Marathon Petroleum MPC

> ## ⚠ RESULTS RELEASED THE MORNING OF THIS DIVE

>

> Marathon Petroleum issued its second-quarter 2026 release on 2026-08-04 (8-K, Item 2.02, signed by CFO Maria A. Khoury) and filed its 10-Q the same day. Both are in this archive with tables preserved, so every figure below is filing-verified through 2026-06-30 — this is the freshest data set in the batch.

>

> The result: EPS $17.73 against a $14.27 consensus (+24.2%); revenue $52,337M against $40,866M (+28.1%). The stock closed +1.82%.

>

> A 24% beat that moves a stock 1.8% is a market telling you it does not believe the number is repeatable. The verdict below is set with that reaction as a direct input.

Reference table

1. The June 2026 quarter — filing-verified, and it is extraordinary

All figures from the consolidated statements of income in the 10-Q filed 2026-08-04, with tables preserved:

Note the composition of that leverage. Revenue rose 53.5% while cost of revenues rose 43.4%. The gross spread went from $3,774M to $8,930M. Depreciation, SG&A and other taxes rose a combined 4.4% — essentially fixed. That is a fixed-cost business receiving a price shock, and the operating leverage is 4.4x: a 53.5% revenue increase produced a 233% increase in operating income.

And note what did NOT change: noncontrolling interests took $400M against $394M, essentially flat, because that is MPLX's public unitholders' share of a midstream business whose earnings did not move.

The one number that explains it

From the 10-Q's Refining & Marketing segment tables:

Refining margin per barrel more than doubled; segment EBITDA per barrel rose 3.7x. Note that costs per barrel rose only 7.1% (operating) and 6.5% (distribution) — so essentially every dollar of margin expansion fell to EBITDA.

The 10-Q's own explanation, verbatim:

> "Our Refining & Marketing segment results for the second quarter of 2026 versus the second quarter of 2025 reflect higher realized refining margins supported by stable demand and higher product prices driven by global crude oil supply disruptions as a result of increasing regional conflicts, particularly in the Middle East. Longer term, global demand growth is expected to outpace the net impact of refining capacity additions and rationalizations through the end of the decade. We anticipate these fundamentals, as well as the U.S. refining industry's current structural advantages over the rest of the world, will support a constructive environment for U.S. refiners."

And the risk factors name the specific event: "increased pricing volatility or supply disruptions due to the U.S.-Iran conflict and market reactions thereto" and "regional conflicts such as hostilities in the Middle East and in Ukraine."

The second and third sentences of that management quotation are company framing and are half-weighted per house rule. The first sentence is a statement of fact about the quarter and is taken at face value.

The sensitivity the company publishes

From the 10-Q: "Blended crack spread sensitivity (per $1.00/barrel change) | 1,125" — $1,125 million of annual EBITDA for every dollar of blended crack spread, calculated on a blend of 42% MEH, 40% WTI and 18% ANS crude with Gulf Coast, Mid-Continent and West Coast product pricing.

That single disclosure is the risk score. The year-on-year margin expansion of $18.75 per barrel is therefore worth approximately $21 billion of annualised EBITDA, against a market capitalisation of $91.3 billion and an enterprise value near $125 billion corrected for noncontrolling interests. No balance sheet, buyback or asset base offsets a sensitivity of that magnitude, and this dive does not pretend otherwise.

2. Segments — one of three is doing all of it

From the 10-Q's segment adjusted EBITDA reconciliation:

Read that table and the company divides cleanly in three.

Refining & Marketing produced $8,032M of the $11,704M half-year total — 68.6% — against $2,379M of $5,679M (41.9%) a year earlier. Its EBITDA rose 3.4x. This segment is the entire story and it is the entire risk.

Midstream produced $3,376M against $3,361M — a 0.4% change. This is the ballast: fee-based, contracted, and completely indifferent to the war. It is also where the noncontrolling interest lives, so roughly 36.5% of it belongs to MPLX's public unitholders rather than to MPC shareholders.

Renewable Diesel turned positive for the first time in this series: $296M against MINUS $61M, helped by "[r]ecognition of 2025 clean fuel production tax credits as a result of proposed regulatory guidance issued in February of 2026 which clarified the qualification criteria for 45Z credits" — a $32M item in the reconciliation. A segment whose profitability depends on the qualification criteria of a tax credit is not yet a business, and it is 2.5% of segment EBITDA.

3. The MPLX stake — 40% of the market capitalisation, disclosed in one sentence

From the 10-Q, verbatim:

> "We owned approximately 647 million MPLX common units as of June 30, 2026, with a market value of $36.47 billion based on the June 30, 2026 closing price of $56.33 per common unit. On July 28, 2026, MPLX declared a quarterly cash distribution of $1.0765 per common unit, payable on August 14, 2026 to unitholders of record on August 7, 2026. MPC's portion of this distribution is approximately $697 million."

Four consequences, all material and none of them in the vendor payload.

The stake is 40.0% of MPC's $91.26 billion market capitalisation. At MPLX's current price of $60.51 (from the vendor's own peer block), 647 million units are worth approximately $39.15 billion, 42.9%. A reader valuing MPC on a refining multiple alone is valuing a company of which two-fifths is a separately listed midstream partnership.

MPC receives approximately $697 million per quarter — roughly $2.8 billion a year — in cash distributions that do not depend on the crack spread. Against a $91.3 billion market capitalisation that is 3.1% a year, more than twice the common dividend yield.

MPLX is growing by acquisition and MPC consolidates it. The 10-Q discloses the Northwind Midstream Acquisition: "On August 29, 2025, MPLX completed the acquisition of 100 percent of the outstanding membership interests of Northwind Delaware Holdings LLC... for $2.4 billion in cash... financed with a portion of the net proceeds from MPLX's $4.5 billion senior notes issuance in August 2025." The system has "over 200,000 dedicated acres, more than 200 miles of gathering pipelines" and treating capacity expanding "to over 400 MMcf/d by the second half of 2026" from 150 MMcf/d, "partially supported by minimum volume commitments by regional producers." Purchase-price allocation: $1,167M of property, plant and equipment, $957M of intangibles with a 15-year weighted-average amortisation period, and $360M of goodwill. Also disclosed: the Rockies gathering and processing disposal to Harvest Midstream for $980 million (2025-11-12), and the sale of MPC's 49.9% interest in The Andersons Marathon Holdings ethanol joint venture for $427 million (2025-07-31).

And the consolidation distorts the cash-flow picture in MPC shareholders' disfavour, which Section 5 quantifies. The 10-Q also flags a specific asset risk: MPLX may have to contribute its pro rata share of interest and principal if a pipeline easement vacatur causes a shutdown, and its pro rata share of remediation costs to reinstate the easement.

4. Capital return — and the share count is the real story

From the 10-Q, verbatim: "On May 5, 2026, MPC announced that our board of directors approved a $5.0 billion share repurchase authorization in addition to the $5.0 billion share repurchase authorization announced on November 5, 2024. As of June 30, 2026, we had $6.13 billion remaining under the share repurchase authorizations. The share repurchase authorizations have no expiration date."

The diluted share count, from inc_a and the 10-Q:

A 55.3% reduction in six years. Cumulative repurchase from cf_a: $11,922M (2022), $11,572M (2023), $9,189M (2024), $3,488M (2025) — plus $3,250M in the first half of 2026.

The dividend was raised on 2026-07-29. From the 10-Q's subsequent-events disclosure: "On July 29, 2026, our board of directors declared a dividend of $1.00 per share on common stock. The dividend is payable September 10, 2026 to shareholders of record as of the close of business on August 19, 2026." $4.00 annualised, a 1.28% yield. The vendor's dividendPerShareTTM of $3.91 is trailing and understates the forward rate by 2.3%.

Combined shareholder yield: $6,500M annualised buyback (at the H1 run rate) plus $1,170M of dividend against a $91.26 billion market capitalisation = approximately 8.4%. That is the largest shareholder yield in this batch by a wide margin — and it is funded by earnings that consensus expects to halve.

5. Free cash flow — the field is corrupt, the number is right, and the number is still misleading

This dive was directed to treat capex and free cash flow as guilty until proven innocent. MPC produces the most interesting result in the programme so far: the diagnostic field is corrupt while the underlying figures are correct.

The corrupt field. capexToDepreciationTTM is −5.996. A negative capex-to-depreciation ratio is arithmetically impossible for a company with positive capital expenditure and positive depreciation, and it is the exact signature that flagged COP, DUK, CSX, FCX and SO.

But the underlying figures check out against the filing. The 10-Q's cash flow statement reports additions to property, plant and equipment of $2,099M for the six months ended 2026-06-30 and $1,358M for the prior-year half. Reconstructing the trailing twelve months:

The capex figure is correct. Similarly, D&A: FY2025 $3,353M less H1 2025 $1,582M plus H1 2026 $1,647M = $3,418M TTM, so the true capex/D&A ratio is 1.24x, not −5.996. The ratio field is broken in isolation; the inputs are not. Free cash flow: OCF TTM reconstructed at $17,126M (FY2025 $8,253M − H1 2025 $2,575M + H1 2026 $11,448M) less $4,227M capex = $12,899M, against the vendor's implied $13,095M — a 1.5% difference. freeCashFlowYieldTTM of 14.30% is arithmetically defensible.

And it is still misleading, for three reasons that have nothing to do with data quality.

First, roughly a quarter of the trailing operating cash flow is working capital, not earnings. The H1 2026 cash flow statement shows current receivables −$5,394M, inventories +$156M, current liabilities and other current assets +$7,552M, derivatives +$876M, right-of-use/lease +$6M, all other +$30M — a net +$3,226M contribution. In a quarter when product prices spiked, payables ballooned faster than receivables. That reverses.

Second, roughly 36.5% of MPLX's cash flow is not MPC's. Distributions to noncontrolling interests were $813M in the first half, roughly $1.6 billion annualised, and consolidated free cash flow includes MPLX's cash before that leakage.

Third, and most importantly, the earnings are at a war-driven peak.

We publish 8.8%, not 14.3%, and we show the bridge. 8.8% is still a very high free-cash-flow yield — and it is being earned at $36.33 per barrel of refining margin against $17.58 a year ago.

6. Balance sheet

Three observations.

Leverage is adequate but not trivial: net debt of $30,686M against FY2025 EBITDA of $11,678M is 2.63x, against the vendor's netDebtToEBITDATTM of 1.64x (which uses the far higher trailing EBITDA including the June quarter). Interest coverage is 11.13x and net interest costs were $340M in the quarter against $7,322M of operating income. A substantial part of the debt sits at MPLX, which is separately financed — the Northwind acquisition was funded by MPLX's own $4.5 billion senior notes issuance in August 2025 — so consolidated leverage overstates the recourse position at MPC. The filings in this archive do not split the debt by entity and this dive does not estimate it.

Cash more than doubled in six months, from $3,673M to $7,768M including restricted cash, as $11,448M of operating cash flow outran $2,441M of investing and $4,912M of financing outflows.

Equity fell from $24,404M (2023) to $17,314M (2025) despite cumulative profits — because $24,249M of stock was repurchased across 2023-2025. bookValuePerShareTTM of $88.38 against a $312.61 price is 3.54x on the shareholders'-equity-per-share basis of $65.56, or 4.77x on the vendor's priceToBookRatioTTM; note the vendor's bookValuePerShareTTM ($88.38) and shareholdersEquityPerShareTTM ($65.56) differ by 34.8%, the gap being the noncontrolling interest.

7. Valuation — priced in or room?

At $312.61 (market cap $91.26B, ~291.9M shares; 290M diluted weighted-average in the June quarter):

Estimate coverage and dispersion are the headline here. 10 analysts on FY2026 EPS, 11 on FY2027, only 5 on FY2028. The FY2027 range is $22.33 to $40.04 — a 79% spread, five times wider than any other name in this batch — and FY2028's is 113%. That is not analyst laziness; it is an honest representation of a variable nobody can forecast. The FY2029 row rests on ONE analyst ($23.58) and FY2030 on ONE ($40.10, a 70% jump on a single estimate); both are excluded from every conclusion in this dive.

The shape of the consensus is the single most important thing on this page: $40.00 → $29.21 → $20.36, a 49% decline over two years. Consensus is explicitly modelling the current earnings as a peak. The trailing multiple of 10.6x and the FY2026 multiple of 7.8x are therefore not "cheap" in any sense a screen would recognise — they are the classic refiner configuration in which the multiple is lowest at the top of the cycle.

est.ebitdaAvg and est.ebitAvg are REJECTED — fixed-ratio fabrication. ebitdaAvg is exactly 9.834% of revenueAvg in every year FY2026 through FY2030; ebitAvg is exactly 7.496%. Tested against reality: FY2025 actual EBITDA of $11,678M on $132,699M is 8.80%, so the fabricated ratio overstates the actual margin by 11.8%. More damningly, the block is incapable of representing a cycle at all: a fixed percentage of revenue cannot capture a business whose EBITDA per barrel went from $6.79 to $24.84 while revenue rose 54%. All forward valuation uses epsAvg. Sixteenth consecutive name.

Peer set is directionally reasonable — VLO and PSX are the two direct comparables and both are present, alongside MPLX (MPC's own subsidiary, which is an odd inclusion) and six upstream/midstream/services names. The payload carries no peer estimates, so no multiple comparison is computed and none is asserted. Note that VLO closed at $308.73 and MPC at $312.61, within 1.3% of each other — a coincidence worth recording since both are in this batch.

7a. What today's price assumes (the inversion)

At $312.61 — 7.8x FY2026 consensus, 10.7x FY2027, an ~8.4% shareholder yield — the price embeds:

• EPS of $40.00 in 2026, falling to $29.21 in 2027 and $20.36 in 2028. (Consensus; 10, 11 and 5 analysts.) The first half delivered $19.30, so FY2026 needs $20.70 in the second half, of which the Q3 consensus is $16.42 — leaving roughly $4.28 for Q4, against $5.13 in Q4 2025. Consensus already models a collapse back to near-normal margins by the December quarter. This bullet is nearly arithmetic and is not the contested part.
• Refining & Marketing margin per barrel does not stay at $36.33. (Our derivation from the company's sensitivity.) This is the most fragile assumption in the price, and it is fragile in BOTH directions. At $1,125M of EBITDA per $1.00/bbl, the gap between the current $36.33 and the prior-year $17.58 is roughly $21 billion of annualised EBITDA. If margins hold, FY2027 consensus of $29.21 is far too low; if they revert, FY2028's $20.36 is optimistic. The 79% FY2027 estimate range is the market's honest statement that it does not know.
• The buyback continues at roughly $6.5 billion a year. (Our number, from $3,250M in the first half.) That is 7.1% of the market capitalisation annually and it is the largest single contributor to the expected return. $6.13 billion of authorisation remained at 2026-06-30, so roughly eleven months of capacity at the current pace before a new authorisation is needed.
• The MPLX stake retains its value. (Filing.) 647 million units at $56.33 on 2026-06-30 was $36.47 billion, 40% of MPC's market capitalisation, paying MPC ~$697M a quarter. A 20% fall in MPLX is roughly 8% of MPC's market capitalisation, independent of anything the refineries do.
• The market keeps paying roughly 10-11x the following year's consensus. (Our number.) At 8x FY2027E the stock is $234; at 13x it is $380. Refiners have historically traded 6-10x at peak earnings and 15-25x at trough earnings, which is why the multiple is a poor anchor here and why the fair-value range below is deliberately wide.

7b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (MINUS 27.0%, from FY2026E $40.004 to FY2027E $29.209) + multiple re-rating (EXPANSION of roughly 37%, from 7.8x FY2026E to 10.7x FY2027E at a constant price) + shareholder yield (+8.4%).

Read that carefully, because it is the arithmetic of a peak-cycle stock and it is easily mis-stated. The forward multiple rising from 7.8x to 10.7x as the estimate year rolls forward is NOT re-rating — it is the mechanical consequence of a falling estimate at a constant price, and the guardrail in the format addendum exists precisely for this case. The real question is what multiple the market pays on the then-current year at each date.

Our base of $315 is 10.8x FY2027E, essentially the multiple implied by today's price on next year's number. We assume no re-rating and no de-rating, and we let a 27% earnings decline be offset by an 8.4% shareholder yield and a 7% reduction in the share count. The result is a fair value 0.8% above spot — which is another way of saying that at these estimates and this multiple, the buyback exactly offsets the expected earnings decline.

That is a genuinely interesting equilibrium and it is worth naming: MPC's shareholder yield is large enough to neutralise a 27% consensus earnings decline. It is not large enough to produce a return, and it will shrink with the cash flow that funds it.

7c. Variant perception (where we differ, what would surprise)

• On the level we are 3.4% above the street's $304.56 and 0.8% above spot, so we have no meaningful variant perception — but the street has one with itself and it is worth naming. 25 of 33 analysts rate MPC a Buy, and the consensus price target is BELOW the closing price. Only the median ($321) is above it. That is a ratings distribution that has not been updated for an 88.9% twelve-month move, and it means the "Buy" consensus label carries no information.
• We think the knowledge base's thematic lane is the most decision-relevant material in this file and we are deliberately not underwriting either side of it. Six independent 2026-dated voices argue the capacity destruction is structural — doomberg ("With the Strait of Hormuz shut and refining/storage buffers being rebuilt, energy is higher-for-longer in the coming years, refining margins too", 2026-04-01), mike_green ("Refining capacity is being hit in Russia, Ukraine and the Middle East, so the crack spread... is rising rapidly", 2026-07-24), eurodollar_university ("The record ~$60 3-2-1 crack spread means gasoline and diesel stay expensive despite far lower crude", 2026-07-16), money_of_mine ("crack spreads up 3-4x pre-war levels", 2026-05-22), andreas_steno ("a structural capex story with legs for another 2-3 years; near-term move already in the price", 2026-01-05) and forward_guidance on an export ban (2026-03-20). One argues the opposite with a dated test — forward_guidance, 2026-06-26, speaker Quinn: "once past peak summer driving season the gap should close", test: "crack-spread gap closes after the summer months." Watchable event: the 2026-11-03 print, which covers exactly the post-summer period the bearish test names.
• We differ from the vendor on free cash flow by 38%, and the difference is economic rather than arithmetic. The vendor's 14.30% free-cash-flow yield is defensible arithmetic; our 8.8% strips $3,226M of working-capital and derivative benefit and roughly $1.6 billion a year of distributions to MPLX's public unitholders. A screen sorting on free-cash-flow yield will place MPC near the top of the S&P 500 on a number that includes a payables build and other people's cash.
• Positive surprise that would force a re-rate: Refining & Marketing margin per barrel above $30 in the September quarter, i.e. the war premium persisting past the summer driving season, which would invalidate the one bearish claim in the knowledge base and make the FY2027 consensus of $29.21 visibly too low; or a new repurchase authorisation above $6.13 billion, signalling that management expects the cash flow to continue.
• Negative surprise that would break the thesis: a Middle East ceasefire or a Strait of Hormuz reopening, which is the whole thesis in one headline; Refining & Marketing margin per barrel back below $20; a US refined-product export ban or windfall levy, which the knowledge base records as "still on the table" and which is the specific policy risk to a domestic refiner at record margins; a buyback pause; or an MPLX de-rating, which is 40% of the market capitalisation.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $29.209, low $22.326, high $40.037, 11 analysts), cross-checked against FY2026, FY2028 and the shareholder yield. The range is deliberately the widest in this batch because the underlying estimate range is 79% wide.

• Bear ~$200 — 9.0x the FY2027 consensus MEAN of $29.209, or 8.96x the FY2027 low of $22.326 grossed for a lower multiple; 9.8x FY2028E. Cross-check: 26.1% above the 52-week low of $158.59; 5.0x FY2026E. The scenario: the conflict de-escalates, crack spreads revert toward the historical $15-20 3-2-1 norm that doomberg's own 2022 claim identifies, Refining & Marketing margin falls below $20 per barrel, and the market applies a trough multiple to a normalising earnings stream while the buyback is throttled. −36.0%.
• Base ~$315 — 10.8x the FY2027 consensus MEAN of $29.209. Cross-check: 7.9x FY2026E; 15.5x FY2028E; 10.5x the simple three-year consensus average of $29.86; 1.5% below the 52-week high. The scenario: margins normalise roughly as consensus models, the ~7% annual buyback offsets the earnings decline, MPLX holds its value and keeps distributing ~$2.8 billion a year, and the multiple on the then-current year does not move. +0.8%.
• Bull ~$415 — 10.4x the FY2027 consensus HIGH of $40.037, i.e. the same multiple applied to the top of the estimate range rather than a higher multiple applied to the mean. Cross-check: 10.4x FY2026E; 29.8% above the 52-week high. The scenario: the six-voice structural argument is right — capacity destruction is permanent, Hormuz stays disrupted, no new refineries are built, and FY2027 lands at the top of a very wide range while the share count falls another 7%. +32.7%.

Base is 0.8% above spot; asymmetry 0.91:1 (36.0% down, 32.7% up), plus an 8.4% shareholder yield. A base case inside 1% of spot with a payoff ratio slightly below 1:1 is a Hold, and the very wide range around it is the honest representation of a business whose earnings are set by a war.

8. Knowledge base — no name, and the best theme lane in the batch

Raw hits: 129. After a case-sensitive entity re-run: 19. Name-level claims on Marathon Petroleum: ZERO. Entity claims on MPLX: 1. Used for conviction: 0. Used as thematic context: 18.

The sweep ran entity terms MPC, Marathon, Marathon Petroleum and MPLX, plus free text on refiner, refining, crack spread and diesel, across the full distilled claim store.

The case-sensitive discipline was essential and is worth showing. The raw run returned 129 hits across 34 channels. "MPC" is one of the most collision-prone three-letter strings in a macro-heavy store — it is also the standard abbreviation for the marginal propensity to consume and for model predictive control — and "Marathon" collides with Marathon Digital, the bitcoin miner, and with the footrace as a metaphor. The case-sensitive entity run collapses the lane to 19 and eliminates every one of those.

Not one claim in the store names Marathon Petroleum. The single entity-bearing survivor names MPLX:

> 2023-10-03 · bullish · conviction 80 · horizon: thesis · entities: Energy Transfer, ET, Magellan, MMP, MPLX, NuStar, NS · channel: odd_lots

> "~40% of portfolio in oil/gas pipeline MLPs like Energy Transfer yielding ~9%, tax-deferred until sale — favorite risk-adjusted, tax-advantaged holding."

That is a claim about the MLP asset class from three years ago, and it is used only as background to the MPLX discussion in Section 3.

But the remaining eighteen are the most decision-relevant material this knowledge base has contributed to any name in this batch, and they are all about the one variable that sets MPC's earnings. Six are dated 2026. The 2026 lane runs six bullish to one bearish.

The 2026 bullish lane, verbatim:

> 2026-04-01 · bullish · conviction 55 · horizon: fact · entities: oil · channel: doomberg · speaker: Doomberg · role: independent · skill 1.0

> "With the Strait of Hormuz shut and refining/storage buffers being rebuilt, energy is higher-for-longer in the coming years, refining margins too."

> 2026-07-24 · bullish · conviction 62 · horizon: thesis · entities: gasoline, oil · channel: mike_green · speaker: mike_green · role: independent · skill 0.9

> "Refining capacity is being hit in Russia, Ukraine and the Middle East, so the crack spread — gasoline prices over oil prices — is rising rapidly."

> 2026-07-16 · bullish · conviction 68 · horizon: thesis · entities: gasoline, diesel · channel: eurodollar_university · role: independent · skill 0.7

> "The record ~$60 3-2-1 crack spread means gasoline and diesel stay expensive despite far lower crude; drained inventories can't be rebuilt quickly, so pump pain drags on for some time longer."

> TEST: "crack spreads stay near record and pump prices remain high even as crude stays low, in the coming months"

> 2026-05-22 · bullish · conviction 75 · horizon: thesis · entities: VEA, ALD · channel: money_of_mine · speaker: Manny Datt · role: independent · skill 1.0

> "Rotated evenly into Viva and Ampol refiners; Middle East refinery damage cut global capacity, crack spreads up 3-4x pre-war levels; government lifted minimum refining margins, lowering their WACC."

> 2026-01-05 · bullish · conviction 62 · horizon: thesis · channel: andreas_steno · speaker: Andreas Steno · role: independent · skill 0.8

> "US refiners handling heavy Venezuelan crude are a structural capex story with legs for another 2-3 years; near-term move already in the price."

> 2026-03-20 · bullish · conviction 60 · horizon: fact · channel: forward_guidance · skill 1.0

> "Refined-product export ban still on the table — diesel/crack spreads far outpacing WTI; that, not a crude ban, is where consumer-affordability action lies."

The single bearish 2026 claim, and it is the most useful one in the file because it carries a dated, resolvable test:

> 2026-06-26 · bearish · conviction 55 · horizon: fact · entities: gasoline · channel: forward_guidance · speaker: Quinn

> "Gasoline hasn't followed crude down and crack spreads are surging on depleted product reserves, but once past peak summer driving season the gap should close."

> TEST: "crack-spread gap closes after the summer months"

The pre-2026 lane is a historical calibration and it is instructive. doomberg (2022-06-01): "The 321 crack spread is at ~54-55 vs a historical $15-20 range." doomberg (2022-10-28): "US refiners run at 94-95% capacity with no new major refinery since the 1970s." doomberg (2022-11-04): "The US hasn't built a scale refinery since the 1970s; chronic refining shortfall versus demand structurally supports refined-product prices." warren_pies (2022-05-13): "Refiners rally hard as 3-2-1 crack spreads blew out from ~$10 to $50." darius_dale/Eric Townsend (2022-06-24): "No room for crack spreads or refining margins to come back down." real_vision (2024-07-24, bearish 75): "Skeptical of oil into H2 — late cycle, economically sensitive, weak crack spreads."

Note what that history shows: the same structural argument was made at the 2022 peak, and crack spreads DID come back down — MPC's Refining & Marketing margin per barrel was $15.57 in the first half of 2025. The "no new refinery since the 1970s" argument is true and permanent, and it did not prevent a three-year margin trough. That is the most important calibration in this lane and it is why the structural claims are recorded rather than underwritten.

Concentration test, run and reported. doomberg supplies 6 of the 19 case-sensitive survivors — 31.6%, at the edge of the one-third threshold — and every doomberg claim in the lane is bullish on refining margins. A single channel supplying nearly a third of a lane, all in one direction, is a concentration finding and is discounted accordingly. The 2026 bullish lane without doomberg is four voices, not six.

Conclusion. Breadth 0, name-level claim count 0, net conviction NONE — the conviction score is zero because no claim names the company. The thematic lane enters no score and is the single most useful contribution the knowledge base makes to this dive, principally through the one bearish claim whose test resolves in the exact quarter MPC next reports.

9. Data integrity — what we rejected and why

Nine findings.

1. capexToDepreciationTTM is −5.996 — arithmetically impossible — BUT the underlying capex and free cash flow are CORRECT. This is the first case in this programme where the corruption signature fires without the free-cash-flow number being wrong. Verification against the 10-Q: additions to property, plant and equipment were $2,099M for the six months to 2026-06-30 and $1,358M for the prior-year half; reconstructing TTM capex from cf_a's FY2025 figure gives $4,227M against the vendor's implied $4,241M — a 0.3% match. TTM D&A reconstructs to $3,418M, so the true ratio is 1.24x. The ratio field alone is broken. Free cash flow is computed by subtraction and is right. The lesson recorded for the programme: the diagnostic and the datum can fail independently, and the diagnostic must be checked against the filing rather than trusted to condemn.

2. freeCashFlowYieldTTM of 14.30% is arithmetically defensible and economically misleading — CORRECTED to approximately 8.8%. Three adjustments, all filing-sourced: the H1 2026 cash flow statement shows a +$3,226M net contribution from working capital and derivatives (receivables −$5,394M, inventories +$156M, current liabilities and other current assets +$7,552M, derivatives +$876M, other +$36M); distributions to noncontrolling interests were $813M in the half, roughly $1.6 billion annualised, and belong to MPLX's public unitholders rather than to MPC shareholders; and the earnings are at a war-driven peak. Adjusted MPC-attributable free cash flow is approximately $8,073M, a yield of approximately 8.8%. Both figures are published with the bridge shown.

3. enterpriseValueTTM OMITS noncontrolling interests — the KKR/FCX/SPGI defect, and it is large here. The vendor reports $117,786M, consistent with market cap $91,262M plus roughly $34,000M of debt less the 2026-06-30 cash of $7,768M. It does not add the $6,772M of minority interest that represents the ~36.5% of MPLX owned by the public. Corrected enterprise value is approximately $124,558M — 5.7% higher — and EV/EBITDA moves from 7.281x to approximately 7.68x. Both are stated; the corrected figure is used.

4. seg_geo FY2025 contains a SINGLE line reading "Midstream" $11,534M — REJECTED outright. It is not a geography, it is a segment name; it is the only entry in the block for any year; and $11,534M is 8.7% of FY2025 revenue of $132,699M. The block is unusable and no geographic revenue analysis appears in this dive.

5. seg_prod FY2025 ties and is ACCEPTED. Refining & Marketing $124,252M + Renewable Diesel $2,799M + Midstream $5,628M = $132,679M against $132,699M of revenue, 99.98% — an essentially exact tie. Recorded as a clean result, and noted for contrast with the geography block in the same payload. Note that the segment revenue figures are gross of intersegment eliminations in the company's own presentation and that the dive uses segment adjusted EBITDA from the 10-Q rather than these revenue lines.

6. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature — REJECTED, and here the defect is conceptual as well as arithmetic. ebitdaAvg is exactly 9.834% of revenueAvg and ebitAvg exactly 7.496% in every year FY2026-FY2030. Tested against reality: FY2025 actual EBITDA of $11,678M on $132,699M is 8.80%, so the ratio overstates by 11.8%. More fundamentally, a fixed EBITDA margin cannot describe a company whose segment EBITDA per barrel moved from $6.79 to $24.84 while revenue rose 54% — the block is structurally incapable of representing this business. All forward valuation uses epsAvg.

7. The FY2029 and FY2030 estimate rows rest on ONE analyst each and are EXCLUDED. FY2029 epsAvg is $23.58 (1 analyst) and FY2030 is $40.10 (1 analyst) — a 70% increase between two single-estimate years, which is noise rather than forecast. Neither appears in any conclusion.

8. dividendPerShareTTM of $3.91 is trailing and misses the 2026-07-29 declaration of $1.00 per quarter. The forward rate is $4.00 and the forward yield is 1.28%; both are used.

9. Two receivables fields are zero and cannot be. receivablesTurnoverTTM = 0 and daysOfSalesOutstandingTTM = 0 for a company reporting $10,317M of net receivables at 2025-12-31 and a $5,394M increase in the first half of 2026. operatingCycleTTM (26.75 days) and cashConversionCycleTTM (−24.83 days) are computed from those zeros and are therefore also wrong. Separately, stockBasedCompensationToRevenueTTM is 0 and cf_a reports stockBasedCompensation of 0 in all four years, which is implausible for a company of this size. None of these fields is used.

Corporate actions absent from the payload — THREE, all found in the filings. (a) A NEW $5.0 billion share repurchase authorisation approved 2026-05-05, in addition to the November 2024 authorisation, leaving $6.13 billion remaining at 2026-06-30. (b) The dividend declaration of 2026-07-29 at $1.00 per share, payable 2026-09-10. (c) The death of director Abdulaziz F. Alkhayyal, reported in the 8-K of 2026-06-29; he had served since 2016 on the Compensation and Organization Development Committee and the Sustainability and Public Policy Committee. Also absent, though disclosed in the 10-Q rather than an 8-K in this window: the Northwind Midstream acquisition ($2.4 billion, 2025-08-29), the Rockies disposal ($980 million, 2025-11-12) and the sale of the 49.9% TAMH ethanol interest ($427 million, 2025-07-31).

Vendor versus filing on the quarter — small discrepancies, FILING WINS. inc_q reports Q2 2026 operating income of $7,198M and diluted EPS of $17.68; the 10-Q reports $7,322M and $17.73. earn_cal reports revenue of $52,337M, matching the filing's total revenues and other income, while inc_q reports $51,994M, matching sales and other operating revenues. All figures in this dive use the filed statements and the basis is stated where the two differ.

Quote versus technical blocks. quote.yearHigh/yearLow report $326.92 / $158.00; tech.hi52/lo52 report $319.76 / $158.59. Differences of 2.2% and 0.4%. tech is used throughout. Note that tech.max_dd_from_peak (−2.236%) is identical to tech.pct_from_hi, meaning the deepest drawdown in the measured window is the current one — for a stock up 97.1% from its low, that describes an uninterrupted advance.

Share count — CHECKED AND MATCHED. Market capitalisation ÷ price implies 291.9M shares against the 10-Q's 290M diluted weighted-average for the quarter and 289M basic. A 0.7% match against the diluted average, which is expected given the buyback ran through the period.

Non-equity tripwire — checked and passed. Common stock, par value $0.01, NYSE-listed, per the 8-K cover page. Price $312.61; beta 0.508; volume 2.35M shares (~$734M of turnover); 52-week band $158.59-$319.76, a 102% range — by far the widest in this batch and an accurate representation of what this security is. This is common equity.

10. Technicals, insiders and verdict

Technicals

• Price $312.61, +13.1% above the 50-day moving average of $276.31 and +40.1% above the 200-day average of $223.06. Extended on every measure.
• −2.2% from the 52-week high of $319.76; +97.1% above the low of $158.59. Position in the annual range: 96th percentile — the highest in this batch.
• RSI 60.2 — strong but not overbought, which is itself notable after a 97% move and reflects a steady rather than parabolic advance. MACD +11.21.
• Maximum drawdown from peak in the measured window: −2.2%, identical to the current distance from the high.
• Relative performance: 3-month +23.8% vs SPY +7.6%; 6-month +76.7% vs SPY +11.1%; 12-month +88.9% vs SPY +24.3% and QQQ +30.8%. The best absolute and relative performance in this batch at every horizon.
• Sentiment: 0 strong buy, 25 buy, 8 hold, 0 sell across 33 analysts; consensus label Buy. Consensus target $304.56 — 2.6% BELOW the closing price; median $321 (+2.7%); high $344 (+10.0%); low $225 (−28.0%). A 25-Buy rating distribution with a consensus target below the price is stale, and this dive treats the "Buy" label as carrying no information.

Today's move and what it means. MPC reported before the open, beat consensus EPS by 24.2% and revenue by 28.1%, and closed up 1.82% at $312.61 from $307.03, having opened at $308.83 and traded $301.57 to $315.12 on 2.35M shares. A 24% beat producing a 1.8% move is the single most informative fact in the technical section: the market has already discounted these earnings as non-recurring.

Insiders — every discretionary transaction is a sale

The file contains transactions by two officers. Michael A. Henschen II (Executive Vice President, Refining) exercised 4,964 options at $49.94 on 2026-06-04 and sold 4,964 shares at $268.845 and a further 1,372 at $268.75 the same day, ending at 16,900 shares — an exercise-and-sell with no retention of the exercised stock. Ricky D. Hessling (Chief Commercial Officer) sold on 2026-05-13, ending at 6,525 shares.

There is not one open-market purchase in the file. The Henschen exercise-and-sell was executed at approximately $268.80, 14.0% below today's close, and the option strike of $49.94 is a reminder of where this stock traded when the grant was made. Small amounts, mechanical structure, and a uniformly one-directional pattern: mildly negative, and reported as mildly negative.

Verdict

Hold.

The arithmetic. At $312.61 MPC trades at 7.8x the FY2026 consensus of $40.004 and 10.7x FY2027's $29.209, against a consensus that models a 49% earnings decline over two years. Our base fair value of $315 is 0.8% above spot with 0.91:1 asymmetry ($200 bear, $415 bull) — a payoff ratio slightly below one. The street's consensus price target of $304.56 is below the closing price. And the market's own verdict was delivered this morning: a 24.2% earnings beat moved the stock 1.82%.

What is genuinely excellent. A June quarter with $52,337M of revenue, $7,322M of operating income, $5,138M of net income to MPC and $17.73 of diluted EPS against $3.96. Refining & Marketing margin per barrel of $36.33 against $17.58; segment adjusted EBITDA per barrel of $24.84 against $6.79. $3,250M of stock repurchased in six months with $6.13 billion of authorisation remaining, a share count down 55% since 2020, and a dividend just raised to $1.00 a quarter — a combined shareholder yield near 8.4%. A $36.47 billion MPLX stake paying roughly $2.8 billion a year regardless of the crack spread. Midstream segment EBITDA of $3,376M for the half, essentially unchanged, providing genuine ballast. And the freshest data in this batch: a 10-Q filed the morning of this dive with tables preserved.

What keeps it at Hold rather than anything stronger or weaker.

First, the earnings are a war. The 10-Q attributes the margin expansion to "global crude oil supply disruptions as a result of increasing regional conflicts, particularly in the Middle East", and the risk factors name "the U.S.-Iran conflict." At $1,125 million of EBITDA per dollar of blended crack spread, the $18.75/bbl of margin expansion is roughly $21 billion of annualised EBITDA against a $91.3 billion company. A ceasefire is the single biggest risk and it is not forecastable.

Second, consensus already models the decline and the price already reflects it. $40.00 → $29.21 → $20.36. The 7.8x current-year multiple is not cheapness; it is the standard configuration of a refiner at the top of a cycle.

Third, the asymmetry is very slightly negative. 32.7% up against 36.0% down. The 79%-wide FY2027 estimate range is an honest statement that neither we nor the eleven covering analysts can narrow it.

Fourth, the entry is at the 96th percentile of the annual range after a 97.1% advance from the low, 13.1% above the 50-day average.

Fifth, the knowledge base has zero name-level conviction, its best thematic lane is 32% supplied by a single uniformly bullish channel, and the same structural argument was made at the 2022 peak before margins fell to $15.57 per barrel by the first half of 2025.

Against all of that, the reason this is a Hold and not an Avoid: the shareholder yield of ~8.4% is large enough to neutralise the entire consensus earnings decline, the balance sheet is sound at 11.13x interest coverage, the MPLX stake is 40% of the market capitalisation and does not depend on the crack spread, and six independent voices argue the capacity destruction is permanent. We do not know who is right and we decline to pretend.

Pre-registered KILL criteria — what would take this to Avoid:

• Refining & Marketing margin per barrel below $20 in any quarter, i.e. back toward the H1 2025 level of $15.57.
• A Middle East ceasefire or a Strait of Hormuz reopening, which is the thesis in one headline and which the knowledge base's own bullish lane names as the operative condition.
• A US refined-product export ban, windfall levy or refinery-specific tax, recorded by forward_guidance (2026-03-20) as "still on the table."
• The buyback pausing or falling below roughly $1 billion a quarter, which removes most of the shareholder yield that currently offsets the earnings decline.
• An MPLX de-rating of 20% or more, worth roughly 8% of MPC's market capitalisation independent of refining.
• Renewable Diesel returning to losses if the 45Z clean fuel production credit qualification changes.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A price below roughly $240, which is 8.2x FY2027E consensus and restores asymmetry to better than 2:1 against the same base case. This is the operative trigger and it is a price.
• Refining & Marketing margin per barrel above $30 in the September quarter, which would falsify the one bearish knowledge-base claim on its own stated test and make the FY2027 consensus of $29.21 visibly too low.
• A new repurchase authorisation materially above the $6.13 billion remaining, signalling management's own view that the cash flow persists.
• Any structural simplification of the MPLX relationship — a roll-up, a monetisation or a clean sum-of-the-parts disclosure — which would let the market value 40% of the company on its own merits.
• Insider buying, of which there is currently none.

Where MPC fits in the Synthos Framework Portfolio. No position. This is a well-run company that is currently a leveraged, unhedged position on a geopolitical conflict, and the price already reflects the good version of that. The disciplined action is a limit at roughly $240 and patience; the undisciplined action is to buy 7.8x current-year earnings at the 96th percentile of the annual range four months into a war premium. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $312.61, with the fair-value anchors, kill criteria and the $240 trigger all gradeable. We note explicitly that for an existing holder this is a trimming decision rather than a selling one, because the shareholder yield is real and the structural argument is not frivolous.

Single biggest risk: a ceasefire. MPC earned $36.33 of refining margin per barrel in the June quarter against $17.58 a year earlier because, in the 10-Q's own words, of "global crude oil supply disruptions as a result of increasing regional conflicts, particularly in the Middle East." The company publishes the sensitivity: $1,125 million of annual EBITDA per $1.00 per barrel. A return to last year's margin is roughly $21 billion of annualised EBITDA on a $91.3 billion company. The knowledge base's structural voices say the capacity is physically destroyed and will not return; the same argument was made in 2022 and margins nonetheless fell to $15.57 per barrel by the first half of 2025. We are not taking a view on a war, and a company whose earnings are set by one should not be owned as though we were.

Provenance & disclosures

• Traceability: ZERO knowledge-base claims name Marathon Petroleum (raw hits 129, case-sensitive survivors 19, name-level entity claims 0, MPLX entity claims 1, used for conviction 0; breadth 0, claim count 0, net conviction NONE). The case-sensitive discipline was essential: "MPC" collides with the marginal propensity to consume and with model predictive control in a macro-heavy store, and "Marathon" collides with Marathon Digital and with the footrace as metaphor; the case-sensitive run removed all 110 such hits. The one entity-bearing survivor is odd_lots 2023-10-03 (bullish 80, entities Energy Transfer/ET/Magellan/MMP/MPLX/NuStar/NS) on holding ~40% of a portfolio in pipeline MLPs — used only as background. The remaining eighteen constitute a dense, dated thematic lane on the crack spread, the single variable that sets this company's earnings, and it enters no conviction score while being the most decision-relevant material in the file. The 2026 lane runs six bullish to one bearish: doomberg 2026-04-01 ("With the Strait of Hormuz shut... refining margins too"), mike_green 2026-07-24 ("Refining capacity is being hit in Russia, Ukraine and the Middle East"), eurodollar_university 2026-07-16 ("the record ~$60 3-2-1 crack spread", with a stated test), money_of_mine/Manny Datt 2026-05-22 ("crack spreads up 3-4x pre-war levels"), andreas_steno 2026-01-05 ("structural capex story with legs for another 2-3 years; near-term move already in the price") and forward_guidance 2026-03-20 ("Refined-product export ban still on the table") against forward_guidance 2026-06-26, speaker Quinn ("once past peak summer driving season the gap should close", TEST: "crack-spread gap closes after the summer months") — a dated, falsifiable claim that resolves in the exact quarter MPC reports on 2026-11-03. Concentration test RUN and REPORTED: doomberg supplies 6 of 19 case-sensitive survivors (31.6%) and every one is bullish on refining margins; the lane is discounted accordingly and the 2026 bullish count without doomberg is four, not six. Historical calibration recorded: the identical structural argument ("no scale US refinery since the 1970s", "no room for crack spreads to come back down") was made at the 2022 peak by doomberg, warren_pies and darius_dale, and MPC's Refining & Marketing margin per barrel nonetheless fell to $15.57 in the first half of 2025.
• Data as-of: fundamentals — the June 2026 quarter, filed the morning of this dive: the 10-Q filed 2026-08-04 supplies the income statement, cash flow statement, segment adjusted EBITDA, Refining & Marketing per-barrel metrics, the crack-spread sensitivity, the MPLX unit count and market value, the repurchase authorisations, the dividend declaration and the acquisition and disposal disclosures; annual statements through 2025-12-31 from the 10-K filed 2026-02-26 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873603 = 2026-08-04T20:00:03Z ($312.61, +1.82%; 50-DMA $276.31; 200-DMA $223.06; RSI 60.2; MACD +11.21) · knowledge-base claims 2026-08-04. The fiscal year is the calendar year. All figures come from the Synthos vendor data file for MPC or from the SEC filings in the MPC archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-26 (fiscal 2025); 10-Q filed 2026-05-05 (March 2026 quarter); 10-Q filed 2026-08-04 (June 2026 quarter — the primary source for this dive); 8-K filed 2026-05-05 (Item 2.02, first-quarter results; exhibit not extracted); 8-K filed 2026-06-29 (Item 8.01, the death of director Abdulaziz F. Alkhayyal on 2026-06-26); 8-K filed 2026-08-04 (Item 2.02, second-quarter results, signed by CFO Maria A. Khoury; Exhibit 99.1 not extracted). All six documents carry preserved [TABLE] statement data, which is why every quarterly figure in this dive is filing-verified rather than vendor-derived. The two earnings 8-Ks carry only the cover and item text; the press-release exhibits were not extracted, so all quarterly detail comes from the 10-Q itself.
• Where the filing CONTRADICTED the vendor and the filing won: inc_q reports Q2 2026 operating income of $7,198M and diluted EPS of $17.68 against the 10-Q's $7,322M and $17.73; inc_q revenue of $51,994M is sales and other operating revenues while earn_cal's $52,337M is total revenues and other income — both are correct on their own basis and this dive states which basis it uses at each point. dividendPerShareTTM of $3.91 is trailing and misses the $1.00 per share declared 2026-07-29 ($4.00 annualised, 1.28% forward yield). quote.yearHigh/yearLow ($326.92/$158.00) differ from tech.hi52/lo52 ($319.76/$158.59) by 2.2% and 0.4%; tech is used.
• Where the vendor was REJECTED: capexToDepreciationTTM of −5.996, arithmetically impossible — though uniquely in this programme the underlying capex is CORRECT (reconstructed TTM $4,227M against the vendor's implied $4,241M, a 0.3% match, verified against the 10-Q's $2,099M and $1,358M half-year figures; true ratio 1.24x). freeCashFlowYieldTTM of 14.30%, arithmetically defensible but economically misleading — corrected to approximately 8.8% after removing $3,226M of H1 working-capital and derivative benefit and roughly $1.6 billion a year of distributions to MPLX's public unitholders, with the full bridge published. enterpriseValueTTM of $117,786M, which omits $6,772M of noncontrolling interests (7.4% of market capitalisation) — corrected to approximately $124,558M, taking EV/EBITDA from 7.281x to approximately 7.68x; the KKR/FCX/SPGI defect class. seg_geo FY2025, a single line reading "Midstream" $11,534M — not a geography, and 8.7% of revenue; unusable, and no geographic analysis appears in this dive. est.ebitdaAvg/ebitAvg, exactly 9.834% and 7.496% of revenueAvg in every forward year against an FY2025 actual EBITDA margin of 8.80% — and structurally incapable of describing a business whose segment EBITDA per barrel moved from $6.79 to $24.84; all forward valuation uses epsAvg. The FY2029 ($23.58) and FY2030 ($40.10) EPS rows, each resting on ONE analyst with a 70% jump between them — excluded from every conclusion. receivablesTurnoverTTM and daysOfSalesOutstandingTTM, both ZERO on a company with $10,317M of net receivables, which also invalidates operatingCycleTTM and cashConversionCycleTTM; and stockBasedCompensationToRevenueTTM of 0 with cf_a stockBasedCompensation of 0 in all four years, implausible at this scale. None of these fields is used.
• Where the vendor was CLEAN and it is recorded: seg_prod FY2025 ties at 99.98% ($132,679M against $132,699M) — a clean result on a defect class that failed on MDT, NEM, JCI, CDNS, CSX, UPS and, in this same batch, MRSH; and the share count, 291.9M implied against the 10-Q's 290M diluted weighted-average, a 0.7% match consistent with a buyback running through the period.
• Corporate actions absent from the payload — THREE in the 8-K window: a new $5.0 billion share repurchase authorisation approved 2026-05-05 (leaving $6.13 billion of combined authorisation at 2026-06-30); the $1.00 per share dividend declared 2026-07-29, payable 2026-09-10; and the death of director Abdulaziz F. Alkhayyal on 2026-06-26. Also absent from the payload though disclosed in the 10-Q: the Northwind Midstream acquisition ($2.4 billion in cash, 2025-08-29, funded from MPLX's $4.5 billion August 2025 senior notes issuance, producing $1,167M of property, plant and equipment, $957M of 15-year-amortising intangibles and $360M of goodwill); the Rockies gathering and processing disposal to Harvest Midstream ($980 million, 2025-11-12); and the sale of MPC's 49.9% interest in The Andersons Marathon Holdings ($427 million, 2025-07-31).
• Estimate coverage: 10 analysts on FY2026 EPS, 11 on FY2027 — the anchor for all three fair values, with a range of $22.326 to $40.037, a 79% spread that is five times the widest of any other name in this batch; only 5 on FY2028 (range 113% wide); 1 each on FY2029 and FY2030, excluded. The dispersion is not a data defect — it is the honest state of knowledge about a crack spread.
• Peer note: the vendor peer set — E, EOG, EPD, EQNR, ET, KMI, MPLX, PSX, SLB, VLO — contains the two direct comparables (Phillips 66, Valero) and, oddly, MPC's own consolidated subsidiary MPLX. The payload carries no peer earnings estimates, so no multiple comparison is computed and none is asserted. Recorded for interest: VLO closed at $308.73 and MPC at $312.61, within 1.3% of each other, and both are in this batch.
• Fair-value caveat: the $200 / $315 / $415 anchors are multiples of the FY2027 consensus EPS distribution — 6.8x the mean at the bear (equivalently a trough multiple on a normalising number), 10.8x the mean at the base, and 10.4x the high of $40.037 at the bull — cross-checked against FY2026 and FY2028 and against the 52-week range. Stated arithmetic, not a discounted cash flow. The range is deliberately the widest in this batch because the underlying FY2027 estimate range is 79% wide; a narrower range here would be false precision about a variable set by a geopolitical conflict. The base assumes no re-rating: it applies today's implied multiple on next year's number, so that a 27% consensus earnings decline is offset by an approximately 8.4% shareholder yield and a roughly 7% annual reduction in the share count. The rolldown from 7.8x FY2026E to 10.7x FY2027E at a constant price is MECHANICAL and is not described anywhere in this dive as re-rating.
• Basis note: consensus estimates and earn_cal epsActual are on an adjusted basis; for the June 2026 quarter the reported $17.73 matches the 10-Q's GAAP diluted EPS exactly, so no adjustment gap exists in the most recent period. Segment adjusted EBITDA as used throughout is the company's own measure, defined in the 10-Q as excluding depreciation and amortisation, net interest and other financial costs, turnaround expenses and other adjustments. The trailing multiple of 10.6x is computed on the vendor's trailing per-share figure.
• Timing: second-quarter 2026 results were released on the MORNING of 2026-08-04, the day this dive is struck, and beat consensus EPS by 24.2% ($17.73 against $14.27) on revenue 28.1% above estimate — and the shares closed up 1.82%. The next print is 2026-11-03, 91 days away, against consensus of $16.42 in EPS and $40,709M of revenue. The most recent insider transaction is dated 2026-06-04 and is an exercise-and-sell; there is no purchase in the file.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.