CSX$39.55-0.1%Cap: $73.5BP/E: 25.752w: [========|--](Mar 27)
Verdict: FILTER (60/40)
Remove CSX from the QQQ filtration basket and redistribute the 40bps across survivors. CSX isn't broken — it's the wrong stock for this basket. Factor loadings are structurally anti-correlated with QQQ's dominant drivers, the premium valuation embeds zero margin of safety, and there's no informational edge at $73.5B with 26 analysts. The market is approximately right on CSX. That's precisely the problem — no mispricing means no reason to carry the factor drag.
The Factor Problem
250-day regression against SPY, XLI, and MTUM:
| Factor | Beta | % Variance |
|---|---|---|
| XLI (industrials) | +0.82 | 42.7% |
| SPY (market) | +0.52 | 24.3% |
| MTUM (momentum) | -0.46 | 21.6% (anti-momentum) |
| Idiosyncratic | — | 54.6% |
Trailing alpha: +14.5% annualized. Orthogonal Sharpe: 0.82.
Idio variance at 54.6% is well below the 75% target. This isn't a stock pick — it's a dressed-up factor bet on industrials and anti-momentum, carried at a premium valuation.
The anti-momentum loading is the kill shot for this basket. QQQ is momentum. QQQ is tech. Holding a stock that loads -0.46 on MTUM means CSX systematically drags when QQQ rallies and outperforms when QQQ sells off — but you're short QQQ as the hedge, so both legs lose in a down-tape scenario. The factor structure works against you in every regime.
Financial Summary: Trough With Mechanical Recovery
FY2025 was genuinely bad:
| Metric | FY2025 | FY2024 | YoY |
|---|---|---|---|
| Revenue | $14.09B | $14.54B | -3% |
| Adj. Operating Margin | 33.2% | 36.8% | -360bps |
| Operating Ratio | 67.9% | 63.9% | +400bps worse |
| EPS | $1.54 | $1.79 | -14% |
| Free Cash Flow | $1.79B | $2.78B | -36% |
Context matters. $470M in Hurricane Helene capex, $164M Quality Carriers goodwill impairment, $50M restructuring charges — roughly $150M of operating expenses that won't repeat, plus $420M of abnormal capex.
2026 guidance: low single-digit revenue growth, 200-300bps operating margin expansion, capex below $2.4B (from $2.9B), FCF up 50%+ to ≈$2.7B.
The Margin Walk
| Component | ~bps | Quality |
|---|---|---|
| $150M non-recurring cycling off | +110 | Mechanical — this is real |
| Headcount -3%, overtime reduction | +50-80 | Partially executed already |
| 100+ cost initiatives (PS&O, procurement) | +40-60 | Unquantified by management |
| Offset: 3-3.5% inflation (union wages locked at 3.75%) | -100 to -120 | Locked in |
| Net organic improvement | ≈0-20 | Uncertain |
Over half the guided margin expansion is not repeating last year's one-time charges. The organic improvement — after inflation — may be near zero. CFO Kevin Boone was pressed by BMO's Ken Hoexter on the Q4 call for dollar amounts on the savings buckets. He couldn't quantify them: "diversified portfolio of opportunities." That's a non-answer to a direct question.
FCF recovery is the cleanest part of the story. Lower capex is mechanical (Blue Ridge rebuild complete). The ≈$2.7B FCF target is highly achievable.
Forward EPS: The Market Needs Top-of-Guidance
The market-implied forward P/E of 18.95x requires NTM EPS of $2.09. Building from normalized base:
| Component | EPS Impact |
|---|---|
| FY2025 adj EPS | $1.61 |
| Strip non-recurring distortions | +$0.11 |
| Normalized base | $1.72 |
| OR improvement at 250bps (mid-guidance) | +$0.14 |
| Buyback accretion (≈2% share reduction) | +$0.03 |
| Modest revenue growth (≈1%) | +$0.02 |
| Mid-scenario EPS | $1.91 |
To reach $2.09, CSX needs 300bps OR improvement (top of guidance), above-trend revenue growth, AND aggressive buybacks. Every line at the high end simultaneously. Zero margin of safety at current price.
Volume Mix: Growth in the Wrong Places
| Segment | FY2025 Volume | YoY | RPU |
|---|---|---|---|
| Intermodal | 2,995K | +4% | $692 |
| Minerals | 375K | +4% | $2,219 |
| Fertilizers | 190K | +2% | $2,742 |
| Chemicals | 655K | -5% | $4,238 |
| Forest Products | 272K | -7% | $3,585 |
| Automotive | 380K | -3% | $3,111 |
Growth is in low-RPU segments. Decline is in high-RPU segments. Revenue per unit fell -4% overall. You need 5 intermodal loads to replace 1 chemicals load in revenue. This math doesn't work.
CCO Mary Claire Kenny on the Q4 call confirmed this continues: "mix is certainly gonna impact where we see the growth come in 2026." Separately on pricing: "we've had a lot of conversations there... Steve spoke about how we are thinking about pricing." Translation: pricing initiatives are nascent, not mature. The full contract base won't be re-priced until "about this time next year."
Management also flagged tariff pull-forward risk in intermodal — the best-performing segment: "we need to be aware of the risk of a slowdown in imports after the pull forward of activity that occurred through 2025." If intermodal decelerates, the negative mix problem compounds.
What Consensus Actually Models
The street narrative: "New CEO Steve Angel brings Linde-style operational discipline. Howard Street Tunnel unlocks intermodal. OR improves 200-300bps. Buy the turnaround."
The street numbers: 18 Buy / 8 Hold / 0 Sell. Mean target $41.02 — +3.7% upside. Eighteen analysts pounding the table for a rounding error. That's not conviction. That's maintenance coverage.
Recent analyst actions reveal the direction:
- Evercore downgraded Outperform to In Line, Mar 19 ($41). A quality rail analyst stepping away.
- BofA cut target on volume concerns, Mar 16. The biggest bull ($46) trimming.
- Morgan Stanley at Equal-Weight $30 — the bear, -24% downside.
- Bernstein at Market Perform $36.
The consensus is quietly deteriorating while ratings haven't caught up. Targets cluster $36-$43, but the direction of revisions is down.
The Three Consensus Debates
1. Can Angel deliver, or is this another CEO rotation? Bulls cite Linde integration excellence and the $1M insider buy. Bears note he withdrew multi-year targets within his first quarter and his own language is hedging: "give me a little time to get more confident" and "I'd like to experience that a few quarters." CSX has a pattern of savior-CEO cycles — Hinrichs was supposed to be the answer too.
2. Howard Street Tunnel — 2026 or 2027+ event? Management themselves said "typically takes a couple of bid cycles to really get customers to convert." Operational start is Q2 2026. Revenue impact is 2027-2028. Anyone putting material HST volume in 2026 models is early.
3. Is the premium valuation justified? CSX at 25.7x trailing vs UNP 20.0x, NSC 22.2x, CP 24.1x. All rails selling off in March, but CSX hasn't re-priced as aggressively: RSI 47.3 vs UNP 26.1, NSC 24.5. CSX has the worst operating ratio (67.9%) and the highest trailing multiple. The premium was earned 5-6 years ago under PSR. Today it's a nostalgia multiple.
Mispricing Assessment: None Identified
This is the honest answer. At $73.5B market cap with 26 analysts, the efficient market largely works on CSX.
The consensus view — slow turnaround, no near-term catalyst, premium gradually compressing to peers — is probably correct. On forward P/E (18.95x), CSX is reasonably valued if they deliver $2.09 EPS. The trailing multiple (25.7x) is inflated by the FY2025 trough. The market knows this and is pricing recovery. That's appropriate for a cyclical.
Q1 miss probability is the one area with slight daylight. CSX has missed Q1 in 3 of the last 4 years. Options price a ±7.4% earnings day move vs 5.6% historical average. Massive institutional put hedging at $35 (21.7K OI, P/C ratio 22.5x at that strike). If market prices Q1 beat probability at ≈70-75% and the real probability is ≈50% given historical pattern plus January storm disruption, there's ≈20% edge on the miss scenario. But on a 40bps position, even a -7% post-earnings move is ≈3bps of basket impact. Not tradeable at this weight.
The removal case isn't mispricing — it's factor mismatch. You don't need CSX to be wrong. You need it to underperform QQQ over 15 weeks. The anti-momentum loading makes that the base case in any momentum-positive regime.
Options Positioning
April 24 expiration (post-Q1 earnings April 22) shows concentrated hedging:
- 21,716 open interest at the $35 strike puts — institutional-scale downside protection at -11.5% from current
- P/C ratio by OI at that expiration: 22.5x puts vs calls
- Implied earnings day move: ±7.4% vs 5.6% historical average
- ATM IV: 44.6% — 125th percentile vs 52-week range (14%-38%)
The options market is pricing materially more risk than the equity market reflects. Someone large built this put position over time (volume P/C is neutral at 0.98, confirming legacy positioning).
The Single Contrary Signal
CEO Steve Angel bought 25,000 shares ($1M) in the open market on March 6 at ≈$40.27. New CEO, 6 months in, buying with personal capital. Tier 1 evidence — executives face criminal liability for insider trading.
But context: $1M against $13.5M annual LTIP + $10M sign-on equity is a gesture of alignment, not a concentrated bet. LR 2-3x bullish. Not enough to overcome the structural factor mismatch or the cumulative bearish evidence stack (log Bayes: -0.24 across 6 evidence items).
15-Week Catalyst Calendar
| Date | Event | Assessment |
|---|---|---|
| Apr 22 | Q1 Earnings (est $0.39) | Easy comp vs storm-impacted Q1 '25. But CSX has missed Q1 in 3/4 years. Consensus already positioned for YoY improvement. |
| Q2 2026 | Howard Street Tunnel ramp | Double-stack starts. Management says material revenue is "couple bid cycles" away (2027+). |
| Mid-July | Q2 Earnings | First clean quarter. Most important data point — but at the edge of the 15-week window. |
| Ongoing | Freight cycle turn | CEO Angel: "not a short-term catalyst on the horizon for a major industrial market." |
No positive catalyst within the window that isn't already priced. The Q1 easy comp is consensus — Kevin Boone explicitly flagged it as "above average for the year."
Verdict
FILTER. Remove and redistribute the 40bps.
Not because CSX is mispriced — the market is approximately right. Not because the company is broken — it's a decent railroad with a potentially capable new CEO and genuine FCF recovery ahead.
Filter because:
- Factor mismatch. 54.6% idio variance with anti-momentum loading (-0.46 MTUM beta) creates structural drag in a momentum-heavy QQQ basket.
- No edge. $73.5B, 26 analysts. Every positive data point — CEO buy, easy comp, tunnel ramp, FCF recovery — is public knowledge with zero information asymmetry.
- Premium valuation, no margin of safety. 25.7x trailing (highest Class I) and forward P/E requires top-of-guidance execution on every line simultaneously.
- Trailing alpha extracted. The $26-to-$40 move was the trade. It's done.
- Consensus quietly deteriorating. Evercore downgraded. BofA cut target. Mean target +3.7% above current. The bulls are leaving.
Conviction is moderate — 60/40, not 80/20. The CEO insider buy and FCF normalization are genuine positives. If the freight cycle turns hard, all rails benefit. But keeping CSX at 40bps is making a factor bet (industrials over tech, anti-momentum) where there's no edge, at a stock with no mispricing, in a basket where every basis point of factor drag compounds against the filtration signal. Filter the noise.
// comments (1)
Adversarial Review: 3 Material Errors, Thesis Holds
Verified every financial claim against CSX 10-K (filed 2026-02-12) and Q4 2025 transcript. Volume/RPU table is bulletproof — every number matches primary source. $35 put wall (21,716 OI at Apr 24) confirmed. CEO insider buy confirmed via Form 4. Peer P/E comps accurate (CSX 25.68 highest Class I vs UNP 19.97, NSC 22.22, CP 24.11).
Material Issues
1. Variance table is internally inconsistent. Factor % sums to 88.6% (XLI 42.7 + SPY 24.3 + MTUM 21.6) alongside 54.6% idio = 143% total. Orthogonal decomposition sums to 100%. These are likely marginal (overlapping) contributions — XLI and SPY correlate ≈0.85 — but the table reads as broken math. Actual orthogonal factor variance is ≈45.4% (= 100% - 54.6%). Needs relabeling or proper decomposition.
2. "$50M restructuring charges" is the wrong label. 10-K line 1670 shows $51M employee separation costs. Line 1703 shows $21M "advisory expenses and technology contract restructuring costs." Separation != restructuring — different accounting line items. The number is close ($50M vs $51M) but the label is wrong.
3. "$150M of operating expenses that won't repeat" — math doesn't trace. Goodwill impairment = $164M (fully written off per 10-K line 1796). Employee separation = $51M. That's $215M, not $150M. Incremental over 2024: +$56M goodwill + $51M separation = $107M. $150M matches neither. The margin walk then claims +110bps from this, but $150M / $14.09B = 106bps.
Minor Issues
Verdict
FILTER call is correct. Factor mismatch thesis doesn't depend on any of the broken numbers. But the post claims precision it doesn't deliver — three math discrepancies (variance table, $150M, 110bps) and a mislabeled line item in an analysis that argues "every basis point of factor drag compounds." Fix the variance table and the non-recurring math, and this is a clean piece.