conf 2/10 partial
enthusiasm:18.0 · trend:8 · quantifies:0 · impact:0 · under_radar:14 · credibility:0 · business_impact:4 · disruption:0 · commitment:0 · confirmation:3
Enthusiasm latest 6 / prev 5 (rising)
Owens Corning’s AI story is bifurcated: internal use centers on predictive manufacturing (20k+ sensors, ~40 plants) and, as of the prior call, supply-chain optimization agents in North American insulation with planned enterprise scaling and future generative/agentic AI under a new CIO—without dollarized savings or margin attribution. The latest call adds an end-market angle: insulation nonres demand from AI/data-center buildout is strong enough to constrain some product lines, but management does not quantify AI-linked revenue. Credibility is moderate on operations (concrete deployment scale) and low on financial impact (no P&L numbers, no analyst-driven AI disclosure in Q&A beyond demand color).
Grounded on actual base — revenue $10.1B · net income $-0.5B · net margin -5.2% · diluted EPS -6.26
These are next-fiscal-year annual uplift estimates, not next-quarter numbers.
Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: % · vs analysts: unclear · priced in: low (model's call-read: high; verdict above is the hard-data one used for ranking) · confidence: 2/10
| Claim | Figure | Arithmetic | Next-FY Rev % | Next-FY EPS % |
|---|---|---|---|---|
| 20,000+ process sensors monitored by AI for real-time risk alerts productivity · soft | over 20,000 sensors | Operational count, not a dollar figure. No $ saving, % productivity, or FTE delta disclosed; cross-claim 20,000 sensors ÷ ~40 plants ≈ 500 sensors/plant is deployment intensity only. Cannot compute after_tax_saving = saving_$ × (1-0.21) with saving_$ undisclosed. EPS denominator is GAAP NI = -$522M (loss), voiding any eps_uplift_pct. No revenue/cost base for this claim appears anywhere in inputs — genuinely unanchored. | ||
| AI monitoring deployed in ~40 plants across 3 businesses, expanding productivity · soft | nearly 40 plants | Deployment-footprint count across Roofing, Insulation, Doors, not a dollar figure. Quality/yield/uptime benefit is plausible but unquantified ('predict risk before it impacts safety, quality or productivity'); no revenue lift, cost saving, or margin $ disclosed. Cannot size; same loss-making EPS denominator problem. | ||
| Nonres product lines sold out / near sold out on AI/data-center demand revenue · soft | 'close to being sold out or are sold out in specific product lines' (no revenue $) | Revenue tailwind comes from OC selling building products INTO the external AI/data-center buildout (picks-and-shovels SUPPLIER position), not AI improving OC's own products — tagged supplier-side and excluded from the adopter headline. Sold-out status implies full utilization but not the $ magnitude of incremental vs baseline. Management gives no revenue $, no affected-segment size, no incremental volume: rev_uplift_pct = 100×(undisclosed incr_$/$10,103M) is uncomputable. Unanchored — soft. |
Assumptions: No phasing applied — none of the three claims provide a next-FY revenue or savings $. Defaults not invoked because no claim is dollar-anchored. For any hypothetical revenue claim, incremental margin would use operating margin ($1,720M/$10,103M = 17.0%), not GAAP net margin (-5.17%, distorted by the one-off-driven -$522M net loss); tax rate 21% for any hypothetical cost saving. EPS sizing would use consensus adjusted NI base (~$1,032.8M FY25E, EPS $12.21), NOT GAAP -$522M NI, which makes any EPS-uplift % meaningless per guardrail and is set null. Segment mapping: claims 1-2 → internal manufacturing ops (cost/quality, adopter); claim 3 → nonres roofing/insulation demand (supplier into buildout). Adopter aggregate excludes the supplier-side data-center demand claim.
Top line: Adopter-side topline uplift: 0% computable. The only revenue-adjacent signal is supplier-side — sold-out nonres product lines feeding AI/data-center/reindustrialization construction — but management gave zero revenue $, no affected-line size, and no incremental volume, so 100×claim_$/$10,103M is uncomputable and supplier_rev_uplift_pct is also null. This is a demand tailwind OC benefits from as a materials supplier into the buildout, not AI improving OC's own products; per the adopter-ranking rule it is excluded from the headline and cannot be sized. est_rev_uplift_pct = null.
Bottom line: Adopter-side bottom-line uplift: 0% computable. The 20,000-sensor / ~40-plant operational AI deployment (≈500 sensors/plant) is a real footprint targeting safety/quality/productivity but carries no disclosed opex savings, downtime-avoidance $, or productivity %, so after_tax_saving = saving_$ × 0.79 cannot be computed. Separately, OC is GAAP loss-making (NI = -$522M, -5.17% margin), so EPS-uplift % off that base has no valid denominator; even against consensus adjusted NI of ~$1,032.8M (EPS $12.21) there is no savings $ to flow through at 21% tax. est_eps_uplift_pct = null per the loss-making guardrail.
[impact n/m (all claims soft/unanchored); EPS uplift n/m (loss-making base)] Consensus FY25E: revenue ~$10,132M (+0.29% vs actual $10,103M), EPS $12.21, NI ~$1,033M (12 analysts). Trajectory: FY23→FY24 revenue +10.4% ($9,872M→$10,902M), FY24→FY25E -7.1% ($10,902M→$10,133M); EPS peaked $15.51 in FY24, falling to $12.21 in FY25E (-21.3%). With zero hard adopter-side rev/eps uplift computable (all three claims soft), there is no quantifiable gap above consensus. The supplier-side sold-out/data-center demand narrative is a known cyclical demand fact already embedded in the nonres strength behind the -7% revenue normalization vs the FY24 peak — not new unmodeled upside. Forward statements (supply-chain AI agents, generative/agentic AI via new CIO, plant-monitoring expansion) contain no $ anchors. Net: no positive consensus gap can be demonstrated in numbers — treat as priced in.
partial
Both answers fully agree on all numbers (all null), soft flags, priced_in=high, unclear, confidence 2. Only minor type-label and margin-basis wording differed.
| Field | Opus 4.8 | GPT-5.5 |
|---|---|---|
| Rev uplift % | – | – |
| EPS uplift % | – | – |
| Priced in | high | – |
| vs analysts | unclear | – |
| Confidence | 2 | – |
| Top line | No sizable adopter-side AI revenue is quantified. The only AI-linked revenue signal is supplier-side: nonres product lines feeding data-center/reindustrialization demand are 'sold out or close to sold out,' but management gave zero revenue $, no affected-line size, and no incremental volume — so 100*claim_$/$10,103M is uncomputable. This is a demand tailwind OC benefits from as a materials supplier into the buildout, not AI improving OC's own products; per the adopter-ranking rule it is excluded from the headline and cannot be sized either. est_rev_uplift_pct = null. | – |
| Bottom line | The two adopter AI initiatives (20,000+ sensors; ~40 plants) target safety/quality/productivity and likely yield real but undisclosed cost/yield benefits. None carries a $ saving or % productivity figure, so after_tax_saving = saving_$ x 0.79 cannot be computed. Separately, OC is GAAP loss-making (NI = -$522M), so an EPS-uplift % has no valid denominator — even a known saving would produce a meaningless/negative figure. est_eps_uplift_pct = null per the loss-making guardrail. | – |
| Reasoning | Nothing AI-specific is quantified to dollars, so there is no incremental figure to compare against the consensus trajectory. Consensus already carries the nonres/data-center demand: FY25 revenue avg $10.13B (vs $10.10B actual) and net income avg ~$1.03B — the sold-out commentary is a known cyclical/demand fact embedded in those numbers, not a new uplift. The internal sensor/plant AI program is an efficiency story with no disclosed magnitude, so it adds no measurable gap above consensus. Net: no positive consensus gap can be demonstrated in numbers -> treat as priced in. | – |
Rows highlighted where the two models disagreed.
—/100 (no quantified promises) no-quantified-promises 6 calls reviewed
Across six calls Owens Corning never made a quantified AI promise (number plus timeframe); AI talk is limited to qualitative scaling of supply-chain agents and later operational metrics (20k sensors, ~40 plants) with no prior measurable targets to score.
Est. revisions flat · Fwd P/E 10.1 · EV/Sales 1.6x
AI claim maps to Roofing, Insulation, Doors
AI enthusiasm across 6 calls — trend ↗ rising
Five calls with no AI story; Q1 FY2026 added specific plant-sensor predictive monitoring across ~40 sites.
5/10 qualitative impact moderate medium-term · mixed evidence
Where AI matters: plant predictive maintenance & supply-chain ops
OC has credible adopter-side scale—20k+ AI-monitored sensors across ~40 plants plus insulation supply-chain agents—but no dollarized savings or margin attribution, and data-center/nonres strength is supplier demand into the AI buildout, not AI improving OC’s own products.
Caveats: No disclosed opex, downtime, or EPS uplift from operational AI despite scale metrics; Generative/agentic AI and enterprise agent rollout remain forward narrative without P&L proof; Sold-out nonres lines on data-center demand are cyclical supplier tailwind, not durable adopter-side AI monetization; Industry-wide efficiency gains could modestly cap pricing power without commoditizing the product
AI DISRUPTION / CANNIBALIZATION RISK tailwind · 2/10
Core revenue is physical insulation, roofing, and doors where AI mainly optimizes manufacturing and logistics; it does not automate away installed building-envelope materials or structurally compress billable units the way services/labor-arbitrage models face.
option liquidity: fair
proxy inputs — dollar-ADV $161M · beta 1.35 · px $120.22
source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.