Primary metal manufacturing

The pre-screen called this national-scale capital from the title alone. The filings say the same thing with numbers, and add something the title does not: capital is the entry ticket, and it does not buy a margin. The scale first. Cleveland-Cliffs bought Stelco, one Ontario flat-rolled producer, in 2024 at an enterprise value of about US$2.5B (C$3.4B) [A]. That is the price of one working mill. The 63 Canadian establishments are misleading as a count: 32 have fewer than ten employees and are not steelmakers in any sense that matters, while twelve employ 200 or more, and those twelve are the industry. Then what the capital earns. Algoma Steel, the other Ontario integrated producer, reported 2025 revenue of C$2,085.7M, down from C$2,461.7M, and a net loss of C$984.9M, with adjusted EBITDA of negative C$261.4M [A]. It paid C$225.0M in direct tariff costs during the year — about 11% of revenue — and took C$500M of government-backed liquidity through the Large Enterprise Tariff Loan facility to get through its conversion to electric-arc steelmaking [A]. The price of its product is set by the North American sheet market and its access to that market by US trade policy; neither is in the mill's hands. An electric-arc mini-mill is the smaller version of the same bet, not a different one: it is still a furnace, a caster and a rolling line sunk before the first tonne, selling a commodity. There is no entrant-sized proposition inside this group. What is sold to mills — scrap, refractories, maintenance, process software — is where ordinary resources reach, and those are other records.

The pre-screen cut this on capital — "rolling and finishing plant". That is half right. A wire-drawing line or a tube mill is real money, but it is an order of magnitude below a steel mill, and the size bands show it: 112 of the 180 Canadian establishments, 62%, have fewer than twenty employees. People do own small plants in this group. The better question is what such a plant earns, and one Canadian company answers it in public. Tree Island Steel of Richmond, B.C. has drawn wire from purchased rod since 1964 — galvanized wire, nails, reinforcing mesh, fencing. In 2025 it reported revenue of C$161.8M, down from C$207.0M, and gross profit of C$9.5M — a gross margin under 6% — with adjusted EBITDA of C$3.0M and a net loss of C$5.3M, after a C$4.1M loss in 2024 [A]. It cut 27% of its workforce, withdrew from unprofitable product lines, suspended its dividend, and attributes the decline to expanded US tariffs on wire products [A]. That is the mechanism. A converter buys its steel from mills and sells a product defined by a standard, so its margin is the spread between two prices it does not set: rod or coil on one side, and on the other a finished-wire or tube price anchored by imports and by mills that convert their own steel downstream. Sixty years of operating history, brands and two countries of plant did not protect that spread. An entrant would have the same suppliers, the same standards and none of the history. Pipe and tube for energy markets follows drilling cycles and was not separately examined.

Canada is a serious aluminum country — about 3.3 million tonnes of primary metal in 2024 from ten smelters, nine of them in Quebec, and C$17.4B of aluminum exports, 91% to the United States [A]. The pre-screen called smelting "power-contract economics", which is true and incomplete: the power contract decides whether a smelter lives, but the pot line decides whether anyone but a major can build one. Rio Tinto has just priced that. In 2023 it approved 96 new AP60 pots at its Saguenay complex for US$1.1B, adding about 160,000 tonnes a year; by the time commissioning began in 2026 the company was describing a US$1.5B expansion [A]. That is roughly US$9,400 per tonne of annual capacity, for a brownfield addition by the owner of the technology, on a site that already had the power, the port and the alumina. A greenfield entrant has none of those and would need a multi-decade hydro block that Quebec allocates as industrial policy, not as a product for sale. The rest of the group — rolling, extruding, alloying, secondary remelt — is where most of the 99 Canadian establishments sit, and the size bands are bimodal: 41 have fewer than ten employees, 25 have a hundred or more. Extrusion is the conceivable entry point, a press and a die shop serving window, trailer and solar-racking makers. It was not examined here, and nothing in this record should be read as a verdict on it. It would need its own screen, and that screen would start from billet supply: the press buys its metal from the same smelter owners at the exchange price plus a regional premium, and sells a shape its customer can re-quote.

Copper, zinc, lead, nickel and precious-metal smelting and refining, plus the rod, strip and alloy mills downstream. Canada has real assets here, and 2025 offered an unusually clear look at two of them. Teck's Trail Operations in B.C., one of the world's largest integrated zinc and lead complexes, booked a C$1.1B pre-tax impairment in 2024, when the test put the post-tax recoverable amount of the whole Trail cash-generating unit at C$666M [A]. Teck's stated cause was a "challenging environment for treatment charges due to a global shortage of zinc concentrate". That sentence is the business model: a custom smelter is paid a treatment charge set by the world balance between mine output and smelter capacity, and when concentrate is short, smelters bid that charge down against each other. Trail swung from a C$66M gross loss in 2024 to a C$281M gross profit in 2025 on C$2,489M of revenue — and Teck attributes the recovery to silver, germanium and indium by-product prices and to running stockpiled residues in place of purchased concentrate, while refined zinc output was deliberately cut to 229,900 tonnes [A]. The profit came from not doing the core activity. In Quebec, Glencore had planned nearly C$1B over five years at the Horne copper smelter, C$300M of it for emissions reduction, and suspended all of it in February 2026 for want of regulatory certainty ahead of limits that take effect in March 2027 [A]. So the capital runs to ten figures, is periodically written off by its owners, and is hostage to an air permit. Eighteen of the 78 Canadian establishments have fewer than five employees — most likely small alloyers and precious-metal refiners. That fringe was not examined.

The pre-screen cut foundries on capital. Tested, that is not quite the reason. A jobbing foundry — melt, mould, pour, clean, ship castings to someone else's drawing — is a mid-sized plant, not a mega-project: 81 of Canada's 159 foundries, 51%, employ between twenty and 199 people, and the US average is about 78 per establishment [A]. Plants of that size change hands, and an entrant would buy one and not build one, since a new melt shop would also need a new air permit. The capital is reachable. The cut is what is being bought. The American Foundry Society counts 1,750 metalcasting facilities in the United States, against 3,200 in 1991 and 6,150 in 1955, and says 75% of firms have fewer than a hundred employees [B]. That is a population that has lost 45% of its members in a generation. Some of that will be consolidation into larger plants, which a plant count cannot distinguish from closure — but for a small jobbing shop the direction is the fact that matters: each year there are fewer of you, and the ones that left did not leave because they were bought at a premium. Castings are specified by drawing and alloy, so a buyer can move a pattern to another foundry, in Ontario or in Asia, and the foundry's remaining edge is lead time and proximity on short runs. The screen did not find evidence on the two things that would overturn this: whether surviving Canadian foundries are earning more as competitors close, and whether reshoring and tariffs have turned the order book. The nearest listed comparable points the same way: Ampco-Pittsburgh's forged-and-cast segment turned $292.6M of 2025 sales into a $44.7M operating loss, and its UK roll foundry went into insolvency in October 2025 rather than finding a buyer [A]. That is a roll maker, not a jobbing shop, and no jobbing shop reports. A full study would start there, with a specific plant's customer list.