For years, there was a fairly obvious workaround to the EU's carbon pricing rules: if domestic production got expensive because of the Emissions Trading System, just import the goods from somewhere with looser rules instead. CBAM (the Carbon Border Adjustment Mechanism) exists specifically to close that loophole, by making imported goods pay a carbon price equivalent to what EU producers already pay under the ETS.

Which Goods Are Covered

The mechanism started narrow, focused on six sectors chosen for their carbon intensity and their exposure to this kind of carbon leakage: steel, aluminium, cement, fertilizers, hydrogen, and electricity.

How the Mechanism Actually Works

Importers of covered goods have three obligations to keep up with:

• Report the embedded CO₂ emissions in what they're importing, on a quarterly basis
• Purchase CBAM certificates, priced in line with EU ETS carbon prices
• Surrender certificates annually, matching the verified emissions for that period

There's a relief valve built in: if a producer can show they've already paid a carbon price in their country of origin, the CBAM obligation shrinks accordingly. The mechanism is designed to top up, not double-charge.

The Data Problem Underneath It

None of this works without verified emissions data, and that's where things get complicated for importers. They need real numbers from suppliers, primarily Scope 1 emissions tied to the actual production process. That requirement creates a direct, unavoidable link between a company's supply chain GHG calculations and its import compliance obligations. The two used to live in separate departments; now they can't.

Where Things Stand

CBAM has been in its transitional phase since October 2023, running on quarterly reporting obligations without the financial teeth yet. That changes in 2026, when the full regime (certificate purchases included) comes into force.

At BUME, the same supplier emissions data we capture for Scope 3 reporting feeds directly into CBAM compliance too, since underneath the different acronyms, it's largely the same underlying data problem.