The Machine Nobody Owns

On dollar power and the empire of interoperability

On 26 February 2022, the G7 governments froze something in the order of $300 billion belonging to the Russian Central Bank. One might have expected this sum, large enough to run a medium-sized country for a year, to vanish into some vault of geopolitical consequence, guarded by men with grim faces. Instead it stayed exactly where it had been, sitting in the ledgers of Euroclear, a Belgian clearing house whose previous claim to fame was settling Eurobond trades efficiently and without fuss. And in the administrative non-event of not being able to move lay a rather large revelation — that the international monetary system, which had spent decades cultivating airs of a neutral utility, as politically inert as the plumbing, turned out to have owners after all, and those owners had opinions.

A system had just demonstrated, in the most public way imaginable, that it could vanish your money by administrative fiat, and the response of the world's finance ministries was, on the whole, to keep their accounts open and say nothing. The Indian solution was the most elegant: buy Russian oil in rupees, keep the dollar accounts ticking over as normal, and call the whole arrangement a hedge rather than a defection — strategic ambiguity dressed up as bookkeeping. Everyone else's reasoning was less artful but equally durable. German manufacturers stayed because their supply chains had dollar finance running through them like damp through a wall; too deep to dig out without the structure coming down. The Gulf monarchies stayed because dollar liquidity is, in some functional sense, the thing their political systems are built on top of. Nobody was making a statement. They were doing what people do when the exits are all worse than the room they're in, which is nothing, conspicuously.

The conventional explanation for this stickiness is that America runs the plumbing and the world puts up with it because the alternative is worse — and there is truth in this. Dollar dominance rests on sanctions regimes, swap lines, legal jurisdictions and a structural toll that peripheral economies pay simply for the privilege of borrowing in a currency they cannot print.

But this explanation, true as far as it goes, gets the shape of the thing wrong. The political-economy tradition — Lenin's stages theory, Hilferding on finance capital, and in our own time the careful work of Panitch and Gindin, Harvey, Lapavitsas — tends to read the dollar order as a design, painting a picture of mechanisms that seem intentional, coherent and run by identifiable people in their own interest, rather like a successful family firm.

The trouble is that nobody seems to have been running it. The dollar order behaves like a machine that assembled itself out of spare parts lying around after other emergencies, and which gradually became too large and too central for anyone, including the people nominally in charge, to risk turning off.

Greta Krippner's account of American financialization makes the case well. Wall Street's great expansion did not spring from a master plan for permanent monetary supremacy. It came from the American state trying to solve a string of unglamorous problems — the inflation left over from Vietnam, the fiscal mess of the 1970s, the need to pull in foreign capital without the kind of domestic austerity no government could survive. The gates were opened for reasons that had nothing to do with empire. Yet finance bounded through the gates before it grew, complicated itself, and became too interconnected to fail. Eventually the Federal Reserve found itself the guarantor of an edifice it had not designed, at a scale nobody had budgeted for, serving purposes well beyond anything in its founding charter. Essentially, the machine got built and Washington moved in afterwards, the way a government might discover, some years after the fact, that it owns a railway.

Arrighi spotted the pattern decades ago. His great insight, though hardly foreign to generations of Byzantinists or China scholars, was that a great power, in the late stage of its dominance, tends to narrow its own room for manoeuvre with every exercise of that dominance, so that each use forecloses the next and the costs compound rather than cancel. What Arrighi left unspecified was the mechanism — how, exactly, a capability goes from latent to used to impossible to "unknow."

If the dollar order is an imperial project, the current period of financial coercion represents its fullest expression. If it is a machine that accumulated coercive potential gradually through improvisation rather than intent, the current period is something more unstable: the moment a capability was discovered, demonstrated publicly, and began consuming the trust on which its long-term efficacy depended. The second reading is the more frightening one because it means the damage isn't a policy that can be reversed. It's a disclosure, and disclosures don't go back in the bottle.

The capability had been sitting there for years, waiting for someone to work out the method. The man who worked it out was called Stuart Levey, and the story of how he did it is, in its way, the whole essay in miniature.

In September 2006, Levey, the Bush administration's first Under Secretary for Terrorism and Financial Intelligence, sanctioned a single Iranian bank, Bank Saderat, and then got on a plane. Over the following months he travelled to the boardrooms of the world's largest banks — he wasn't threatening them. He was educating them. Any bank that processed transactions for the sanctioned Iranian entities risked losing its own access to dollar clearing — the pipework, in effect, without which an internationally active bank simply cannot function.

Within four years Iran was unable to move dollars or euros through any bank that mattered, without a single soldier or warship involved. In other words, it was financially excommunicated. Levey had discovered that you could fight a war using nothing but other people's fear of losing their plumbing. The lesson was filed away, and in time it was scaled up to larger targets.

For now, the center holds. In March 2020, as Covid produced a dollar squeeze of startling speed, fourteen central banks rang the Federal Reserve for emergency liquidity, and hundreds of billions moved within days. The interesting question isn't why the Fed answered — it had every reason to. It's why all fourteen rang the same number, and why no other number would have done.

The People's Bank of China could not have played the part given it has no sovereign debt market deep enough to serve as global collateral, no government bonds anchoring the world's repo markets, and its payments system processes only a fraction of SWIFT's volume. When the machine panicked, it called the only number it had ever called — the architecture offered no other.

What has changed isn't the lynchpin's power. It's the weather around it. For thirty years after the Cold War the machine operated on a set of comfortable assumptions: that trade made conflict expensive for everyone, that supply chains were an efficiency problem rather than a security one, and that turbulence was a temporary glitch in an otherwise self-correcting system rather than a permanent feature of the furniture.

These assumptions have not so much eroded as collapsed, more or less all at once. Covid broke supply chains optimised for speed rather than survival. The Ukraine war turned wheat and gas into weapons. The Houthis added several weeks to the journey between Rotterdam and Shanghai simply by making the Red Sea unpleasant. Export controls on semiconductors turned chip fabrication into something guarded with the seriousness once reserved for enriched uranium. After thirty years off, scarcity and friction are back, and they've brought their luggage.

The machine still runs. But it runs now in weather it wasn't built for, tended by a power whose relationship to its own infrastructure looks less like ownership and more like an increasingly expensive and contested lease.

II. The Disobedient Appendages

Somewhere on the container route between Shenzhen and Istanbul, sometime in mid-2022, a perfectly ordinary shipment of semiconductors changed its mind about where it was going. The chips themselves had the usual cosmopolitan upbringing — designed in California, fabricated in Taiwan, packaged in Malaysia — and they arrived, as such things do, with paperwork. The paperwork said the cargo was bound for one place, while it was in fact bound for another. Somewhere along the way a compliance officer looked at the forms with the level of attention compliance officers are generally paid to apply, which is to say not enough, and the shipment cleared. Nobody lied, exactly. The truth simply declined to keep up with the paperwork, which is a different thing, and a much more common one.

This happened at scale. Investigators later traced nearly $4 billion in export-controlled chips into Russia after the invasion, moving through a network of more than 6,000 companies spread across dozens of jurisdictions. Japanese manufacturers turn up with embarrassing regularity in Russian customs data, despite Tokyo's enthusiastic participation in the sanctions regime — and Japan has since been implicated in the business of Nvidia chips finding their way into China. The geography of all this — Hong Kong, Turkey, the UAE, India, Singapore — has the smooth, well-oiled look of a mature commercial operation, not the nervous shuffle one associates with men doing something illegal. These were distributors, doing what distributors do, which is find a margin and follow it.

The Lapavitsas account of subordinate financialization — peripheral firms locked into dollar production chains, disciplined by lead firms, exposed to currency risks they can't hedge, dependent on credit they can't negotiate — is accurate as a description of the cage. What it doesn't quite predict is that the bird, on this particular occasion, simply flew through the bars. The machine's own appendages, from chip distributors to shippers and insurers, from cloud providers to law firms, carried on pursuing their own incentives with only the most glancing reference to what Washington wanted, and Washington discovered it had no lever long enough to reach them.

Even at the center, the obedience turns out to be patchier than advertised. BlackRock sits about as deep inside American financial power as it is possible to sit — during the pandemic it ran several of the Federal Reserve's own bond-buying programmes, and it holds significant stakes across most of the world's listed companies. And yet researchers combing through more than 24,000 of its proxy votes found it had quietly and consistently backed Communist Party committee structures inside Chinese state-owned enterprises, voting the opposite way to Vanguard and State Street on the same resolutions. BlackRock's business depends on assets under management; its access to the Chinese market is worth real money; and Washington's decoupling agenda would cost it some of that money. So it didn't decouple. There was, in the end, no button marked comply that the imperial apparatus could press.

Albert Hirschman's old framework (exit, voice and loyalty) explains this better than the political-economy tradition tends to. When exit is too costly, people don't leave and they don't quite stay either: they hedge, duplicate, build the workaround they're not yet committed to using. This is the mode the world has been in for some time now, and it started earlier than 2022 suggests — it started, more or less, with Levey.

Russia began dedollarising in earnest after Crimea in 2014: diversifying its reserves, settling more trade outside dollars, building SPFS as a homegrown alternative to SWIFT, launching the Mir card network. Between 2013 and 2021 the central bank halved its dollar holdings and moved into gold and renminbi, so that by mid-2021 gold made up 22% of reserves against the dollar's 16%. None of it was enough — half the total was still sitting in Western custody when the freeze landed, a miscalculation of heroic proportions — but the direction of travel was unmistakable, and each fresh demonstration of what the West could do produced a roughly proportionate effort to make sure it couldn't be done again.

China's own version of this accelerated after the 2018 trade war, and the 2022 freeze widened the audience considerably — India, the Gulf states, smaller economies across Asia and Africa all discovering, more or less simultaneously, that they too were the kind of country this could happen to.

The result has been a gold-buying spree, and it's worth being precise about what that spree actually represents, because it isn't a flight to safety in any ordinary sense. Central banks do not imagine that gold can anchor a modern payments system — it can't serve as collateral, can't settle a trade, can't do any of the things a reserve currency needs to do. What gold can do is sit in a vault that nobody in Washington has the keys to. Buying gold at this scale is a vote of no confidence, expressed in the only currency — literally — that cannot be frozen by someone else's keystroke. And the buying has not slowed down.

The economics of all this are not subtle given redundant supply chains cost more than efficient ones; stockpiles tie up capital that would otherwise be working; parallel payment systems duplicate infrastructure the world has already built and paid for once. Reshoring trades efficiency for survivability, and somebody — eventually everybody — picks up the bill. The world now forming is more inflationary, more capital-hungry and less productive than the one it's replacing. This is insurance, and insurance is expensive in exact proportion to how much you've come to distrust the thing you're insuring against.

When Donald Trump threatened 100% tariffs against the BRICS countries merely for building alternatives to the dollar system, he was trying to stop the conversation about the exit. In Hirschman's terms, he was going after voice rather than exit, and Hirschman's framework is fairly unforgiving on this point: suppressing the conversation when the exit door is genuinely blocked doesn't restore anyone's loyalty. It just drives the dissatisfaction underground, into forms that are quieter and much harder to see coming.

The Levey doctrine, taken to its logical end, has produced exactly the counter-infrastructure it was designed to prevent. This is the Arrighi mechanism again, now visible at the level of plumbing: the more often the coercive option gets used, the more it forecloses the cooperative arrangements that used to make coercion unnecessary in the first place.

III. The Interoperability Stack

Consider what geopolitics has done to a product roadmap. When Washington restricted the export of Nvidia's A100 chip to China, Nvidia built the A800 — the same chip, with its interconnect bandwidth quietly dialled down to fall just inside the new limit. When the A800 was restricted in turn, along came the H800, recalibrated to the next set of tolerances. Then another compliant variant. Then another. The world's most valuable semiconductor company now does a fair portion of its engineering not in dialogue with physics and customers, but with the Commerce Department's Office of Export Controls, a body whose expertise lies in neither. Export law has become, alongside transistor density and heat dissipation, a design spec for the hardware that runs the world's most advanced AI. This is simply how things are made now.

Meanwhile, in Eindhoven, one company, ASML, builds the only machines on earth capable of the extreme ultraviolet lithography needed for leading-edge chips. About fifty of these machines come off the line each year. Each costs hundreds of millions of dollars and represents, in effect, the accumulated output of a small civilization's worth of specialised engineering, none of it readily duplicated elsewhere. Under sustained American pressure, the Dutch government blocked their export to China — and Huawei's best AI chip still trails Nvidia's by a wide margin, and will go on trailing for as long as ASML remains a single point of failure and The Hague remains willing to take instructions about it. The gap between Huawei and Nvidia is not a technology gap. It's a policy outcome, achieved by exactly the method Levey pioneered on Iranian banks in 2006, now applied a few floors further up the building.

The political-economy tradition is good on the extraction running through all this — Lapavitsas in particular is meticulous about the interest-rate tribute the periphery pays the core, the balance-sheet hierarchies that let American firms borrow long on terms nobody else gets, the collateral chains that reproduce dollar dominance in every repo market on earth. None of this is wrong. But extraction is not the same thing as control. Control is the power to decide what plugs into what — which standards apply, who gets access, and at what level of capability they're allowed to operate. And increasingly that power runs not through the financial circuits Lapavitsas describes so well, but through the architecture itself.

Picture the whole thing as a building with several floors, each one load-bearing. The dollar system is the ground floor. Semiconductors are another floor up. Cloud computing — AWS, Azure, Google Cloud, between them running a large share of the world's business — is another, and its exposure to American jurisdiction has not yet been seriously tested but is not, by any means, theoretical. Maritime insurance is another given the Lloyd's market quietly underwrites global shipping while doubling, when called upon, as a sanctions enforcer. GPS, undersea cables, the legal jurisdictions that decide whether a dollar claim can actually be enforced — these are all floors of the same building, and its foundations remain disproportionately American.

What's happening now is that several floors are being shaken at once — sanctions, chip controls, cloud access, satellites, cables — and the cumulative effect of the shaking has turned out to matter more than any single tremor. This is the same process Levey started in 2006 with one bank and a plane ticket, now propagated all the way up the building until the building itself has become visible as a building, rather than as the weather.

Infrastructure works by being invisible. Roads function as such because nobody using them thinks about who built them or under what terms. The internet felt apolitical for decades because almost nobody using it had any reason to think about the cables and jurisdictions underneath. The dollar order worked the same way — people treated it as weather, a fact of economic life, rather than a system built and maintained by people with addresses.

The Iran sanctions were the first crack in that illusion — proof, for anyone paying attention, that there was a man twitching behind the curtain. The 2022 freeze was the announcement, made simultaneously to every finance ministry on earth. And what's underway now is the audit: thousands of institutions, in dozens of countries, working out which floors of the building they're standing on, which ones they could survive losing, and what it would cost to build a staircase to somewhere else. Europe's reaction to the proposed AI Diffusion Rule in 2025 — when European firms abruptly discovered they were inside the hierarchy rather than merely next door to it — was this same realisation, compressed into a single news cycle, and it produced the same response everywhere it happens, i.e. a sudden, urgent interest in redundancy.

IV. China and the Art of Not Winning

The assumption running through most Western strategic commentary is that Beijing wants Washington's job — to issue world money, run the reserve system, set the standards everyone else has to follow. The evidence for this is thin and overall the reverse looks true: China has looked at the job and doesn't want it.

Lapavitsas gets closer than most here given he correctly identifies the blockage, seeing that the same domestic arrangements that powered China's industrial rise are precisely the ones that rule out reserve-currency status, because world money requires open capital markets, independent legal protections, and a surrender of monetary control to the preferences of foreign creditors, none of which Xi's China is remotely interested in offering. But even Lapavitsas writes as though China were straining against a constraint it hasn't yet found a way around, a superpower waiting for the technology of escape. The Chinese leadership doesn't read like an institution waiting for anything. It reads like one that has already chosen.

World money requires the kind of credibility that comes only from tying your own hands — deep markets, legal constraints on the state, monetary policy run partly for the benefit of people who aren't you. Xi's programme has spent a decade moving in exactly the opposite direction, embracing tighter party control of finance, tighter capital controls, more direct state influence over who gets capital and on what terms. The very controls that protect the Party from the disciplines of global finance are the controls that would have to go for the renminbi to become global finance. Beijing knows this. It isn't trying to solve it. It has simply decided which of the two things it would rather have.

What China is building instead is insulation; infrastructure that lowers the cost of operating inside the existing system while widening the range of conditions under which China could survive outside it. CIPS processed something like 175 trillion yuan, roughly $26 trillion, in 2024, with digital yuan functionality increasingly built into the cross-border plumbing. Project mBridge, linking China, Hong Kong, Thailand, the UAE and Saudi Arabia through a shared digital-currency platform, grew rapidly until 2024, when the Bank for International Settlements withdrew, citing sanctions-evasion concerns with a candour that rather gave the game away about how seriously the project was being taken. Yuan settlement for energy has expanded with Russia, Iran, and a growing list of Gulf producers. None of this adds up to a rival system. What it adds up to is China having fewer single points of failure than it used to.

And this creates its own trap, which loops straight back to Arrighi. China builds parallel infrastructure to reduce its exposure to Western coercion. The infrastructure, once built, becomes available to sanctioned states looking to evade that same coercion. Its use for that purpose confirms Western suspicions about why China built it. The confirmation accelerates the pressure that motivated the insulation in the first place. Round and round — a cycle neither side designed and neither side controls, each move making the next one look reasonable to whoever's making it.

China accounts for nearly 30% of global manufacturing value added and around 15% of world exports. And yet its firms still struggle to borrow long-term abroad in their own currency, its government bonds don't anchor anyone's collateral, and its payment networks process a fraction of what the dollar system handles without trying. The productive superpower of the age remains boxed in by a monetary hierarchy it didn't build and can't yet leave — not for want of will, but because leaving would mean giving up the domestic controls Beijing has already decided are not on the table.

So, why exaggerate this as succession maneuvers? The dollar isn't collapsing, because nothing else has the scale or the depth, and the countries most eager to see it fall — Russia, Iran, the usual list — don't have the financial mass to make it happen. But stable continuation is getting harder too, because every freeze, every export control, every fresh disclosure peels back another layer of the stack and reveals it as owned rather than neutral — and neutrality, once visibly gone, can't be argued back into existence.

What's left is attrition: selective interoperability, rising redundancy, accumulating friction, and a growing share of the world's capital going not into production but into hedges against systems nobody can quite leave and nobody quite trusts any more. Call it a tax — one first levied, in effect, the day a man got on a plane to Bahrain in 2006 to demonstrate what the plumbing could be made to do.

The twentieth century's globalisation machine promised that integration meant enrichment, and for a few decades it delivered enough of that promise to keep the question from being asked too closely. The twenty-first is discovering that the same integration is a liability in exact proportion to its depth, and that discovering this makes the liability worse. This is not the impersonal logic of capital working itself out, however much the political-economy tradition might prefer that story. It's the result of a specific sequence of choices — improvised, then institutionalised, then overextended, and now impossible to walk back — about what the machine could be made to do, and how often, before the doing became the problem.

Every reserve manager buying gold, every state laying its own payment rails, every chip company designing to the tolerances of an export-control office rather than the laws of physics, is adding friction to a system whose entire value lay in having none. And that quality, once spent, doesn't come back by spending more of whatever spent it.

Arrighi's mechanism is simple enough. Visibility breeds insulation, and insulation is the one thing the machine was never built to absorb. Washington didn't plan any of this. It found a capability, used it, scaled it, and now lives in the world that using it made. The machine still runs but the world it was built for doesn't exist any more.