.57S7b Slow default cycle−$4.50S7a Credit liquidity shock−$3.91S11 Energy supply shock−$0.09S9a Dollar spike$0.00S9b Dollar slide$0.00S10 Melt-up+$7.32S5 Soft-landing cuts+$9.04

What would change the call

Upgrades if:

Downgrades (conviction up) if:

Watch items

Sources