top of page
Search

Withdrawal Is Not an Exit

  • Writer: Aişe Gül Akkoyun
    Aişe Gül Akkoyun
  • Oct 31, 2025
  • 4 min read

Updated: 3 days ago


Romania leaves the Energy Charter Treaty, and the twenty-year clause that makes leaving take a very long time.

Background

Romania's Government has approved a draft law authorising withdrawal from the Energy Charter Treaty, citing inconsistency with the European Green Deal, the Paris Agreement, and the Court of Justice's judgment in Komstroy (C-741/19).

The decision follows the European Commission's recommendation for a coordinated exit and places Romania alongside a list of Member States that have already left or announced departure: Italy, which withdrew earliest and whose exit took effect in 2016, together with France, Germany, the Netherlands, Spain, Poland and others. The European Union itself announced its withdrawal in June 2024, and the EU and twenty-six Member States declared their intention to sign an agreement disapplying the Treaty's investor-State arbitration provisions to intra-EU disputes, giving effect to Komstroy.

Romania had separately communicated to the Energy Charter Secretariat in September 2025 that it would not provisionally apply the Treaty's modernisation amendments, on the basis that doing so would be inconsistent with its national legislation.

So the direction of travel is clear. What is less often stated clearly is how long the journey takes.

The arithmetic of leaving

The Energy Charter Treaty has two provisions that govern departure, and they operate in sequence.

Article 47(2) provides that withdrawal takes effect one year after the Secretariat receives written notice. So a decision taken in late 2025 does not remove a State from the Treaty in 2025. It starts a clock.

Article 47(3) is the one that matters. Investments made before the date withdrawal takes effect remain protected by the Treaty for a further twenty years. That is the sunset clause, and it is unusually long. Most bilateral investment treaties use ten or fifteen years.

Put those together for a State that notifies in 2026. Withdrawal takes effect in 2027. Investments existing at that date remain protected until 2047. A tribunal constituted in the 2040s could be applying a treaty the respondent State left two decades earlier.

This is not a drafting oversight. Sunset clauses exist to prevent States from stripping protection retroactively from investors who relied on it. The policy logic is sound. The practical consequence is that treaty withdrawal is a slow-acting instrument, and any political claim that leaving the ECT ends exposure to ECT claims is simply wrong for a generation.

The intra-EU complication

Running alongside the sunset problem is a distinct one. Komstroy held that the Treaty's arbitration provision does not apply to disputes between an investor from one Member State and another Member State.

The difficulty is that arbitral tribunals are not bound by the Court of Justice, and several have proceeded with intra-EU ECT claims notwithstanding Komstroy, reasoning that their jurisdiction derives from the Treaty rather than from EU law. The resulting awards then face enforcement problems inside the EU and rather fewer problems outside it.

So there are effectively two legal orders operating on the same treaty text and reaching different answers. For a State in Romania's position, withdrawal addresses the extra-EU exposure over twenty years and does very little about the intra-EU question, which is being fought out in parallel between tribunals and courts.

Why this is a good dataset

ECT withdrawal is one of the rare phenomena in international investment law that is fully observable, precisely dated and repeated across many States. That combination is unusual enough to be worth exploiting.

Every withdrawal generates the same set of recorded facts: the date of the domestic decision, the date of notification to the Secretariat, the effective date one year later, the sunset expiry twenty years after that, and the stated justification. Nothing is confidential. Everything is on the public record.

That makes it possible to ask questions that are usually unanswerable:

  • Sequencing: does withdrawal follow adverse awards against the State, or precede them? Spain and Italy make an instructive contrast here.

  • Justification drift: the stated reasons have shifted over a decade from budgetary and sovereignty arguments to climate and EU-law arguments. That shift is traceable in the text of the domestic instruments.

  • Coordination: how tightly clustered are notification dates, and does clustering follow Commission recommendations or domestic political cycles?

  • Residual exposure: how many claims are filed against a State after its withdrawal takes effect but within the sunset period, and does the rate decline over the twenty years or stay flat?

The last question is the one nobody can currently answer, because not enough time has passed. Italy is the only Member State far enough into its sunset period to provide meaningful data, and one State is not a sample.

The open question

The ECT withdrawals are usually described as States reclaiming regulatory space for the energy transition. That is accurate as a statement of intent.

Whether it is accurate as a description of effect is a different matter, and one that will not be settled for twenty years. In the meantime, every withdrawing State remains bound for existing investments, tribunals continue to constitute themselves under a treaty that a growing number of parties have left, and the legal position of intra-EU claims depends on which forum is asked.

The interesting question for research is therefore not whether States are leaving. It is whether leaving does anything measurable, and on what timescale. Treaty exit is a natural experiment with a twenty-year lag, and almost nobody is set up to measure it.

Sources

Comments

bottom of page