Saudi Arabia steps away from the payments project Washington feared, and the dollar data refuses to cooperate

Saudi Arabia's central bank, the operator of the world's largest crude export balance sheet and the Gulf's largest economy, has confirmed that it is no longer part of mBridge, the digital currency platform built with the People's Bank of China to move money across borders without touching the dollar-based correspondent banking system. The Saudi Central Bank told the Financial Times it completed its planned proof-of-concept work in May 2025 and stopped participating. The statement was published on Saturday and has been picked up across the region since.

The news arrived in most of the world's feeds as a single line, usually framed as Riyadh abandoning a dollar alternative. The record behind it is more specific, and more interesting. mBridge was the payments project that Washington worried about most, because it was the one designed to make sanctions enforcement genuinely harder. Its membership has now lost the international institution that built it, and the Gulf state whose participation gave it economic weight. Understanding what that means requires separating three things that are being blurred together: what the platform actually achieved, why participants keep leaving it, and what the hard data says about the dollar's position, which is not what the de-dollarization narrative predicts.

What mBridge actually was

The primary record sits on the Bank for International Settlements' own project pages. mBridge grew out of experiments begun in 2021 between the BIS Innovation Hub's Hong Kong centre, the Hong Kong Monetary Authority, the Bank of Thailand, the Digital Currency Institute of the People's Bank of China and the Central Bank of the UAE. Its purpose was direct settlement: participating banks pay each other in their own central bank digital currencies across a shared ledger, collapsing the chain of correspondent banks, delays and fees that moves money between jurisdictions today.

The 2022 pilot was real money on real rails. Twenty commercial banks across the four jurisdictions executed 164 payment and foreign-exchange transactions worth over $22 million in six weeks, settled directly on the platform, according to the BIS's own release. In June 2024 the project reached what engineers call minimum viable product, the Saudi Central Bank joined as a full participant the same month, and more than thirty institutions had signed on as observers, including the European Central Bank, the IMF, the World Bank and the Federal Reserve Bank of New York's innovation centre.

That observer list is the detail worth holding onto. The New York Fed was watching the platform as a peer. Western central banks were interested in the technology, and wary of the same platform's potential at the same time.

The departures, in order

The BIS pulled out in October 2024, announcing it was handing the project over to the partner central banks. The timing was noted at the time: the withdrawal came days after a report that the BIS was weighing mBridge's future amid concern in Washington that a payments network connecting China, the Gulf and Southeast Asia on a shared ledger could be used to evade sanctions. The BIS's public framing was administrative, an innovation hub concluding its phase. The effect was not: the Western institution that had anchored the project's credibility stepped back to the sidelines, and the platform passed to the founding central banks, with China's Digital Currency Institute the largest of them.

Saudi Arabia joined after that debate, in the same month the MVP launched, and its exit is now confirmed. What the bank has said publicly is narrow: its planned trial finished in May 2025. Some reporting attributes to people familiar with the matter the view that Saudi officials found the platform unusable for real activity at scale; that characterization has not been confirmed by the bank, and no participant has published transaction volumes since the BIS left, so the platform's operating scale cannot be verified from public data either way.

What can be verified is the direction of travel. The BIS's own project page now lists mBridge as concluded. The remaining participants are the Digital Currency Institute of the People's Bank of China, the Hong Kong Monetary Authority, the Bank of Thailand and the UAE central bank. A project that spent three years recruiting international observers has ended up with a smaller membership than it had at its pilot.

Why the exits matter more than the launch

Payments infrastructure is a scale business. A cross-border settlement platform has value exactly in proportion to how many banks with real trade flows use it, which is why the correspondent banking system it aims to replace, whatever its inefficiencies, has held: everyone is already on it. mBridge never crossed the volume threshold that would make it an alternative rather than an experiment, and the two departures that carried the most weight are the institution that supplied technical legitimacy and the member whose oil trade supplied potential volume.

The Washington angle is inseparable from this. mBridge was explicitly designed so that settlement happens directly between central banks on a shared ledger, outside the Swiss-correspondent-New York clearing chain through which dollar enforcement operates. Reporting around the BIS withdrawal described concern that the platform could become exactly that bypass. Whatever the stated reasons for each exit, the sequencing leaves the bypass unsolved and unstaffed.

Now put the dollar data next to it

The temptation with any story like this is to file it under the decline of the dollar. The measured record refuses to cooperate, and the refusal is itself the finding.

The IMF's own quarterly survey of currency composition in official reserves, the dataset central banks report into, put the dollar's share of allocated foreign exchange reserves at 57.13 percent in the first quarter of 2026, up from 56.42 percent in the fourth quarter of 2025. Total allocated reserves stand near $13 trillion. The dollar's share is lower than it was two decades ago, and the direction of the last few quarters is up, not down.

Central bank behaviour meanwhile shows real diversification, but concentrated in one asset. Net official-sector gold buying has run at roughly a thousand tonnes a year for most of the past decade, sixteen consecutive years of net purchases, with the buying wave accelerating sharply after the freezing of Russia's reserves in 2022. Our investigation into what the world's central banks have actually done with gold since 2010 traces that record in full, and the conclusion there is the same one the mBridge exit points toward: the behaviour is hedging at the margin, not abandonment.

The two datasets reconcile only if you read them as different answers to different questions. Reserve composition measures where the stock of official wealth sits, and there the dollar's network effects, Treasury market depth and the absence of a comparable alternative keep the share rising on valuation effects alone. Payments infrastructure measures how money moves, and there the search for routes outside Western jurisdiction is genuine, underfunded and so far unsolved. Saudi Arabia leaving mBridge weakens the second front and leaves the first untouched. The gold buying is the visible expression of the same hedge: buy an asset nobody can freeze, while keeping the dollar flows that price your oil.

What changes the picture

Three things would move this story from a membership footnote to something bigger, and all three are checkable.

The first is transparency from the remaining members. mBridge has published no transaction volumes since mid-2024. If the platform processes meaningful real settlement between the UAE, Thailand, Hong Kong and the Chinese mainland, a single published figure would establish that. The absence of one, three years into a minimum viable product, is evidence of its own.

The second is the successor experiment. The BIS's Project Agora, built with seven major central banks and dozens of commercial institutions on tokenized deposits within existing regulatory frameworks, conducted real-value transactions in July 2026 with 28 institutions across three continents. If wholesale tokenized settlement reaches production in Western frameworks while mBridge stalls, the technological question the platform raised gets answered without it.

The third is the next quarterly reserve survey. If the dollar's share reverses from 57.13 percent while Gulf and Asian central banks keep adding gold, the diversification story gets teeth. If it keeps rising, the mBridge exit will end up as a data point about the difficulty of building financial infrastructure, not about the dollar's decline.

The Saudi decision is best read in exactly that register. Riyadh spent a year trialling the most credible sanctions-resistant settlement platform anyone has built, and its bank now says the trial is finished. The project Washington feared is smaller than it was. The dollar the project was meant to hedge against ended the quarter with a larger share of world reserves than it started with. Both of those facts are true at once, and the second is the one the data will keep scoring.

Updated 20 September 2026. The reserve-composition figures come from the IMF's 30 June 2026 data brief; mBridge project details come from the BIS's own project pages and media releases; the Saudi Central Bank's confirmation was reported by the Financial Times.