Fortescue (FMG)

Two Reporting Traps Before Any Number

Get these wrong and every comparison you build on Fortescue is off. First, the fiscal year ends 30 June: FY2026 means the year to 30 June 2026, and the June 2026 quarter is Q4 of it. Vale reports on a calendar year, so a casual "FY2025 versus FY2025" comparison between the two is six months out of line. Second, the shares trade on the ASX in Australian dollars while the accounts are kept in US dollars. Convert at a market rate, never at the company's own guidance rate. Fortescue assumes AUD:USD of 0.65 for FY2026 and 0.70 for FY2027; run the A$61.2-63.2 billion market cap of 9 June 2026 through 0.65 rather than the RBA's 0.7020 that day and you lose more than US$3 billion of it.

The Cost Position

Fortescue is the simplest business in the bulk-commodity set: dig iron ore in the Pilbara, ship it to China, and do it cheaper than anyone else of comparable size. FY2026 shipments were a record 201.3 Mt. Of that, 192.3 Mt was hematite, the ore rich enough to sell after little more than crushing and screening, and 9.0 Mt came from Iron Bridge, where a lower-iron magnetite rock has to be ground and concentrated first. The hematite C1 cash cost, meaning the direct cost of mining, processing, rail and port before royalties, freight and capital, was US$18.74 per wet metric tonne. That is bottom-of-the-curve territory alongside BHP.

The other three disclose on their own year-ends and their own definitions, so read the basis column as part of the number:

The footnote matters more than usual here: each company defines C1 differently, so this is a ranking with caveats, not a like-for-like measure. And C1 is only the start of the cost stack. Freight, royalties, sustaining capital and corporate costs all sit on top before you reach a breakeven. Vale is the only one of the four to publish an all-in figure, US$54.2/t against its US$21.3/t C1, but do not borrow that ratio for Fortescue: most of Vale's gap is the voyage, and Brazil to China is roughly three times the Pilbara run. Fortescue's own step from C1 to all-in is considerably shorter. Our iron ore cost curve guide works through the definitions and how to footnote them properly.

What the cost position is worth shows up on the earnings line: US$7,941 million of underlying EBITDA in FY2025 on 198.4 Mt shipped, about US$40 a tonne after freight, royalties and everything else C1 leaves out. That is the iron ore parallel to the EBITDA-per-tonne metric used to compare steelmakers.

Costs are drifting up, and the company is guiding for more of it. Hematite C1 went from US$17.99/wmt in FY2025 to US$18.74 in FY2026, reached US$19.37 in the June 2026 quarter on higher diesel prices, and FY2027 guidance is US$20.50-21.75/wmt. Read that step carefully before calling it cost blowout. Fortescue spends in Australian dollars and reports in US dollars, and the FY2027 range assumes AUD:USD 0.70 against 0.65 the year before. Translating unchanged Australian spending at the higher rate adds about US$1.40/wmt by itself, out of a US$2.40 rise to the midpoint of guidance. The remaining dollar is the real deterioration.

The Index Discount Is the Lesson

Fortescue's ore grades below the benchmark, so it never receives the index price. In FY2026 it realised US$90.66 per dry metric tonne against a Platts 61% Fe index averaging US$102.90, or 88 cents in the index dollar. Any model that multiplies Fortescue's tonnes by the benchmark overstates revenue by more than a tenth before it has done anything else.

The discount is not stable either, and the thing that moves it is the profitability of Fortescue's customers. When Chinese steel mills are earning well, the constraint on them is furnace capacity rather than input cost, so an extra point of iron content is worth paying up for and the discount on low-grade ore widens. When their margins are squeezed they hunt the cheapest iron units available, and the discount narrows. Fortescue's own history is the clearest demonstration: it realised around 65% of the index in FY2018, when Chinese mill margins were the strongest in a decade, against the mid-to-high 80s through the thin-margin years since. The grade discount is a partial hedge, widest when the market is hottest. That is also why a realisation line has to be a flexed input in an iron ore model rather than a fixed assumption.

One trap in reading a trend into those percentages. Platts changed its flagship index on 2 January 2026, from 62% Fe to 61% Fe. Measure a below-benchmark producer against a lower-grade reference and its realisation improves on the same ore at the same price, so FY2026's 88% is not straightforwardly better than the 84% of FY2025 against the old 62% mark. The June 2026 quarter alone came in at 84%, US$88.85/dmt against an index averaging US$105.29. Check which index a figure is quoted against before comparing periods. Note the units too: C1 comes per wet metric tonne (wmt) and realised prices per dry metric tonne (dmt), so the two cannot be netted without a moisture adjustment.

The Cleanest Balance Sheet in the Sector

Net debt was US$0.8 billion at 30 June 2026, down from US$1.1 billion a year earlier, against FY2025 underlying EBITDA of US$7,941 million. Pair those however you like and the answer rounds to about a tenth of one year's earnings. For bulk producers we screen below 1.0x as conservative, 1.0-2.5x as normal, and above 2.5x as stretched; Fortescue barely registers. The contrast with the steelmakers it sells to is stark: ArcelorMittal carried US$7.9 billion of net debt at around 1.2x, and Cleveland-Cliffs held US$7.2 billion against trough EBITDA that made the ratio meaningless.

For a single-commodity producer selling into one dominant customer market, the balance sheet matters more than the income statement in a downturn. But do not read the gap between the iron ore price and C1 as the margin, which is the standard way to get Fortescue badly wrong. Our planning price of US$90/t is a benchmark price delivered to China; Fortescue realises about 88% of the benchmark, so call it US$79 for a tonne of its product. A C1 of roughly US$19 does not leave US$60 of that. C1 excludes the ocean freight, the Western Australian state royalty, sustaining capital and head office, and it is struck on wet tonnes while the price is per dry tonne. Take those out and FY2025 lands at US$40 a tonne of EBITDA. At US$90 iron ore Fortescue is comfortably cash-generative and solvency is not the question at this level of debt. Free cash flow, against US$3.6 billion of capital spending in FY2026, is a live one.

What to Watch

Where C1 settles. FY2027 guidance of US$20.50-21.75/wmt would leave Fortescue only US$2-3 under Rio Tinto's FY2025 Pilbara figure, on a gap that was US$5.50 in FY2025. Most of that step-up is the guidance exchange rate rather than cost. If the operating part sticks, the one advantage that justifies a single-commodity, single-market structure gets thin.

The realised price, not the index. Realisation was 88% across FY2026 and 84% in the June quarter, so the two lines can move in opposite directions and often do. The index tells you about the market; only realised price per dry tonne tells you what Fortescue banked.

Iron Bridge. The magnetite operation is Fortescue's partial answer to its own grade problem: concentrate earns a premium rather than a discount, realising 117% of the index in FY2026 against hematite's 88%. It is also where the execution risk has been. Shipments rose 27% to 9.0 Mt against a 22 Mtpa nameplate, and in July 2026 Fortescue said it would take a US$750 million pre-tax impairment (US$525 million after tax) on the asset after revisiting the ramp-up schedule. Guidance is 11-14 Mt for FY2027, with an annualised rate above 16 Mt targeted during FY2028. Fortescue owns 69% of the joint venture; those tonnages are quoted on a 100% basis.

The iron ore price. We plan at US$90/t, against a 2021-2025 average of US$123 and an index averaging US$105 in the June 2026 quarter. Fortescue's margin survives all of those. Its free cash flow and dividend capacity do not treat them equally.

Key Risks

One commodity, one customer market. Chinese crude steel output is the demand side of the entire revenue line, and there is no second business to absorb a structural decline in it. The buy side is consolidating against that, too. China Mineral Resources Group was created in 2022 to negotiate on behalf of hundreds of mills rather than leave each to deal with the miners individually, and Fortescue's June 2026 quarterly reports talks with it as ongoing. The exposure lands harder here than at the other majors because Fortescue's biggest product is its cheapest: Super Special Fines were 86.4 Mt of FY2026 shipments, 43% of the book.

Grade discount widening. Realisation has run in the mid-to-high 80s recently and printed 84% in the June 2026 quarter. It has been as low as around 65%, which takes a third off the revenue a benchmark-priced model would have shown. Nothing about the recent range makes it a floor.

Cost inflation. Labour, diesel and rising strip ratios, the waste rock that has to be moved for each tonne of ore as a pit deepens, all push C1 up over a mine's life. Diesel alone was the main reason the June 2026 quarter came in 6% above the one before.

The energy spending, and what it is for. Fortescue cancelled its Arizona and Gladstone green hydrogen projects in July 2025 at a US$150 million pre-tax write-down, and the money has gone into its own operations since: the Pilbara green grid, the 690 MW Turner River solar farm, the Nullagine wind project. That is a bet on removing diesel from the cost base, and it is a large one for a company whose case rests on being the cheapest producer. Judge it as a cost programme, because it is no longer a second revenue line.

Native title. In July 2026 the Federal Court ordered Fortescue to pay the Yindjibarndi Ngurra Aboriginal Corporation A$150 million for cultural loss, plus economic loss and interest. The sum is small against the balance sheet. The precedent, for a compensation claim over mining on native title land, is not.