SWIFT gpi

SWIFT gpi is the correspondent system’s answer to its own reputation. Cross-border payments were slow, opaque, and untraceable; gpi, rolled out across the SWIFT network from 2017, attacked the opacity directly. Its core device is the UETR, a unique end-to-end transaction reference that travels with the payment through every intermediary bank, so that sender and beneficiary banks can see where a payment is, which hop holds it, and what each hop deducted. Member banks additionally commit to same-day processing where operating hours allow, to fee and FX transparency, and to confirming final credit.

Within its terms, it worked. SWIFT’s published figures have nearly 60% of gpi payments reaching the end beneficiary within 30 minutes, almost 100% within 24 hours, and over $300 billion a day flowing through the service. The tracking data had a second-order benefit: it made delay measurable, and therefore attributable, which changed behaviour along correspondent chains that had never faced a stopwatch.

What gpi does not do is change the machine. SWIFT carries messages; the money settles separately, across nostro accounts, hop by hop, exactly as before. Every structural property of the correspondent model survives intact: pre-funded liquidity, per-hop compliance checks, deducted fees, local cut-offs, and a moment of finality that arrives only when the last domestic leg credits the beneficiary. The 24-hour tail in SWIFT’s own statistics is that structure showing through the service layer.

The fair summary is that gpi made the correspondent system honest about itself: visible, measured, and considerably faster at the median. The gap it left, settlement that is still sequential, still pre-funded, and still bounded by the slowest hop, is the gap the newer families of rails, from interlinked instant payment systems to shared-ledger settlement, exist to close.