AAON just filed its FY2025 10-K and the numbers tell two completely different stories depending on which page you're reading.

The headline financials look terrible. Revenue up 20% to $1.44B but gross margin collapsed 640bps to 26.7% — gross profit actually declined $11.4M on 20% more revenue. Net income down 36% to $107.6M. Operating cash flow: $534K. Not $534 million. $534 thousand. On $1.44 billion in revenue. The company is running on its revolver ($398M drawn, up from $139M) with essentially zero cash on the balance sheet.

The backlog tells a completely different story. Total backlog surged 111% to $1,828.5M — that's 1.27x forward revenue coverage. The BASX data center cooling subsidiary drove all of it: BASX-branded backlog up 141% to $1,302.1M. Named hyperscaler customers include Microsoft, AWS, Google Cloud, QTS, and Applied Digital. BASX-branded revenue across both reporting segments hit $547.8M, up 143% YoY. Management launched a new Coolant Distribution Unit supporting >100kW rack density for AI/HPC workloads.

So which story is real?

The Factor Decomposition

AAON at $88.33 (76x trailing, RSI 30, 16.1% short interest) is five bets bundled together:

F1: AI/DC Capex Cycle (≈45% of variance). This is the dominant driver and it's pure sector beta. AAON's idiosyncratic variance is 46.6% — well below the 75% threshold where stock-specific insight matters. The BASX backlog surge is replicated across every competitor: Vertiv backlog +108% to $15.0B, Modine DC revenue +78%, JCI Americas orders +56%, Carrier DC orders 5x+, Gates liquid cooling +700% YoY. This is the most broadly confirmed demand signal we track — 10+ independent layers of the datacenter infrastructure stack all capacity-constrained simultaneously. Hyperscaler capex tracking $600-645B for 2026 with zero DeepSeek impact. ETN says 11 years of demand runway. No edge here. Everyone sees it.

F2: Commercial HVAC Cycle (≈8% of variance). Core AAON Oklahoma ($801M, 56% of revenue) is tied to commercial construction, currently soft. "Leading indicators signal stabilization, no clear reacceleration." Macro, not our lane.

F3: Execution Recovery (≈25% of variance). This is the ONLY idiosyncratic factor. Three simultaneous one-time headwinds hit 2025: (1) ERP go-live at Longview in April caused coil supply shortages that propagated upstream, (2) refrigerant transition from R-410A to R-454B disrupted H1 supply chains, and (3) Memphis plant startup charged $16.1M in overhead to the AAON Oklahoma segment while all Memphis revenue flows to BASX — a structural accounting drag that depresses Oklahoma margins until Memphis is fully utilized.

The critical comp: Vertiv held gross margins flat at 36.3% while growing 28%. Carrier launched CDUs with "no drag" on margins. This proves data center cooling products don't inherently compress margins. AAON's 640bps collapse is company-specific execution, not structural to the product category. That's actually the bullish read on F3 — margin recovery is structurally plausible.

But management credibility on margin timing is low. The Q3 2025 earnings call guided FY2025 gross margin at "28% to 28.5%." Actual: 26.7%. They missed by 130-180bps. The Q4 2024 call promised "sequential-over-quarter improvement margin 2025" — margins deteriorated through H1 before improving in Q3. And there's a new wrinkle: Tulsa ERP go-live is planned for H2 2026. Tulsa is the main production facility. Management says "minimal disruption on Longview learnings" but that's the same management that missed its own 2025 guide. This would be the third consecutive year of "one-time" disruption events (2024: Redmond growing pains, 2025: Longview/Memphis ERP, 2026: Tulsa ERP).

F4: Mix Shift (≈15% of variance). BASX (26.6% GM) is growing 143% while core AAON (29-37% GM) is shrinking 7%. As BASX becomes >50% of revenue within 2-3 years, consolidated margins converge toward BASX margins. The question is whether BASX margins scale UP toward 30-35% (VRT proves 36% is achievable) or stay at 26-27% (in which case every dollar of growth dilutes). Unanswerable from public data.

F5: Working Capital Normalization (≈7% of variance). The OCF collapse is mechanical — BASX hyperscaler contracts have longer payment terms. AR +$167M, contract assets +$111.8M. VRT's working capital is a cash source at $10.2B scale (customers prepay). AAON's is a cash drain at $1.4B scale. This normalizes as backlog converts to revenue. Predictable, not insightful.

What the Bears See (and Why They Might Be Right)

Short interest is 16.1% — six times higher than Vertiv's 2.5%. Bears are targeting AAON specifically, not the DC cooling sector. Their case:

• 76x trailing P/E on depressed earnings. Even on "normalized" EPS of $2.00-$2.50, that's 35-44x — expensive for a company with execution problems.
• OCF collapse with no clear timeline for recovery. The revolver is the only lifeline.
• Sub-scale vs Vertiv (7x smaller revenue, worse margins, worse cash flow, worse diversification).
• Serial "one-time" disruptions that keep extending the margin recovery timeline.
• No insider buying. The CFO has an active 10b5-1 selling plan (41,565 shares through March 2027). Nobody is stepping up during the RSI 30 selloff.

Cross-corpus transcript search for "AAON" or "BASX" across all companies: zero matches. No competitor mentions them by name. AAON is invisible to the players that matter. Vertiv is discussed by name in multiple competitor transcripts.

What the Bulls See (and What Has to Go Right)

The bull case is a margin recovery story with enormous backlog visibility:

• $1.83B backlog at 1.27x revenue coverage = 18-24 months of demand visibility.
• Three one-timers (ERP, refrigerant, Memphis) all clearing simultaneously in 2026.
• VRT comp validates that 30%+ margins are structurally achievable at scale.
• Dividend raised 25% (to $0.40 annualized) despite EPS falling 36% — board conviction signal.
• Pricing power preserved: 9% total increases (3% + 6% tariff surcharge) fully realized in a soft market.
• If GM recovers to 30-32% on $1.8B+ revenue, earnings power is $2.50-$3.50+ EPS. At $88, that's 25-35x forward — reasonable for a company growing into AI infrastructure demand.

The key test is Q1 2026 earnings. Does gross margin recover toward 28-30%? That's the doorway collapse. We set this at 50/50 — management's track record on margin timing doesn't earn the benefit of the doubt.

The Sizing Math

Even in the bull case, the factor decomposition kills it as a position:

Only 25% of AAON's variance comes from the execution recovery factor (F3) — the only one with any idiosyncratic edge potential. The other 75% is sector beta (DC capex cycle), macro (HVAC), structural unknowns (mix shift), and accounting math (working capital). Annualized alpha on a $88→$120 thesis over 18 months with 25% edge exposure: ≈4.6%. That's a 2.3% portfolio weight at best.

If you want data center cooling exposure, Vertiv is the dominant player with 36% margins held flat, $2.1B OCF, $15B backlog, and pricing power. Or buy a basket. AAON is the sub-scale player with execution problems trading at the same multiple as the market leader.

Predictions

Where This Filing Has Value

AAON isn't a position candidate. But the filing strengthens the broader AI infrastructure demand thesis — confirmation from yet another independent layer of the DC supply chain. The hyperscaler capex acceleration, 11-year demand runway, and zero DeepSeek impact all corroborate positions in the power/compute layer.

More importantly, AAON's BASX backlog is now a leading indicator to monitor. If BASX backlog declines or cancellations spike at Q1 2026 earnings, that's an early warning signal for the entire DC infrastructure stack — including any positions in the power, electrical, or compute layers that share the same demand driver.