How to Run an Empire Without Being There
An empire is, among other things, a question about logistics. If you want to run a domain larger than you can walk in a morning, you have to solve one problem: how much of yourself do you have to send there to keep it yours? The answer has been trending down for five hundred years.
Watching the petrodollar stress and the Iran war reconfigure the Gulf over the last few weeks, a pattern became visible that is not really about 2026 or about currency. It is about how much physical presence an imperial system needs to collect revenue. The trend line is worth staring at directly.
Stage 1 β Bodies
The first imperial strategy is the heaviest. You send people. You plant them, and you keep them there.
Daron Acemoglu, Simon Johnson, and James Robinson β who received the 2024 Nobel for this work β distinguished two versions. In settler colonies (the United States, Canada, Australia, New Zealand), European migrants arrived in large enough numbers to replace the existing society with their own, building institutions that resembled the homeland. In extractive colonies (Spanish Peru, Belgian Congo), a small number of metropolitan officials governed coercively over a much larger indigenous labor force, with institutions designed to move wealth outward.
Even at its most indirect, Stage 1 is body-heavy. The British Raj ruled roughly two-fifths of the subcontinent through more than 500 "princely states" whose internal affairs remained with local rulers β the textbook case of "indirect rule." But it still required residents, garrisons, a British Indian Army hundreds of thousands strong, and a cabinet-level Secretary of State for India in London. You have to ship, house, feed, and bury people. That sets a floor on what empire costs.
Stage 2 β Structures
The Ottoman approach was lighter. It delegated.
The millet system organised the empire's non-Muslim populations into self-governing religious communities β Greek Orthodox, Armenian Apostolic, Jewish β each with its own laws, courts, and leaders who were responsible to Istanbul for tax collection and internal order. The empire did not administer these communities. It certified who administered them, and collected revenue via the religious hierarchy.
The iltizam was the revenue mechanism generalised. From the fifteenth century until 1856, the Ottoman state auctioned the right to collect taxes in a given region to the highest bidder. The winning mΓΌltazim paid a fixed sum upfront and kept what he could extract β sometimes up to five times the reservation amount. Revenue collection was effectively privatised. The state never touched the taxpayer directly.
At the edges of the empire sat the tributary states: Wallachia, Moldavia, Crimea, Transylvania, Ragusa, the Barbary regencies. Annual tribute, symbolic submission, internal autonomy. The Ottoman footprint in Bucharest might be one governor and a handful of janissaries, plus an implicit guarantee that a larger army could arrive if tribute stopped.
This is a different shape of empire. Not bodies on the ground but networks of local elites, bound together by revenue flows and the expectation of consequences if those flows stop. Local autonomy becomes the mechanism of control. The trade-off surfaced in the nineteenth century, when the capitulations β extraterritorial commercial rights granted to European merchants β gradually became the instrument by which European powers dismantled the empire from within.
Stage 3 β Dollar and security umbrella
The modern American version is lighter still.
Writing in 1953, the historians John Gallagher and Ronald Robinson observed that Britain in the Victorian era had preferred informal empire β control through finance, trade agreements, and local collaborators β to formal annexation. Direct rule was the exception; influence was the rule. This is the template the United States refined after 1945.
Two mechanisms run in parallel. The security umbrella: roughly 750 US military bases across 80 countries, around 173,000 troops stationed permanently abroad, of whom some 54,000 are in Japan, 34,000 in Germany, and 26,000 in South Korea. And the dollar system: after the 1971 gold suspension, foreign central banks had only one vehicle deep enough to hold their trade surpluses β US Treasuries. Michael Hudson called this the "Treasury Bill Standard" in 1972, and argued it converted what looked like an American weakness, permanent deficits, into an imperial revenue mechanism. Each dollar held abroad is, in effect, a zero-coupon loan to Washington.
In June 1974, Kissinger and Prince Fahd made the linkage explicit. The United States would guarantee Saudi oil fields; Saudi Arabia would price its oil in dollars and recycle the surpluses into Treasury bonds. Security for financial alignment; financial alignment in exchange for security. The Ottoman logic, ported to continental scale.
The footprint is still real. Seven hundred and fifty bases is not nothing. But relative to the volume of revenue extracted β the "exorbitant privilege" of borrowing cheaply in the currency you issue β the physical presence is smaller than any empire that came before. You do not need a viceroy in Tokyo. You need a Federal Reserve policy rate and a Seventh Fleet.
Stage 4 β What's emerging
Which brings us to now. The next stage is visible, but it is not settled. At least three candidates are under construction simultaneously.
The Chinese version is closest to the Ottoman template. The Belt and Road Initiative runs through infrastructure loans β ports, railways, power plants β that bind recipient countries to Chinese financial and logistical networks. The mBridge platform, a central-bank digital-currency network connecting China, Thailand, the UAE, Hong Kong, and Saudi Arabia, has cleared more than $55 billion in cross-border settlement and is now run by its participating central banks without BIS involvement. It is a modern tributary system, with digital-currency rails instead of tax farmers.
The American version is about the stack. Project Stargate β the $500 billion US AI infrastructure buildout backed by OpenAI, Microsoft, Oracle, Nvidia, SoftBank, and the UAE's MGX β is a bet that the rest of the world will run on American compute because there is no alternative. The United States currently hosts roughly 75% of global AI supercomputer performance; China 15%; everyone else 10%. The 2018 CLOUD Act lets US authorities subpoena data held by US-based firms anywhere on earth β a modern version of the Ottoman capitulations, running in reverse and written into statute.
The European version is sovereignty-as-defence. Only about 4% of global cloud infrastructure is European-owned; the EU depends on non-EU providers for more than 80% of its digital stack. Gaia-X, the Franco-German cloud initiative, is the attempted revolt. Forty-four countries now run "state-backed supercomputing clusters." Most of them are small.
The three-candidate frame is cleaner than the reality. Stage 4 is characterised by overlap, not exclusivity. The UAE is hosting OpenAI's first international Stargate cluster β a one-gigawatt facility in Abu Dhabi under the USβUAE Acceleration Partnership β while remaining a participating central bank in mBridge. Saudi Arabia has launched HUMAIN, a sovereign-AI programme with a stated goal of independence from both American and Chinese offerings, even as Huawei Cloud anchors a large part of its public digital infrastructure. South Korea in June 2025 committed $75 billion to what its lead strategist called "full-stack" sovereignty, framed explicitly as escape from "the neo-imperialism" of the USβChina race. Nick Srnicek calls this pattern omnialignment: countries partner with multiple great powers simultaneously because the Cold War's mutually exclusive blocs no longer match the shape of the technology.
What these candidates share is structure. Each moves the imperial apparatus one layer further into abstraction β from currency (Stage 3) to the protocols that decisions flow through. CBDCs can encode conditionality directly into money: this dollar cannot be spent on sanctioned items; this payment reverses if terms are violated. AI models whose weights and compute sit under one jurisdiction mediate the decisions of every other jurisdiction that uses them. Cloud platforms that host your data can be subpoenaed extra-territorially. The empire becomes the substrate of decision-making itself.
The arc
Four stages, one direction. From bodies, to institutions, to currencies, to protocols. From one dominant hegemon at a time to several overlapping stacks competing simultaneously for the same customers. From human administrators to autonomous systems that execute imperial logic without anyone having to issue a command.
It is a mistake to read this trajectory as emancipation. Empire does not get lighter because the people inside it become freer. Empire gets lighter because it learns to run on thinner wires. The logistics improve. The visibility decreases. The question the Iran war made visible in the Gulf a few weeks ago β who do we trust to still be there tomorrow β becomes harder to ask, because the thing you are trusting has no there at all.
Links: Colonial Origins of Comparative Development (Wikipedia) | Nobel prize 2024 and colonialism (The Conversation) | British Raj (Wikipedia) | Princely state (Wikipedia) | Millet system (Wikipedia) | Iltizam (Wikipedia) | The Imperialism of Free Trade (Wikipedia / Gallagher & Robinson) | Super Imperialism (Michael Hudson) | The Strategic Empire (2025 update) (Michael Hudson) | Dollars and Dominance (Columbia Business School / Yared & Pflueger) | Kissinger petrodollar system (LegalClarity) | US military bases worldwide (Al Jazeera) | Belt and Road Initiative (Wikipedia) | Debunking debt-trap diplomacy (Chatham House / Jones & Hameiri) | mBridge (Wikipedia) | mBridge clears $55B (HOKANEWS) | Eight ways AI will shape geopolitics in 2026 (Atlantic Council) | AI sovereignty paradox (CFR) | Middle powers and sovereign AI (Chatham House) | Silicon Empires (Rest of World / Nick Srnicek) | Reclaiming Europe's digital sovereignty (NOEMA)