The Treaty That Outlived Three Wars

The Indus basin was partitioned in 1947 in the worst possible way: the headworks ended up in India and the irrigated land that depended on them in Pakistan. Canals that had been built as one system now crossed an international border, and in April 1948 India briefly cut off supplies to canals serving Pakistani Punjab. That early episode set the terms of everything after it.

Nine years of World Bank mediation followed. The treaty signed in 1960 is unusual, and understanding why takes one shift of expectation.

Most river agreements try to share. They set a total, allocate percentages, build a joint commission with real authority and require the parties to agree repeatedly about operations. The Indus Waters Treaty does close to the opposite. It divides the rivers. India received the three eastern rivers, the Ravi, the Beas and the Sutlej, for unrestricted use. Pakistan received the three western ones, the Indus, the Jhelum and the Chenab, with India's use of them limited largely to run-of-the-river hydropower, which passes water through turbines without impounding much of it, and to specified small amounts of storage and irrigation.

The design principle is worth stating plainly: it minimised the number of decisions the two states would have to take together.

That is why it lasted. India and Pakistan fought wars in 1965 and 1971, fought at Kargil in 1999, and came close on several other occasions, and through all of it the Permanent Indus Commission met and the water flowed. Nobody had to trust anybody. Each side operated its own rivers, and the disputes that did arise were technical arguments about specific plants, settled by neutral experts and arbitration panels, over the gates and spillway design of structures like Baglihar and Kishenganga.

For sixty-five years this was cited everywhere as the proof that water agreements can survive hostility. Then it changed.

On 23 April 2025, the day after the attack at Pahalgam, India announced it was holding the treaty in abeyance until Pakistan credibly abandoned support for cross-border terrorism. In practice this meant suspending hydrological data sharing, suspending meetings of the commission, and disengaging from the treaty's dispute machinery. The text contains no suspension clause and no exit, so abeyance is a position rather than a procedure.

Pakistan went to arbitration. On 31 August 2026 the Court of Arbitration in The Hague held that the treaty remains fully in force, that India has no unilateral right to suspend or terminate it, and that the grounds offered did not justify abeyance. India rejects the tribunal as improperly constituted and treats its awards as void.

Now look at what each side can actually do, because this is where the lesson lives.

A tribunal can declare an obligation and cannot operate a gate. India, for its part, cannot withhold large volumes from the western rivers in the near term even if it wanted to, for a mundane engineering reason: storage of that kind requires large dams, and large dams require a decade and enormous capital. What India can do immediately is stop sharing data, stop attending meetings, flush reservoirs at times of its choosing, and accelerate projects on the Chenab, including a link tunnel into the Beas basin that went out to tender in May 2026.

So the dispute has moved to where action is cheap. Data, meetings, tenders and tunnels are the instruments, because they are the ones available on the timescale of a political decision. The rivers themselves are moving roughly as they did.

Two conclusions, and they pull in different directions.

The first is admiring. A treaty that asks enemies to cooperate as little as possible will outlast one that requires them to cooperate well, which is a design lesson more river agreements should take.

The second is sobering. A treaty with no exit survives everything until one party decides it is no longer bound, at which point there is nothing in the text to invoke and no bailiff to send. Sixty-five years of compliance was a very good run, and it was never the same thing as enforceability.