Gerdau's Best Margin in Four Years Is Almost Entirely American — and Management Won't Sell It

North America throws off 77% of EBITDA at a four-year-high margin, yet the spin-off question gets a flat no; the answer instead is a Brazil self-help plan nobody will fully describe until December 4.

GOAU4.SA · Earnings Call · 2026-10-27

A quarter cleaved in two

Gerdau's third quarter of 2026 is one number pretending to be two. Consolidated EBITDA came to BRL 3.8 billion at a 20.2% margin — “the highest since the second quarter of 2023” — Rafael Dorneles Japur, Executive · 2026-10-27 — with adjusted net income of BRL 1.5 billion and earnings per share up from BRL 0.74 to BRL 0.79. Free cash flow of BRL 1.4 billion was six times the second quarter, and nine-month free cash flow swung to BRL 1.7 billion from a BRL 1.0 billion consumption a year earlier. Net debt sits at 0.58x EBITDA, the revolver was renewed for five years and upsized 29% to $1.125 billion, and S&P moved the outlook to positive on a BBB rating. Then the split. North America printed a 27.7% EBITDA margin, the best in four years, on what management calls resilient demand and strong operating performance. Brazil delivered only a "slight recovery," still buried under what the CEO described as pressure from imports and a local oversupply of steel. The cleanest way to see the company is the mix Japur disclosed: 77% of EBITDA now comes from North America, up from 75% the prior quarter.

The margin machine, and the refusal to monetize it

Gustavo Werneck's framing of North American demand was unusually concrete: “a very strong order backlog driven by a solid portfolio from segments such as renewable energy and data centers” — Gustavo Werneck, Executive · 2026-10-27, with naval destroyers and submarines, nuclear plants and transmission lines layered on top. That is a steel story attached to the data centers and grid build-out — and it is worth noting the contrast with the tape. Across the global price history the AI data-center complex is the single worst-performing cluster over 90 and 30 days, with fifty decliners in the basket. Gerdau sells long and structural product into the physical build, not merchant silicon, and the two trades are now moving independently. The risks are nameable, and management named them. Deacero's new beam mill will export into North America; the ramp is slow but real, and a US–Mexico bilateral deal would hurt most: “it would be kind of negative if there is an agreement between Mexico and the U.S., and it would be net positive if we have an agreement between the U.S. and Canada” — Rafael Dorneles Japur, Executive · 2026-10-27. Scrap is the quiet lever — scrap price discipline is why the margin held even as Chinese steel flooded everywhere else. So the question Leo Correa of BTG asked was the right one, and the answer was the story of the call:

But we have no plans to have a spin-off or any major acquisition in North America to increase our exposure there significantly. We don't have that in our radar at this point.
Rafael Dorneles Japur, Executive · 2026-10-27

A business generating three-quarters of group profit, at a four-year-high margin, inside a company whose entire equity is valued at roughly $1.77 billion. The holding-company discount is not discussed. It is simply declined.

Brazil: a plan with a December deadline

The Brazilian half of the call is where the trade defense mechanism language lives — and it is worth being clear that this is not a company-unique insight. The same phrase sits high in the market-wide keyword set this quarter alongside technical rigor, which is the industry's polite way of saying the anti-dumping process is proceeding slowly. Werneck's version: “the fact that they are having this technical rigor is a positive approach by MDIC” — Gustavo Werneck, Executive · 2026-10-27 — progress, but not a lever the company will bet on. With Lula re-elected, the base case is continuity of tax reform and gas-market liberalization, not a sudden wall against imports. What is genuinely company-specific is the self-help plan, deliberately withheld until the December 4 Investor Day, and already visible in the CapEx line. Disbursements are tracking closer to BRL 4 billion than the BRL 4.5–4.7 billion guided, and Werneck was blunt about why capital is migrating: “CapEx return abroad has been much more competitive than in Brazil” — Gustavo Werneck, Executive · 2026-10-27. That single sentence is a more consequential disclosure than any tariff headline — CapEx return is now the ranking function for where the company grows, and Brazil is currently losing that contest on its own merits. The projects are also slipping. iron ore expansion at Miguel Burnier is in final integrated tests, but the ramp drags: “especially now with El Niño, I think this will delay things a bit” — Rafael Dorneles Japur, Executive · 2026-10-27. The roughly BRL 400 million benefit lands mostly in 2027, not 2026. Midlothian's maintenance shutdown will show up as idle cost in the fourth quarter, capping margin even as realized prices carry a $70 tailwind. And South America was the quarter's weakest geography on Argentine volumes, with management steering expectations to a recurring mid-teens margin — Peru-like, not Argentina-like.

What actually changed

Three things, none of them in the headline EBITDA. First, the buyback is now explicitly valuation-contingent: “while our shares are below the book value, we will see good opportunities to continue to invest” — Rafael Dorneles Japur, Executive · 2026-10-27, with shares cancelled rather than warehoused, and total distributions above 100% of nine-month free cash flow. Second, the balance-sheet policy is quietly up for revision — a gross-debt framework conceived at BRL 3.00–3.50 per dollar is being re-examined at BRL 5.30, which is a tell that leverage capacity will be redeployed rather than hoarded. Third, and least comfortable: in the strongest margin quarter in four years, the Brazil narrative contained no volume growth story at all. It was defense, cost, and a promise of detail to come on mining and competitiveness. The trap for a reader here is to treat the tariff debate as the thesis. The market already prices that debate across every Brazilian steel name. The unpriced item is the other one: a company that has stopped pretending Brazil earns its cost of capital, has started underwriting marginal dollars abroad, and has a 77%-North-America profit base it refuses to separate. December 4 is the next real catalyst.