FIX — Comfort Systems USA, Inc.

Near-term market sentiment is clearly negative on price: as of 2026-10-15 the stock sits below its 20-, 50- and 200-day moving averages with RSI at 31.01 and a -16.8% one-week drawdown on no company-specific bad news, which points to a sector-wide de-rating of high-multiple AI/data-center-adjacent names plus profit-taking after a multi-year run from $770.47 to $2,073.99. Sell-side sentiment runs the opposite way — the consensus rating is Buy (5 buy / 4 hold / 0 sell / 0 strong buy) with a $2,082.50 average target, ~45% above the 2026-10-15 price. That divergence is valuation-driven, not fundamental. The only company-specific negative is routine: CEO Brian Lane trimmed a portion of his stake in late August 2026 while retaining a substantial holding.

The trend is unambiguous acceleration, not a blip. Revenue rose from $1.378B in Q3 2023 to $3.266B in Q2 2026 — a 2.4x increase in under three years — with YoY growth re-accelerating from ~19-20% in H1 2025 to +56.5% in Q1 2026 and +50.3% in Q2 2026. Diluted EPS tells the same story: $2.93 (Q3 2023) → $6.53 (Q2 2025) → $10.51 (Q1 2026) → $12.53 (Q2 2026), with H1 2026 EPS of $23.04 versus $11.28 in H1 2025 (+104%). Margins are expanding alongside volume, not just riding it: operating margin climbed from 13.8% in Q2 2025 to 17.1% in Q2 2026, and gross margin from 23.5% to 25.9%. Cash conversion is excellent — Q2 2026 operating cash flow of $1.139B against net income of $442M produced a deeply negative accruals ratio (-8.22%), i.e. earnings running behind cash, the conservative/healthy pattern. Backlog of $14.06B (vs $12.45B at Q1 2026) covers roughly 1.25x TTM revenue of $11.23B. The one honest caveat: this is a cyclical contractor riding an unusual capex wave, and 17% operating margins should not be extrapolated indefinitely.

Expensive on both absolute and own-history measures, though less extreme than a week ago. As of 2026-10-15 the TTM P/E is 35.2x versus a 5-year average of 25.2x (~40% premium), and TTM P/S is 4.49x versus a 5-year average of 1.87x (~140% premium); P/B is 15.7x and EV/EBITDA is 24.6x. On the forward number — $49.01 next-FY consensus EPS — the P/E compresses to ~29.2x, which is defensible for a business growing revenue 50% with 33.6% ROIC and $1.8B net cash. The headline PEG of 0.32 looks cheap only because it mechanically pairs a trailing multiple with trailing hyper-growth; it will normalize fast as growth decelerates. Verdict: not cheap — this is a pay-up-for-quality name where the dominant risk is multiple compression rather than an earnings miss.

The key signal is that the ~31% decline from $2,073.99 to $1,432.54 happened while estimates went UP — Q2 2026 beat and price targets were raised. That is pure multiple compression: the market is pre-emptively pricing a data-center capex digestion phase or peak-cycle earnings, not discounting current results. Practically, watch the backlog line in the 2026-10-22 Q3 print more than the EPS line. If backlog keeps climbing past $14B and margins hold near 17%, the stock now trades at ~29x forward earnings on a business compounding EPS 50%+ — a re-rating candidate back toward $1,700-2,000. If backlog growth stalls, the multiple has room to fall toward the 20-25x a cyclical contractor historically commands, which on $49 EPS is $1,000-1,225. The asymmetry has improved after the selloff, but it rests entirely on the next two backlog prints.

Decisively broken down as of 2026-10-15. RSI(14) at 31.01 is just above oversold, and price is -14.3% below the 20-day SMA (implied ~$1,672), -13.9% below the 50-day (~$1,663) and -10.6% below the 200-day (~$1,601) — below every major moving average, with a -16.8% one-week and -15.5% three-month loss. This is momentum-driven selling, not a gentle pullback, and RSI at 31 has not yet produced a confirmed washout bounce (sub-30 plus a reversal candle is the classic tell). Entry-timing implication: no technical support signal exists yet; reclaiming the 20-day SMA would be the first evidence of a bottom, and until then adding is catching a falling knife. These levels are as of 2026-10-15 and will have moved by the time this is read.

• AI/data-center capex cycle is the single dominant driver — hyperscaler and colocation construction spending is the swing factor for FIX's backlog, and any announced pause or digestion phase would hit the stock before it hits the income statement.
• Interest rates and cost of capital: elevated rates are still a drag on traditional commercial real estate and non-residential construction, but the mission-critical/data-center segment is largely rate-insensitive, which is why FIX is insulated from the broader construction slowdown.
• Bifurcated non-residential construction backdrop — conventional office/commercial spending is sluggish or down YoY while high-tech manufacturing and data-center work is booming; FIX is levered almost entirely to the strong side of that split.
• Skilled trades labor scarcity and wage inflation — MEP labor is the largest input cost, and fixed-price contract structures mean wage surprises hit margins directly.
• Rotation out of high-multiple AI-adjacent industrials: the -16.8% week through 2026-10-15 looks like a positioning/multiple unwind across the theme rather than anything FIX-specific, which makes the stock hostage to sentiment on AI capex broadly.
• Long-lead electrical equipment (switchgear, transformers) and tariff/trade exposure — supply-chain timing and input cost inflation can delay project recognition even with a strong backlog.

• Q3 2026 earnings on 2026-10-22 (official) — the backlog figure, margin trajectory and any commentary on data-center project timing are the primary valuation catalyst.
• Modular/prefabrication capacity expansion toward ~5 million square feet by mid-2027, which management projects unlocks up to $800M of incremental annualized sales capacity.
• Continued backlog growth: $14.06B at Q2 2026 versus $12.45B at Q1 2026 (+13% sequentially) — sustained sequential increases would validate the growth runway.
• Further sell-side upward revisions — consensus target of $2,082.50 (recent average $2,094, range $1,910-2,200) versus a 2026-10-15 price of $1,432.54, with KeyCorp and others raising targets after consecutive beats.
• Capital deployment optionality from a ~$1.8B net cash position — buybacks, bolt-on MEP acquisitions, or a dividend increase off a token 0.17% yield.

• Customer concentration in hyperscale data centers — a single large customer slowing or pausing projects would hit a $14B backlog quickly and violently; this is the biggest single risk to the thesis.
• Valuation compression: at 35.2x TTM P/E, 4.49x P/S and 15.7x book versus 5-year averages of 25.2x and 1.87x, FIX can fall hard on good news simply by de-rating, as the last week demonstrated.
• Cyclicality — engineering & construction is a cyclical industry and the current 17.1% operating margin is likely peak-cycle; a return toward mid-cycle margins would cut EPS materially even on flat revenue.
• Fixed-price contract and execution risk on large, fast-moving data-center projects, plus labor availability and wage inflation in a tight skilled-trades market.
• Insider selling — the CEO trimmed his stake in late August 2026; routine in isolation but a sentiment negative when it coincides with a 31% drawdown.
• Thin working-capital cushion: current ratio of 1.214 is only modestly above 1.0 and has been trending near that level; the $1.8B net cash position offsets this but does not eliminate project-timing cash swings.

This is an excellent operating business at a full price that has just been de-rated 31% — which makes it a reasonable, not a screaming, entry. Revenue +50%, EPS +92%, 17% operating margins, 33.6% ROIC, $1.8B net cash and a $14.06B backlog are about as good as engineering & construction gets, and the AI/data-center build-out is a multi-year rather than multi-quarter driver. The offset is that FIX is a cyclical contractor whose current margins are likely peak-cycle, and at ~29x forward EPS there is no valuation cushion if hyperscaler capex digests. I would be a buyer on further weakness or on a Q3 backlog confirmation, not a chaser at $1,432.54 given the broken technicals heading into the 2026-10-22 print.

Blended forward-multiple approach: 32x forward EPS of $49.01 (a ~27% premium to the 5-year average P/E of 25.2x, justified by structurally higher margins — 17.1% operating in Q2 2026 versus ~8% in 2023 — plus $1.8B net cash and 1.25x backlog coverage, but capped by contractor cyclicality) gives ~$1,568; cross-checked at 21x NTM EBITDA of ~$2.6B, which is $54.6B EV + $1.8B net cash over ~35.2M shares ≈ $1,600. Blended to ~$1,580, roughly 10% above the 2026-10-15 price of $1,432.54 and well below the $2,082.50 sell-side average.