The EU's Carbon Border Adjustment Mechanism entered its definitive phase on January 1, 2026 — meaning importers now face actual certificate purchase obligations, not just reporting requirements. European Commission guidance published August 14 confirms the transition is complete: the mechanism now covers cement, iron and steel, aluminium, fertilisers, electricity, and hydrogen, with third-party verification requirements and mandatory compliance cycles in place.
The macro-level debate about CBAM's trade effects has been running for years. What's becoming clearer now, as firm-level data accumulates, is that the competitiveness question was never really a country-level story. It's a plant-level one.
National Averages Are the Wrong Unit of Analysis
The instinct in trade policy is to model CBAM exposure by country: take a nation's export volumes, apply average emission intensities, multiply by the ETS price, and call it a day. That approach produces numbers that look alarming in aggregate and miss almost everything that matters.
A Sandbag report on India published August 4 illustrates the problem precisely. In a business-as-usual scenario using national average emission intensities, India faces roughly €762 million in CBAM fees by 2034 — a figure that has circulated widely and generated predictable headlines about CBAM as a trade weapon against developing economies. But the report argues that applying national averages to India's steel sector "makes little sense in a country with such a heterogeneous steelmaking industry."
Once you disaggregate, the picture shifts. India exported approximately 4 million tonnes of CBAM-covered steel in 2025, mostly flat products. The country has sufficient low-emission production capacity — using iron produced from natural gas — to cover its flat steel export volumes. The constraint is not capacity; it's geography and the newly introduced anti-circumvention rules, which prevent factories with multiple production lines from reporting emissions only from their cleanest process. ArcelorMittal's NS Hazira plant, the report finds, clears both hurdles: right location, right technology mix, sufficient capacity. That single plant's competitive position matters more than India's national average emission intensity.
The long steel picture is less favorable, because those exports compete against EU scrap-based production that will face lower carbon costs. But the point stands: CBAM's competitiveness impact is a firm-by-firm calculation, not a national one.
Kazakhstan Shows What Happens When the Rent Leaves the Country
A Terawatt Times Institute analysis of Kazakhstan published August 8 makes the same disaggregation argument from a different angle, and arrives at a more uncomfortable conclusion.
Kazakhstan's aggregate CBAM exposure, calculated bottom-up from actual physical tonnes rather than trade-share extrapolation, is approximately 1.49 million tonnes of CO₂ per year — roughly an order of magnitude below what macro-level estimates implied. That exposure falls almost entirely on four enterprises: ferrochrome producer Kazchrome, aluminium smelter KAS (both part of Eurasian Resources Group), the Temirtau steelworks Qarmet, and fertiliser producer KazAzot.
The divergence between those four cases is sharp. Kazchrome holds a strong competitive position and can capture what the report calls a "data dividend" — the advantage that comes from having granular, verified emissions data that lets you report actual intensities well below sector defaults. The EU's default emission factors are set conservatively high; producers who invest in monitoring and verification can undercut them and pay less. That's worth roughly €74 per tonne to Kazchrome, according to the analysis.
Qarmet is a different story. Its steel emission intensity, while below the sector default, still leaves CBAM costs above the level at which comparable exporters have historically exited EU markets. The obvious decarbonisation route — switching to natural gas-based production — is blocked by pipeline constraints that cannot be resolved before 2030. The plant is the economic anchor of Temirtau, a single-industry city, so the employment and district-heating consequences extend well beyond the plant's balance sheet.
The fiscal point the report makes is the one that should concentrate minds in Astana: Kazakhstan has no functioning domestic carbon price. Under CBAM's Article 9, a domestic carbon price is deductible against the EU certificate obligation. Without one, the full carbon rent flows to Brussels rather than staying in Kazakhstan. The mechanism is, in that sense, exactly what the FT called it — a tariff by another name — and the revenue destination is determined by whether the exporting country has bothered to price carbon itself.
The Design Choice That Determines Everything
The underlying architecture of CBAM — mass-based rather than rate-based — is what makes the firm-level granularity matter so much. A mass-based system prices total embodied emissions, which means every tonne of CO₂ in the product carries a cost. A rate-based alternative would price emissions intensity relative to a benchmark, which creates implicit subsidies for producers who are merely less bad than average and weakens the signal for producers who are already clean.
The current design rewards early movers on decarbonisation and punishes producers who rely on national-average reporting. That's the right incentive structure, even if it creates transition pain for places like Temirtau where the decarbonisation pathway is physically constrained by infrastructure that won't exist until the end of the decade.
Watch for the European Commission's 2028 downstream expansion proposal — whether it extends CBAM to manufactured products that contain CBAM-covered inputs. That's where the competitiveness math gets genuinely complicated, because the firm-level data infrastructure required to track embodied emissions through supply chains doesn't yet exist at scale. The Hazira plant can document its own emissions. Tracking those emissions through to the finished goods that incorporate its steel is a different problem entirely.
