US Upstream Natural Gas Producers Compared (2026)

Five US upstream natural gas producers compared, as of 9 September 2026. A point-in-time snapshot, not an evergreen guide. Price deck: base US$3.00/MMBtu Henry Hub — the fixed-grid price nearest the representative trailing average (3-month US$2.94, 6-month US$2.93) — with every price on the fixed US$2.00–4.00 grid run as a scenario and the EIA’s 2027 forecast as a 0%-weight cross-check; no spot deck is carried anywhere. Currency: US dollars throughout — all five companies report in USD, so no FX conversion enters this post. Fiscal basis: each company’s fiscal-2025 Form 10-K plus its latest quarterly results; market data at the 1–4 September 2026 closes. Units: gas volumes are quoted in bcfe (billion cubic feet equivalent, liquids converted at 6 mcf per barrel), reserves in Tcfe (trillion), and unit economics per mcfe (thousand cubic feet equivalent); the gas price is US$ per MMBtu. Ratings (recomputed on one weighting, Section 4), in the post’s fixed order: EQT 4.3/5 · Expand 4.2/5 · Antero 3.7/5 · Range 4.5/5 · CNX 3.7/5. Value reads: all five Overvalued (wide band) at the shared deck. Timing spread: the five underlying analyses were struck between 4 and 6 September 2026; EQT’s market data is at the 1 September close and the other four at 4 September, and EQT’s own analysis is a template version behind the other four. Refreshed when the underlying analyses are refreshed. For information only, prepared with AI assistance — see the disclaimer at the end.

Five US producers pull most of the Lower 48’s gas out of two sedimentary basins, and they have made five different bets about how to get paid for it. EQT owns the pipe it ships through; Expand holds under ten years of reserves and must drill continuously to stand still; Antero earns the basin’s highest price and hands more than half of it to gatherers and pipelines; Range holds twenty-two years of reserves and no hurry; CNX has sold this year’s gas a quarter below the base deck. Put all five on one construction and the headline finding is not about any one of them: at the shared US$3.00 deck every one of the five reads Overvalued, so the choice between them is not which is cheap — none is — but which gas price you are willing to underwrite. Those prices run from US$3.80 for Expand to US$4.56 for CNX — a 76-cent spread, three-quarters of the distance from the base deck to the top of the grid. To screen these five and every other North American upstream name on the same fields, go to Metal Pilot.

1. The peer group

The inclusion rule: US-listed, US-domiciled upstream companies whose production is predominantly natural gas, above roughly 150 bcfe a year, and which already carry a published single-company analysis on this blog. That admits the five large Appalachian and Haynesville names below. It deliberately excludes Tourmaline and Birchcliff (Canadian-domiciled — they price against the AECO discount and report on the NI 51-101 reserve standard, so they belong in a North American, not a US, comparison), Coterra and the Southwestern-successor volumes inside Expand (no separate listing), Comstock and Gulfport (no published analysis yet — the candidates for the next edition), and Antero Midstream and Williams (midstream, not upstream). Every column below is one published analysis; this post adds no primary research of its own. For the same universe ranked on one cost metric rather than compared across nine dimensions, see Lowest-Cost US Gas Producers (2026) .

Every basis difference in the comparison — the reserve-report price bases, the five different cash-cost constructions, the leverage definitions, the hedge treatments — is consolidated in the comparability ledger, Table 10 in Section 6.1. Read it before trusting any single row.

Table 1. Headline figures comparison, five US gas producers

Source: each company’s FY2025 Form 10-K and latest quarterly results, as analysed in the five posts linked in Section 6.1; market data per stockanalysis.com at the 1–4 Sep 2026 closes, per company as dated in the price row. Columns are ordered by market capitalisation, descending — and that one order is used everywhere in this post: every table, every grid, every bar figure, the quality-against-value plot and the ticker list beside the post. No figure re-sorts itself by its own metric; where a ranking is the finding it is named in the prose and marked on the leading bar in place. One construction governs the operating-cost block: realized price including derivatives and after royalty, plus any revenue the company earns moving third-party volumes where it discloses one, less every cash cost line including corporate general and administrative expense, then less depletion — interest excluded for all five as a financing rather than operating cost. Only EQT has such a revenue line (US$0.26/mcfe of third-party pipeline revenue, ledger row 5); for the other four the term is nil, and on gas sales alone EQT’s cash margin is US$2.18. Per-dollar rows are dated by the prices above, since each divides by a market capitalisation struck at that price; they use the group’s canonical units, bcfe per year for flow and bcfe for stock. All five disclose SEC proved reserves at constant trailing-average prices, so the reserve rows are on one standard; the reserve-report price bases behind them differ and are ledger rows 1 and 2, Table 10. The drilling-inventory row is not rankable — three of the five disclose it in a different form and two not at all, so it prints each disclosure as stated (ledger row 12). Gaps print n/d and are never imputed. Every basis difference is in the comparability ledger, Table 10, Section 6.1.

Three structural facts recur through the rest of the post. The market-cap order is not the production order — Expand produces 10% more gas than EQT and is worth a third less. The size spread is 6.3 to 1: EQT’s market capitalisation is more than six times CNX’s, while it produces only 3.8 times CNX’s volume, wide enough that scale becomes a scorecard dimension in its own right. And the per-dollar rows invert the size ranking — the largest company buys the least production and the fewest reserves per dollar of market value, while the leaders on the two rows are the second-largest and the second-smallest, by a factor of 1.7 on flow and 2.3 on stock.

2. Operating and financial position

2.1 Company by company

EQT is the largest company here by market value and the second largest by volume: 2,382 bcfe in FY2025 against guidance of 2,375–2,450 bcfe, or roughly +1% — the flattest profile in the group. It holds 28.0 Tcfe of SEC proved reserves, the most in the set, on an 11.8-year life, entirely in Appalachia and only ~6% liquids, with more than 30 years of gross drilling locations behind that book. Its cash margin of US$2.44/mcfe is the best figure in Table 1, and its US$1.06/mcfe of cash costs is the lowest of the four companies that publish a full cost stack — CNX’s US$1.04 is a field cash cost on a narrower basis (ledger row 6). The structural fact that separates it: it is the only company here whose margin includes revenue earned from moving other people’s gas — US$0.26/mcfe of third-party pipeline revenue, a consequence of owning its gathering and transmission rather than renting it, and the reason its per-dollar rows are the weakest in the group at 68 bcfe/yr and 807 bcfe per US$1 bn.

Expand Energy is the volume leader at 2,622 bcfe, guiding 2,701–2,774 bcfe for about +4%, on 25.9 Tcfe of proved reserves across Appalachia and the Haynesville. Its cash margin of US$2.01/mcfe is second only to EQT’s and its total cash cost of US$1.29 second-lowest of the four companies here that publish a full cost stack. Its structural fact is duration, or the lack of it: a 9.9-year reserve life, the shortest in the group by more than five years, with the inventory beyond that book undisclosed. Producing at a 2,738 bcfe guidance midpoint against a 25.9 Tcfe base makes the drilling programme a maintenance requirement rather than a growth choice — and it is the only company here whose booked life is shorter than the decade its LNG-adjacent thesis is written over.

Antero Resources produced 1,256 bcfe and guides 1,497 bcfe — a +19% step-up, and the growth is bought, not drilled: the HG Energy acquisition closed in February 2026. It is the most liquids-rich name here at ~36% of volume, which earns it the highest realized price in the table at US$3.97/mcfe, and it holds 19.1 Tcfe on a 15.2-year life. Its structural fact is that the premium price does not survive the pipe: total cash costs of US$2.70/mcfe — more than twice EQT’s — leave a cash margin of US$1.27, the lowest of the five.

Range Resources is the second-smallest here by volume at 816 bcfe, ahead only of CNX, guiding about 867 bcfe for +6%, at ~31% liquids and entirely inside Pennsylvania. Its structural fact is duration: 22.2 years of proved reserves against production, the longest in the group by seven years, on 18.1 Tcfe. It converts that into the second-best fully-loaded margin at US$1.26/mcfe despite a mid-table cash margin, because its depletion charge of US$0.45/mcfe is the lowest here by a third — it grew by drilling rather than by buying, so nothing stepped up its carrying basis.

CNX Resources produced 629 bcfe and is the only company in the group guiding volumes down — 605–620 bcfe, about −3%. It holds 9.66 Tcfe on a 15.4-year life in Pennsylvania and West Virginia, at ~8% liquids, and earns the lowest realized price in the table at US$2.75/mcfe while running the lowest cash cost at US$1.04. Its structural fact is that the shrinkage is deliberate: it is converting a long-life asset into buybacks rather than into barrels, which is why it buys the second-most production and the second-most reserves per dollar of market cap of any name here.

2.2 Production

Figure 1. Production, absolute and per dollar of market value

Source: Table 1. Bars are in the post’s fixed market-cap-descending order, not sorted by either series; the leader marker sits on the leader of series A. The two series carry different units and each is scaled to its own maximum, so every bar prints its true value. Per-dollar figures are dated by the 1–4 Sep 2026 prices in Table 1.

For the first time in this series the flip does not happen: Expand leads on both rows. It is the largest producer at 2,622 bcfe and also buys the most production per dollar of market value at 116 bcfe/yr per US$1 bn — the only company in the group to lead a metric absolutely and per dollar at once, and a direct consequence of being worth a third less than EQT while producing 10% more gas. CNX is a close second per dollar at 113 from a quarter of Expand’s volume, and EQT is last per dollar at 68, buying 41% less production per dollar than Expand does. Antero sits third at 104 and Range fourth at 83.

2.3 Proved reserves

Figure 2. Proved reserves, absolute and per dollar of market value

Source: Table 1. Bars are in the post’s fixed market-cap-descending order; the leader marker sits on the leader of series A. Each series is scaled to its own maximum and every bar prints its true value. All five figures are SEC proved reserves at 31 December 2025, on one standard, so the absolute row is on one standard; the per-dollar row is dated by the 1–4 Sep 2026 prices in Table 1.

Here the flip is complete, and it is the widest in the post. EQT holds the most reserves in absolute terms at 28.0 Tcfe and buys the fewest per dollar at 807 bcfe per US$1 bn; Range holds the second-smallest book at 18.1 Tcfe and buys the most per dollar at 1,849 — 2.3 times what EQT’s price buys. CNX is second per dollar at 1,742 and Antero third at 1,580. The ordering of the two rows is almost exactly reversed, for the reason Section 3 makes explicit: the market is not paying for reserves in this group.

2.4 Operating costs

The construction is one line for all five: realized price including derivatives and after royalty, plus third-party midstream revenue where a company discloses one, less every cash cost including corporate overhead, gives the cash margin; less depletion gives the fully-loaded margin. Only EQT carries the midstream term, at US$0.26/mcfe. Interest is excluded from all five as a financing rather than an operating cost.

Figure 3. Cash margin per unit

Source: Table 1, whose operating-cost block carries the realized price and the cash cost behind every bar. Bars are in the post’s fixed market-cap-descending order, not sorted by margin. Every company is included on one construction, but two carry basis caveats stated in Table 1 and ledgered: EQT’s margin includes US$0.26/mcfe of third-party pipeline revenue no peer earns at scale (ledger row 5), and CNX’s cash cost is its own published field cash cost including selling, general and administrative expense (ledger row 6). The fully-loaded margin is not plotted because CNX publishes depletion only by segment, so the group figure does not exist for one of the five (ledger row 7); Table 1 carries the four that do.

The cost spread is 2.6 to 1 and it is almost entirely about the pipe, not the rock. All five lift gas cheaply — CNX for US$0.15/mcfe, Antero for US$0.11 — and what separates them happens between the wellhead and the market. Antero pays US$2.27/mcfe in gathering, processing and transportation, more than four times CNX’s US$0.54, and that one line turns the group’s highest realized price into its lowest cash margin. EQT sits at the other end for the same reason inverted: it owns the gathering and transmission its peers rent, so the toll is revenue rather than cost.

Range and CNX arrive at the same cash margin from opposite directions. Both print US$1.71/mcfe — Range on a US$3.60 realization against US$1.89 of cost, CNX on US$2.75 against US$1.04. One is liquids-rich with a heavy transport stack, the other the cheapest dry-gas operator selling into the weakest realization. The margin is identical and the businesses are not.

2.5 The three metrics side by side

Figure 4. Flow, stock and margin per company

Source: Table 1. Rows are in the post’s fixed market-cap-descending order, as in every other artifact here. Shading is ranked within each column, never across the grid, because the three columns carry different units: a level 9 means the highest value in that column only. Per-dollar columns are dated by the 1–4 Sep 2026 prices in Table 1. No cell is undisclosed.

The five sort into two structural buckets. There are two integrated seniors — EQT and Expand — that between them produce more than the other three combined, own or part-own their midstream, and are priced accordingly. And there are three Appalachian pure-play producers — Antero, Range and CNX — separated less by geology than by how they are connected to the pipe and how much debt sits between the asset and the equity. The boundary that matters is not size but how each company holds its midstream: EQT and Expand consolidate theirs, Antero holds a 29% stake in a listed operator and pays that operator the group’s highest toll, and Range and CNX own their gathering outright without a separate vehicle.

Three different companies lead the three columns, and that is the finding. Expand leads flow per dollar at 116 bcfe/yr, Range leads stock per dollar at 1,849 bcfe, EQT leads cash margin at US$2.44 — and no company leads more than one. EQT is first on margin and last on both per-dollar columns, the only name here to finish first and last in the same figure. Antero leads none of the three and sits third, third and last; CNX places second on both per-dollar columns without leading either.

The spread across the group is 6.3 to 1 on market capitalisation and 4.2 to 1 on production, but only 1.9 to 1 on cash margin and 2.3 to 1 on reserves per dollar. Duration runs from Expand’s 9.9 years to Range’s 22.2. Being the metric leader here means leading that metric; it is not a verdict, and Sections 3 to 5 are where the verdict gets made.

2.6 Balance sheets and capital returns

Table 2. Balance sheet, credit standing and capital returns

Source: each company’s FY2025 Form 10-K, latest quarterly results and guidance-year free-cash-flow bridge, as published in the five analyses linked in Section 6.1. Net debt and enterprise value are in Table 1 and are not repeated here. Leverage ratios are on each company’s own reported basis and are not strictly comparable — GAAP debt/EBITDA for EQT and Range, net debt to adjusted EBITDAX for Expand and CNX — and Antero publishes no current ratio at all, only its mid-2026 target, so its cell prints n/d and it is excluded from Figure 5 (ledger row 8). Free cash flow is each analysis’s own guidance-year figure struck at the US$3.00 base deck, not the trailing FY2025 number, so the yields are on one construction. The share-count row is reported, not ranked: only EQT and CNX publish a latest-year percentage, and Antero’s and Range’s disclosures are cumulative reductions on different windows — so no cross-company percentage exists and none is imputed (ledger row 9). The three remaining n/d cells each name what would fill them: EQT’s capital-expenditure ratio needs a FY2025 capital figure, which its five-year table does not carry alongside the operating-cash-flow line; Expand’s and Antero’s return on invested capital needs the metric itself, which neither analysis publishes. For how these metrics behave through a commodity cycle, see The Commodity Investor, Part 10: Financial Metrics .

Expand and Range carry the group’s two strong balance sheets, from opposite ends of the size range. Expand holds the lowest leverage ratio at about 0.5× after shedding US$1.3 billion of debt in six months; Range carries the smallest absolute load at US$0.88 billion on 0.73× and cut net debt 28% in the same period. At the other end CNX runs about 1.9× against a US$5.55 billion market capitalisation — the tightest coverage in the group, and the reason it screens as one of the two most price-sensitive names in Section 3.2. Read at the low end of Section 3’s ladder rather than at the base, CNX’s own analysis takes its forward EBITDA to roughly US$1.1 billion at the base deck and its free cash flow to about US$70 million one grid step below it, which is where the buyback the equity is owned for gets squeezed.

The free-cash-flow yields reorder the group. At the base deck CNX returns 5.0% of its price as free cash flow and Expand 4.4%, while EQT returns 1.5% and Antero 0.8% — a US$2.65–2.85 billion capital programme in one case, US$1.2 billion spent to lift volumes 19% in the other. Both dividends are still covered by that line: EQT paid US$390 million in FY2025 against US$515 million, Expand US$765 million against US$1,005 million.

One row has no peer anywhere else in the group. EQT carries a US$3.59 billion Blackstone Class B claim ahead of the common — not debt, not equity, and senior to the shareholder in the midstream vehicle Blackstone bought 40% of in December 2024. At 10% of EQT’s market capitalisation it is larger than any other company’s entire net debt load, and it is deducted in EQT’s own net asset value before a share of value reaches the common. Range’s US$278 million retained divestiture obligation is the only other claim of this kind in the set, and it is a thirteenth the size.

Figure 5. Leverage

Source: Table 2. Bars are in the post’s fixed market-cap-descending order, not sorted by leverage — lower is better, so the leader marker sits on Expand in second place. Antero is excluded because it publishes no current leverage ratio, only a mid-2026 target; the remaining four keep their relative fixed order rather than re-sorting (Table 2, ledger row 8). Each ratio is on the filer’s own definition and the definitions differ — see ledger row 8 in Table 10; the bar lengths are indicative of order rather than of a strictly common measure.

2.7 Hedging and price-risk exposure

Table 3. Price-risk position entering 2026

Source: each company’s FY2025 Form 10-K derivative disclosures as published in its own analysis — EQT , Expand , Antero , Range and CNX — at or near 31 December 2025 except EQT’s (11 February 2026) and CNX’s (8 January 2026). Coverage is stated as each filer defines it — quarterly option volumes, annual notional, a daily rate, a percentage of production — and is deliberately neither converted nor ranked (ledger row 10): the four disclosures are not on one tenor or one denominator, and a coverage percentage built across them would be an imputed figure rather than a comparison. Figure 6 is the comparable read, because it uses the one hedge number every analysis publishes on one construction: the book’s mark at each price on the shared grid.

Figure 6. Hedge book mark-to-market per share, across the price deck

Source: each underlying analysis’s own net-asset-value bridge — EQT , Expand , Antero , Range and CNX — which re-marks the hedge book at every price on this grid from the filer’s disclosed volumes, strikes and tenor; divided by that company’s fully-diluted share count as published in the same section (EQT 625.52 m, Expand 231.50 m, Antero 307.44 m, Range 233.67 m, CNX 147.94 m). The columns are the same fixed natural-gas grid as Figure 8, so a cell here can be read straight across against that figure’s net asset value. Rows are in the post’s fixed market-cap-descending order; the base-case column is outlined. Shading is ranked within each row on that company’s own range across the five columns — so, unlike Figure 8, each row reads saturated at the low deck, because that is where a hedge book is worth most. Expand’s row is held flat at +US$1.33 in every column and is shaded at one level throughout: its derivative note publishes notional volumes and fair values but no strikes, so its own analysis could not re-mark the book by column and carries it at the year-end figure instead (ledger row 11). Marks are after tax where the underlying analysis strikes them after tax.

The two ends of this figure are the most quantifiable difference in the group. CNX swings US$8.02 a share across the grid — from +US$4.69 at US$2.00 to −US$3.33 at US$4.00 — against a net asset value of US$14.92. More than half its downside protection and all of its upside cap sit in a single contract book. EQT moves 24 cents across the whole grid, because its options thin to almost nothing after Q1 2027 and it has chosen to keep its upside. Between them, Antero’s book is worth +US$1.11 at the low deck falling to roughly zero at the high one — protection bought at a strike above the base deck — and Range’s is a third of that, the deliberate posture of the company with the longest reserve life and the least reason to fear a weak year.

Read against Figure 8, the hedge book changes who is protected, not who is cheap. CNX’s equity net asset value at US$2.00 is US$1.32 a share against a US$4.69 hedge mark — at that price the reserves are worth less than nothing to the equity once the debt is deducted. The hedges are a drag at the base deck and the entire equity at the bottom of the grid.

Jurisdiction barely discriminates here: all five operate entirely in stable, rule-of-law US states — Appalachia for four, plus the Louisiana and Texas Haynesville for Expand. Concentration is highest at single-state Range and lowest at Expand, and there is no cross-border or currency exposure anywhere in the set.

3. Asset value

3.1 What the market pays

The primary yardstick here is price to net asset value per share, because it is the one measure every company in the group has on a comparable construction: each underlying analysis builds an equity net asset value from that company’s own disclosed after-tax standardized measure — the filed present value of its proved reserves after tax, moves it to the shared US$3.00 deck on a term built from the filing’s own volumes and tax lines, bridges it through net debt and the other claims, and states the price it is measured against. Every ratio below is struck at the same base deck — the base column of Figure 8 — so the five are read on one price assumption. What each net asset value is built from is not uniform, and the basis column is where that shows. None of these is a company-published valuation. For the methods behind them, see The Commodity Investor, Part 11: How to Value Commodity Stocks .

Table 4. Value against the yardstick

Source: each company’s valuation section — EQT , Expand , Antero , Range and CNX — with reserve values from each FY2025 reserve disclosure. Prices are per-company and dated in the table because the group did not close on one day. Every net asset value is read at this post’s shared US$3.00 base deck — the base column of Figure 8 — so no restatement was needed for any company; all five underlying analyses now run the same base. The NAVs are the analyses’ own, not company guidance, and they differ in estimate content: two of the five credit a risked drilling-inventory row and three print n/d for it (ledger row 13), and three carry a material non-reserve asset — EQT’s midstream, Antero’s listed midstream stake, CNX’s environmental-attributes business — that the other two do not (ledger row 14). The final column is not on one price basis: each company’s PV-10 is struck at its own reserve-report deck, and those decks differ (ledger rows 1 and 2); EQT is shown on both its SEC case and its strip case because its own analysis publishes both.

Figure 7. Price against net asset value

Source: Table 4, which carries the per-company price and its date. Bars are in the post’s fixed market-cap-descending order, not sorted by ratio — cheap and expensive are read against the parity marker, not against the bar above, and the leader marker sits on Antero in third place. The parity marker sits at 1.00 ÷ 2.51 = 39.8% of the axis, the same maximum every bar width is divided by. Every company is included; where a net asset value rests on a different basis, that is stated in Table 4’s basis column rather than handled by exclusion.

Not one of the five trades below the value of its own risked net assets, and the spread between the cheapest and the dearest has more than tripled since the July edition of this comparison, from 0.36× to 1.19×. Antero at 1.32× and Range at 1.41× are the closest to their asset values; EQT at 1.51× is a step above; and Expand at 2.36× and CNX at 2.51× are being valued at roughly two and a half times what their own reserve reports, moved to the shared deck, say the equity is worth. The two most expensive names against net asset value are the ones whose net asset values are thinnest relative to their debt — a point Figure 8 makes mechanical.

On the harder reserve-value measure Range separates. At 0.92× it is the only company trading below the discounted pre-tax value of its own proved reserves, and at US$0.59 per proved mcfe it holds the cheapest reserve book in the set. Expand’s 1.33× and CNX’s 1.16× look milder than their P/NAV ratios because PV-10 is struck before tax and before the debt — the gap between 1.33× and 2.36× at Expand is the whole of what the bridge takes out. This is the clearest case in the post for reading two value measures rather than one.

One of these numbers has been marked by an actual buyer. Blackstone’s December 2024 purchase of 40% of EQT’s midstream vehicle is the one arm’s-length price for any asset in this group, and EQT’s net asset value credits it at the 60% share retained. Every other line in Table 4 rests on a reserve report rather than on a transaction — which is the caveat under every ratio in the column.

3.2 Price sensitivity

Figure 8. NAV per share across the price deck

Source: each underlying analysis’s own price × discount-rate sensitivity grid — EQT , Expand , Antero , Range and CNX — read at that analysis’s base discount rate — 10% for all five, the E&P convention for a single-basin producer — so the price axis is the only variable here. Each cell is that company’s equity net asset value per share in US dollars at that deck. The columns are the fixed natural-gas grid, US$2.00–4.00/MMBtu Henry Hub (grid version 2026-09), so this grid lines up column-for-column with each company’s own; the base-case column (US$3.00) is outlined. NAV/share is an intrinsic figure — lay it against each company’s share price (EQT US$55.59 at 1 Sep; Expand US$97.91, Antero US$39.41, Range US$42.00, CNX US$37.49 at 4 Sep 2026) to read discount or premium. Rows are in the post’s fixed market-cap-descending order — the discount-to-price ranking is read out in the prose and in Section 3.3, not built into the row order, because that ranking moves with the share price while these cells do not. Shading is ranked within each row, on that company’s own minimum-to-maximum across the five columns. Every one of the five holds its unit cost flat across the columns, so each row shows the full operating-leverage effect and none of the cost inflation that would damp it at the high prices — a treatment that is uniform across the group and therefore does not distort the comparison, but does make every row’s high end generous (ledger row 16).

Every net asset value crosses its share price above the base deck, and three cross above the top of the grid. Antero crosses at roughly US$3.79 and Range at roughly US$3.78 — the only two inside the published grid. EQT crosses at about US$4.21, Expand at about US$4.45 and CNX at about US$4.66, all read off the grid’s own slope rather than a printed cell. Not one reaches its price at or below the base deck, which is the arithmetic behind five Overvalued reads.

The leverage read is where the group genuinely separates, and it runs opposite to size. Per US$0.50 step on the grid, Expand’s net asset value moves 47.0% of its base value and CNX’s 45.6% — more than twice Antero’s 20.6% and EQT’s 21.3%, with Range at 26.5% in between. The reason is structural rather than geological: both Expand and CNX carry a thin equity net asset value over a fixed claim stack, so the same absolute move in reserve value is a much larger proportion of what is left for the shareholder. At the bottom of the grid it becomes stark — Expand’s equity net asset value falls to US$2.52 a share and CNX’s to US$1.32, against share prices of US$97.91 and US$37.49. Antero and EQT damp the swing for the opposite reason: a third of Antero’s value is a listed midstream stake that does not move with Henry Hub at all, and EQT’s consolidated midstream is valued on a contracted-infrastructure convention rather than on the gas price.

Rank flips are confined to the two ends. In the middle of the grid the ordering of net asset values is stable, but Expand moves from the second-lowest net asset value in the group at US$2.00 to the highest at US$4.00 — the widest swing in the post, and the direct consequence of the thinnest equity cushion in the set. CNX, the lowest at US$2.00, makes the same journey a rung lower and never clears its price anywhere on the ladder. Antero and Range hold their positions across the whole grid, which is what a long reserve life and a diversified revenue mix look like when the deck moves.

3.3 The gas price each share price assumes

Figure 7 says what the market pays against asset value and Figure 8 says how that value moves with the deck. This is the read the two of them make possible: what gas price each share price is already underwriting.

Table 5. The gas price each share price assumes

Source: the market-implied deck and both flip prices are each underlying analysis’s own published figures, taken unchanged from EQT , Expand , Antero , Range and CNX and never recomputed here — the implied deck is the flat Henry Hub price at which that analysis’s blended value returns exactly the share price shown, and the flip prices are where its value read crosses into the next band up. The crossover deck is this post’s own interpolation of Figure 8, between the two printed cells that straddle each share price, and is stated to the nearest cent of gas rather than presented as a precise threshold; three of the five fall above the top of the grid and are read off the grid’s own slope. Distances are against this post’s US$3.00 base deck. For context, the Henry Hub five-year average is US$3.74 and the EIA’s 2027 forecast US$3.49.

Every share price in this group embeds a gas price above the shared deck, and the spread between them is the widest single difference in the post — wider than the cost spread, wider than the quality spread. Expand’s price asks the least of gas at US$3.80, six cents above the commodity’s own five-year average. CNX’s asks the most at US$4.56, 22% above that average and more than half again the base deck. The group’s implied decks run from +27% to +52% over the base, which is the arithmetic version of a simple statement: the value dispersion in this peer group is a disagreement about the price of gas, not about the quality of the companies.

The two decks diverge for one company, and the divergence is informative. For four of the five they sit within about 50 cents of each other. Expand is the exception: its net asset value does not reach its share price until about US$4.45, while its blended value — which weights cash-flow multiples alongside the asset value — returns the price at US$3.80. That gap is the whole of Expand’s thesis: its reserve base does not support the price, its cash flow comes much closer. Table 4’s 2.36× and this table’s US$3.80 are the same fact stated twice.

The flip prices say how far each name is from a different verdict. Expand needs only US$3.14 to move out of the bottom band and US$3.58 to read Fairly valued — the shortest distance in the group, and inside the five-year average. CNX needs US$3.67 and US$4.26, the longest. In between, Range and Antero need about US$3.25–3.28 to move a band and US$3.84–3.96 to reach Fairly valued, and EQT needs US$3.50 and US$4.15. No company here needs a gas price the market has never seen — every threshold sits inside the fixed grid or a little above it — but they are not the same distance away, and that distance is the most useful single number to carry out of this post.

4. Rating Scoreboard

The compared companies are the peer set. Every relative dimension here — asset quality and scale, cost position, reserves and life, balance sheet, capital allocation — is scored against these five US names and no others. That is very nearly what the underlying analyses already did: all five declare this same five-name group as their peer set, each naming the other four. One limitation is real and worth stating: two of EQT’s rationales — its cost position and its capital-allocation record — are argued against “the six producers reviewed in this series”, a wider North American set that includes a Canadian name this comparison excludes. Re-read against these five alone, neither star moves: EQT still holds the lowest cash cost of the four full-stack filers and still funded its scale with equity. Jurisdiction, which the broader North American comparison split across two countries, is fully unanimous here and is flagged as such below.

The weighting is the producer/operator archetype, which is the reference case: the five dominant dimensions — asset quality, cost position, reserves and life, balance sheet, capital allocation — carry 15% each, and the four base-weight dimensions — growth, management, jurisdiction, ESG — carry 6.25% each. All nine apply to all five companies; none is marked not-applicable, because all five are producer/operators.

Table 6. The nine-dimension scorecard

Source: the Metal Pilot Company Scorecard as applied in the five analyses linked in Section 6.1, where every score is substantiated with a sourced figure. Rows are in weight-descending order and carry no dimension numbers — the reordering would put the scorecard’s own numbering out of sequence, and each dimension’s name identifies it unambiguously. Every composite is Σ(weight × score) on the weights in column 2. Range, the highest: 0.15 × (4 + 4 + 5 + 5 + 5) + 0.0625 × (3 + 4 + 5 + 5) = 3.45 + 1.0625 = 4.51. Antero, the lowest: 0.15 × (4 + 2 + 5 + 3 + 4) + 0.0625 × (4 + 3 + 5 + 4) = 2.70 + 1.00 = 3.70. Bands map from the composite rounded to the nearest half-star, which is why EQT at 4.34 lands in High quality. The Delta row is empty in every column, and that is the result rather than an omission: all five source posts now publish the weighted composite this table recomputes, so there is no rating left open across the group.

Reading the table across the rows is where it earns its keep.

Cost position has the widest spread in the table, from 2 to 5, and it does not track size at all. CNX, the smallest company here, scores 4 — the same as Expand and Range, and two above Antero, which is more than twice its size. Antero’s 2 is the only score below 3 anywhere in the group and it rests on one number: US$2.27/mcfe of gathering, processing and transport against CNX’s US$0.54. Scale and unit economics are separate dimensions in this rubric for exactly this reason.

Reserves and growth trade off against each other almost perfectly, and one company breaks it. The three scoring 5 on reserves, life and replacement — Antero, Range and CNX — score 4, 3 and 2 on growth. The two scoring highest on growth, EQT at 5 and Expand at 4, score 4 and 3 on reserves. CNX is the break: a 5 on reserves and the group’s only 2 on growth, which is one fact — a long-life base being deliberately shrunk — pointing in two directions.

The reserves row contradicts the size ranking outright. The three smallest companies in the group all score 5 and the two largest score 4 and 3. Expand’s 3 is the lowest reserves score in a set where reserve life runs from 9.9 to 22.2 years, and it belongs to the largest producer by volume. A reader who thinks of scale as durability has the ranking backwards in this peer group.

Capital allocation is the only 15% row where the smallest company outscores the largest, CNX at 4 against EQT at 3 — and Table 2 corroborates it from a different direction: CNX returns 5.0% of its market value as free cash flow at the base deck against EQT’s 1.5%, and has retired shares while EQT’s count rose 5.93%. EQT assembled its scale by issuing equity, and that is what the 3 is.

Management is the outlier row, and it is one company. Expand scores 2 — the lowest score anywhere in the table, on any dimension, for any company — because it has run on an interim chief executive since February 2026 with no permanent successor named, while integrating the largest merger in the group’s recent history. Its four other base-weight scores are 4, 5 and 5. Neither Expand’s management score nor EQT’s capital-allocation score moves the composite as much as it would move a reader who cares about it, which is what the row reads are for.

One row discriminates nothing, and it is worth flagging as such. Jurisdiction and geopolitics reads 5, 5, 5, 5, 5 — fully unanimous. All five are pure-play US operators in top-tier rule-of-law states, so the dimension does no work in this peer set and yet carries 6.25% of every composite. It is kept in the table rather than dropped, because an absent dimension reads as an oversight; but a reader weighting jurisdiction should know it separates no one here.

Two pairs need separating, and only one of them is stable. Antero and CNX both print 3.7/5, so the tie is broken on the dominant dimensions in weight order: the first on which they differ is asset quality and scale, where Antero’s 4 beats CNX’s 3, and Antero therefore places fourth. Recomputed as a plain unweighted mean of all nine dimensions that order holds — Antero 3.78 against CNX 3.67 — because CNX’s weakest score of all, a 2 on growth, is cheap at 6.25% and equal weights make it bite. At the top the ranking is not stable: EQT and Range tie exactly at 4.44 on equal weights, where the archetype weighting puts Range ahead. So a reader who weights the nine dimensions equally gets the same fourth and fifth places and a tie at the top. Range leads on both schemes only because it is the one company here with no score below 3 on any dominant dimension.

5. Summary

Table 7. Quality × Value

Source: composites from Table 6, value reads and ratios from Table 4, flip prices and triggers from Table 5 and each analysis’s own conclusion. The per-company price behind each ratio and its date are in Table 4 and are not repeated here. The NAVs vary in estimate content — Table 4’s basis column states each one — so the ratio column compares construction, not certainty. Verdict language is the standard Quality × Value matrix from the Metal Pilot Company Scorecard, unchanged. The “(wide band)” qualifier travels with every read: each analysis’s deep-bear blend sits 76% to 93% below its share price.

Figure 9. Quality against value

Source: Table 7. The y-axis runs the full fixed 1-to-5 composite range so peer groups stay comparable across this series; the x-axis is price to net asset value ascending, so cheap is on the left. The x-axis starts at 1.0× parity, which is the left edge: every company in this group trades above the value of its own net assets, so no part of the plotted range is below parity. The shaded bands are the quality bands — High quality at 4.5 and above, Average and below at 3.5 and under. The price behind each ratio and its date are in Table 4, not in the graphic. Every point is on the same footing; the dominance screen below is a read of these two coordinates, not a second class of company.

The value axis has collapsed into one band, and that is the headline of this edition. All five names read Overvalued (wide band) at the shared deck, so the Quality × Value matrix — which spread this group across three columns in July, when two names read Fairly valued and none was more than 1.37× its net asset value — now places every company in a single column and two boxes: Range and EQT in “great company, rich price”, Expand, Antero and CNX in “full”. The entire undervalued half of the matrix is empty, and so is the low-quality row. That is a real finding rather than a failure of the tool: it says the market is not mispricing quality inside this group, it is paying a higher gas price than the reserve reports support, for all five at once. The discrimination the matrix has lost is in Section 3.3, where the same five names spread across a 76-cent range of implied gas prices.

The dominance screen. Comparing the composite out of 5 against price to net asset value as published — no normalisation, no blended score — three of the five are beaten by another peer on both measures at once:

• EQT (4.3/5, 1.51×) is beaten by Range (4.5/5, 1.41×). EQT still has the best unit economics in the group on both margin lines, the largest reserve base, the only third-party mark on any asset in this comparison, and more than 30 years of drilling locations.
• Expand (4.2/5, 2.36×) is beaten by Range and by EQT on both. It is still the volume leader, the only name here with genuine Gulf Coast LNG adjacency, the lowest leverage in the group and the only meaningful dividend.
• CNX (3.7/5, 2.51×) is beaten by Range, EQT and Expand on both. It still earns the highest return on invested capital in the group at 17.8%, the highest free-cash-flow yield at the base deck at 5.0%, and runs the lowest cash cost per unit of any company here.

What survives is Range and Antero — the group’s quality-price frontier, and the same two names that survived in July, now by a wider margin. That is a set, not a ranking: Range is the quality end at 4.5/5, and Antero the value end at 1.32×, the only two coordinates no other peer beats on both axes. Antero survives by 0.09×, the gap between its 1.32× and Range’s 1.41× — thin, but four times the 0.02× margin of the July edition. The screen removes names on arithmetic and cannot order the two it leaves standing. Being out-argued on two coordinates is not the same as being a bad company, and none of the three dominated names is one.

The value-trap quadrant is empty, and the nearest thing to one is the most expensive name. No company here is both cheap and weak — nothing trades below parity at all. The name that most deserves the value-trap discipline is CNX, which pairs the group’s joint-lowest composite with its highest price-to-NAV at 2.51× and its highest implied deck at US$4.56. What would have to go right is specific: the 2027 and 2028 hedge strikes step up to US$3.28 and US$3.25 from this year’s US$2.74 — the most mechanical margin improvement available to anyone in the group — and Appalachian demand would have to hold gas near the price the equity already assumes. What would go wrong is equally specific: at the US$2.00 column its equity net asset value is US$1.32 a share, and the hedge book is the only thing holding it there. A 2.51× price on a Solid-quality company is not a bargain waiting to be recognised; it is a company priced for a gas market that has not arrived.

Which name answers which question. The screen says two names are not beaten on both axes. It cannot say which of the five suits a given reader, because that depends on what the reader is buying. This table answers that directly: one row per question, the name that leads that one published figure, and what leading it costs. Nothing here is weighted, scored or ranked — every cell points at a number printed above.

Table 8. Which name answers which question

Source: every figure in this table is printed in the table or figure named beside it; nothing here is computed, weighted or ranked. The full fixed question set is answered, plus the one addition this archetype earns — most protected at the low deck, which only a group where every peer hedges can answer. This is a map of preferences, not a pick — a name appearing in a row is the leader on that row’s metric and nothing more.

The consensus cross-check.

Table 9. Analyst consensus against this analysis

Source: stockanalysis.com analyst consensus as read by each underlying analysis on 4–6 Sep 2026. Rows are in the post’s fixed market-cap-descending order. Implied upside is against the per-company price and date shown, which is why both dates are carried: consensus targets are frequently struck on a different day than the post’s window. Coverage is thinnest at CNX (12 analysts, half of Expand’s), which is part of why its target sits on top of its price. The market-implied deck each price embeds is in Table 5 and is not repeated here.

Every target is above the current price, and the equal-weight average of the five implied upsides, 17%, is another way of saying the sell side underwrites a gas deck above the one the reserve reports are struck at. The gap is not about company quality — the two Hold ratings belong to the highest-quality name in the group and to one of the joint-lowest. It is about the deck, and Table 5 puts a number on it: every share price here already embeds US$3.80 to US$4.56 gas against a US$3.00 base, so a target above that price underwrites more still. Range’s Hold against the group’s highest composite is the pattern worth naming — broad coverage, modest upside, no argument about the business, which is what “priced for its quality” looks like from the sell side. CNX’s target sitting 0.6% above its price is the other end of the same observation: the thinnest coverage in the group sees no gap either way, while this analysis, on the audited reserve value, sees the largest one in the set.

What this post is not. The shortlist is a set of two names no peer beats on both axes at once — it is not a shopping list, it is not ordered, and it says nothing about which of the two suits any particular reader. The decision map above is a map of preferences, not a set of recommendations: a name leading a row leads that metric and nothing else. The three dominated names are not names to avoid; each was out-argued on two coordinates by a peer, and each carries a real strength named above. And a value read of Overvalued is not an instruction to sell — it is a statement that at a US$3.00 deck the arithmetic does not support the price, together with the gas price at which it would. To run the same nine dimensions, cost lines, reserve metrics and valuation ratios across the whole North American upstream universe rather than these five, explore Metal Pilot.

6. Sources, methodology & disclaimer

6.1 Sources, methodology & data vintage

This post contains no primary research of its own. It is a synthesis of five single-company analyses published on this blog, each built from that company’s fiscal-2025 Form 10-K and most recent quarterly results; the contribution here is putting all five on one construction. The underlying analyses, with every figure sourced and every scorecard score substantiated:

• EQT Corporation (EQT) — Stock Analysis 2026 [4.3] — valuation as of 4 September 2026
• Expand Energy (EXE) — Stock Analysis 2026 [4.2] — valuation as of 6 September 2026
• Antero Resources (AR) — Stock Analysis 2026 [3.7] — valuation as of 6 September 2026
• Range Resources (RRC) — Stock Analysis 2026 [4.5] — valuation as of 6 September 2026
• CNX Resources (CNX) — Stock Analysis 2026 [3.7] — valuation as of 6 September 2026

For the market backdrop these companies operate in, see the Natural Gas — A Complete Market Guide ; the cost ranking for the wider universe is linked in Section 1. Two Canadian gas producers — Tourmaline and Birchcliff — are excluded from this US comparison because they price against the AECO discount and report on the NI 51-101 reserve standard, as stated in Section 1.

Market data, analyst consensus, return-on-capital and share-count figures are from stockanalysis.com , sourced from S&P Global Market Intelligence and Fiscal.ai, at the 1–4 September 2026 closes as read by each underlying analysis. The Henry Hub trailing averages behind the price deck — 3-month US$2.94, 6-month US$2.93, 12-month US$3.59, and a five-year average of US$3.74 — are from the U.S. EIA monthly series; the forecast carried as a 0%-weight cross-check is from the U.S. EIA Short-Term Energy Outlook .

The comparability ledger. Every place the one-construction rule bends, with the direction of the bias:

Table 10. Comparability ledger

Source: this analysis, from the disclosures cited in Tables 1–5 and the assumptions box of each underlying analysis — EQT , Expand , Antero , Range and CNX . Most rows are cited by number from the table or figure they qualify; rows 3, 4 and 17 qualify the net-asset-value column of Table 4 and the timing statement in the opening note collectively rather than any single cell.

Methodology — the choices, and what each costs.

1. One construction for unit economics — realized price after royalty, less all cash costs including corporate overhead, less depletion, interest excluded. Buys: five margins that mean the same thing. Costs: it discards the segment figures several of these companies headline, including the CNX Shale margin of US$1.17/mcfe that three of the underlying posts benchmarked peers against.
2. Every figure is read from a published analysis, never computed here. Buys: every number traces to a section of a source post. Costs: where a company publishes no line on the group basis — CNX’s depletion, Antero’s leverage ratio, three of the five inventory lives — the cell prints n/d and the row loses a company rather than gaining an estimate.
3. Price to net asset value as the primary yardstick, struck at the shared US$3.00 base column, with EV per unit of reserve value as the cross-check — the enterprise value against PV-10, the discounted pre-tax value of the proved reserves the filing itself publishes. Buys: one number per company, already bridged through its own debt and claims. Costs: the NAVs differ in estimate content, which Table 4’s basis column states and no single ratio can express.
4. The base deck is the fixed grid price nearest the trailing Henry Hub average, not a house forecast and never a spot quote. Buys: a price axis that lines up column-for-column with each underlying analysis’s own sensitivity grid, so a reader can carry a number straight across. Costs: the base (US$3.00) sits below the EIA’s 2027 forecast (US$3.49) and well below the five-year average (US$3.74), which is a large part of why all five reads land in one band.
5. Canonical units for the per-dollar rows — bcfe per year for flow, bcfe for stock, per US$1 bn of market capitalisation. Buys: the flip between absolute and per-dollar scale. Costs: every such row is dated by the share price behind its denominator.
6. The producer/operator archetype weighting, dominant dimensions at 15% and the rest at 6.25%. Buys: auditable composites that all five source posts now reproduce. Costs: the weighting is an editorial choice, and Section 4 publishes what changes without it — EQT and Range tie on equal weights, while the fourth and fifth places hold.
7. The hedge book is compared on its mark, not on its coverage. Buys: the one hedge figure every analysis publishes on one construction — the same bridge line at the same five prices — instead of five incompatible coverage percentages. Costs: the figure carries one flat row (Expand), and a mark is a valuation of a position rather than a measure of how much volume is protected.
8. The decision map is a map, not a score. Buys: a reader who can name their question can find the name that leads on it, with the cost of that leadership printed beside it. Costs: a row can be misread as a recommendation, which is why every row carries a cost clause and the closing paragraph says plainly what the table is not.
9. A comparison-specific figure set, and one figure deliberately not built. Buys: every figure answers a cross-company question, as an inline HTML/CSS component. Costs: no share-count-change figure is published — only two of the five disclose a latest-year percentage (ledger row 9), so a bar chart of three different disclosure shapes would chart the disclosure rather than the companies. Table 2 carries each as stated.

This is a dated artifact. Like the analyses it draws on, it carries market capitalisations, enterprise values and valuation multiples that go stale quickly — the deliberate deviation from this blog’s normal practice of keeping company posts free of point-in-time valuations. Every such figure is dated, and the whole post is refreshed when the underlying analyses are. Data as of 9 September 2026.

Timing spread. The five underlying analyses were struck between 4 and 6 September 2026 — EQT’s valuation as of 4 September on market data at the 1 September close, and the other four as of 6 September on the 4 September close. All five run the same US$3.00 base deck on the same fixed grid. EQT’s analysis is one template version behind the other four; the difference does not touch any figure used here. No company in the group has reported results since its own analysis was struck.

Provenance: EQT Corporation — 10-K Filing — 2025; Expand Energy — 10-K Filing — 2025; Antero Resources — 10-K Filing — 2025; Range Resources — 10-K Filing — 2025; CNX Resources — 10-K Filing — 2025.

6.2 Disclaimer & disclosure

This comparison is for informational purposes only and is not investment advice; do your own research or consult a licensed advisor. It is a point-in-time snapshot as of 9 September 2026 — share prices, multiples, analyst targets and value reads all move. Reserve and net-asset-value figures are estimates prepared under stated SEC conventions and do not represent market value. The ratings and verdicts are analytical reads of quality and price, not buy or sell instructions. A name removed by the dominance screen is not a name to avoid, and a value read of Overvalued is not an instruction to sell. This report was prepared with AI assistance; figures were sourced from company filings, the U.S. EIA and market data and reviewed, but readers should verify before acting. The author holds no position in any of the five companies as of the date of writing.