Government in the Cloud
How satellite internet, mesh networks, AI governance, and decentralised finance are migrating authority from territory to infrastructure stacks—and what that means for how we live, who governs us, and whether democratic accountability can survive the transition.
The Urban Bargain and Its Breaking Point
For most of modern history, governance has been a territorial enterprise. Political authority attaches to borders. Obligation flows from jurisdiction. Belonging is defined by location. This is not merely a legal convention; it is an infrastructure story.
The city became the dominant unit of political and economic organisation not because of ideology, but because density rewarded itself. Roads, water systems, power grids, hospitals, and schools all scale more efficiently in places where people concentrate. And people concentrated where infrastructure was. The self-reinforcing logic was irresistible: move to the city for opportunity, find the networks already there, watch them improve because you arrived. The United Nations, tracking this dynamic across decades, found that 55% of the world's population was already living in urban areas by 2018 and projected that share to reach 68% by 2050—an intensifying trend, not a plateau. Cities, the World Bank has noted, generate roughly 80% of global GDP. Urbanisation was not a preference. It was, structurally, the path of least resistance.
Terrestrial telecommunications extended this logic rather than disrupting it. Fibre networks, landlines, and cellular towers are all built out where density makes them commercially viable. The International Telecommunication Union has documented persistently wide urban-rural gaps in internet access, particularly across lower-income regions. In India—a country that has made deliberate public investments in rural connectivity, including the BharatNet programme linking over 214,000 Gram Panchayats with nearly 700,000 kilometres of optical fibre—the gap between coverage and quality remains stark. A Centre for Economic Data and Analysis study found that only 3.8% of rural Indian households have access to high-speed fibre, against 15.3% in cities. Coverage can be decentralised; capability—speed, latency, reliability—tends to stay centralised. The network does not liberate the periphery. It reinforces the centre.
For decades, this was tolerable because the arrangement worked. Cities were messy but productive platforms for life. The bargain held: absorb the costs of density in exchange for access to employment, services, culture, and connections that simply did not exist elsewhere. Leaving the city meant, for most practical purposes, stepping off the map.
But the bargain is fraying. The very success of urban concentration generates stresses that governance cannot outpace. Housing is the most visible symptom. According to the OECD's 2024 Regions and Cities assessment, housing costs in large functional urban areas—cities above 1.5 million inhabitants—rose by 68% over the preceding decade, and by 2023 were 86% more expensive than in very small cities. The IMF has documented that the COVID-19 period pushed house prices to record levels in many countries, compounding an affordability crisis that was already structurally embedded. UN-Habitat, meeting in 2025, reported that more than 2.8 billion people are inadequately housed globally, with over 1.1 billion living in informal settlements—not on the margins of cities but, in many cases, inside them. Congestion, as the World Bank has repeatedly noted, is a material cost that local governance must manage but rarely eliminates. Climate exposure is increasing. Political legitimacy is fraying.
The problem, then, is not simply that cities have become unpleasant. It is that the urban model contains a structural lock-in that has historically made exit very costly. Connectivity sits at the heart of that lock-in. To leave the city was not merely to relocate; it was to sever access to the networks—professional, digital, social—that made modern participation possible. The gravitational pull of the urban grid was as much about infrastructure as it was about opportunity. As long as terrestrial connectivity rewarded density, decentralisation remained largely aspirational. If that reward structure could be altered, a different political geography might become possible.
Infrastructure as Permission
Every now and then, a technology arrives that does not merely improve an existing arrangement but changes the structural incentives that produced it. Satellite internet is one of those technologies.
The shift began, quietly, around 2020. Starlink—SpaceX's low-Earth-orbit broadband service—began commercial rollout that year, and the scaling that followed was rapid by any measure. By mid-2025, the service had passed six million subscribers across roughly 140 countries and territories. By December 2025, it had reached nine million. By February 2026, ten million, operating across 155 countries, with the constellation growing at over 20,000 new subscribers per day. Amazon's Kuiper network is expanding in parallel, and regulatory bodies including the FCC have approved further large-scale satellite deployments. This remains a small fraction of the 5.5 billion people online globally—but satellite's relevance lies less in its current market share than in where it operates. It serves the geographies that terrestrial infrastructure could never justify: remote regions, difficult terrain, sparse populations, mobile operators seeking coverage in dead zones. South Africa's Vodacom partnered with Starlink in 2025 precisely because rural terrestrial build-out was prohibitively expensive. That is not a commercial anecdote; it is the mechanism.
What satellite changes is not speed, exactly. Ookla measurements show Starlink median U.S. download speeds at around 118 Mbps as of 2025, clearing the FCC's 100 Mbps benchmark, though upload speeds and performance at the edges of coverage areas remain uneven. The more important shift is decoupling: broadband access no longer requires proximity to terrestrial infrastructure. A household with a satellite terminal can participate in modern economic and social life from places that the fibre and cellular build-out logic would never have reached. The geography of opportunity—which was always partly a geography of connectivity—becomes more porous.
This decoupling activates a variable that has always existed in human behavior but was previously constrained: agency in co-location. Where you live and with whom are now partially separable from where the network is. Once connectivity is no longer a binding constraint on exit, a different question begins to dominate: who do you want to live near? Values, norms, lifestyle, political disposition—these preferences, which had always existed, now become actionable at a spatial level in ways they previously could not be for most people.
The evidence that this behavioral shift is underway is visible, if still early. The OECD documented a "doughnut effect" in post-COVID housing demand, as remote work shifted preferences and purchasing activity away from urban cores and toward the surrounding rings. The U.S. Census Bureau found that 13.8% of American workers usually worked from home in 2023—more than double the 2019 rate of 5.7%—representing a shift from roughly nine million to over 22 million workers. The OECD also tracked the policy institutional response: "digital nomad" visa schemes, which barely existed before 2020, had spread to 40 or more jurisdictions by mid-2024, with many countries—as the Financial Times reported—now offering pathways from temporary remote residency to longer-term settlement. Governments are adapting, somewhat tentatively, to the reality that a portion of their mobile, economically productive population is choosing to be somewhere other than a legacy urban centre.
The sorting dynamic carries a second-order effect that the spine of this argument calls "cloud geography." Satellite-enabled communities—whether intentional or emergent—do not become isolated. They become nodes in a connected archipelago. The connectivity provider mediating that network is not just a utility delivering packets. It is the primary interface through which attention flows, services are accessed, and, increasingly, social reality is calibrated. When the interface layer becomes this central, its governance implications are no longer trivial. StatCounter's February 2026 data shows Google commanding roughly 90% of worldwide search—a near-monopoly on the act of discovering information. Meta reported 3.58 billion daily active users across its family of platforms as of December 2025. The European Commission, operationalising this concern in law, treats services above roughly 45 million EU monthly users as "very large" under the Digital Services Act, subjecting them to heightened scrutiny. This is not regulatory overreach. It is institutional recognition that the interface has become infrastructure—and that infrastructure that shapes how populations perceive, communicate, and coordinate is a form of governance, whether or not it calls itself that.
Alongside the satellite layer, a parallel connectivity mode is quietly gaining significance at the opposite end of the scale. Local mesh networks restore locality as a unit of resilience—coordination that does not depend on any central infrastructure. Jack Dorsey's Bitchat, announced in July 2025, is the most visible recent example: a Bluetooth mesh messaging application that functions without Wi-Fi, cellular service, accounts, or central servers. Messages route peer-to-peer, encrypted end-to-end, with each device acting as a relay node. The app includes a "panic mode" that erases all stored data, and Dorsey has described its dual transport architecture—Bluetooth mesh for offline local communication, the Nostr protocol for global reach when internet is available—as an experiment in censorship-resistant coordination.
Within weeks of its release, the experiment became something more urgent. In September 2025, Bitchat saw 70,000 downloads in a single week from Madagascar during political protests, against roughly 360,000 total worldwide downloads reported at that point. Similar spikes occurred in Nepal the same month, where nearly 50,000 Nepalese users downloaded the app in a single day. This is not a technology story. It is a governance story: when centralised infrastructure is contested or blocked, communities self-assemble communications capacity from the ground up. The precedent is not new—Bridgefy was downloaded over a million times in a single day during Myanmar's 2021 coup—but Bitchat's design makes the political implications explicit. A dual connectivity stack is emerging: satellite networks provide global reach and cluster-to-cluster coordination from above; mesh networks provide hyperlocal resilience and proximity-based trust from below. Together, they represent a fundamental change in the infrastructure of collective life.
The New Order — Governing the Stack
The communities enabled by this connectivity shift are not simply scattered households. They are, increasingly, emergent governance experiments. Some of this is explicit: the WIRED-documented "network state" movement and associated billionaire-backed enclave projects represent a cultural and financial push toward building governance-lite communities outside conventional political contracts. Honduras's ZEDE zones—highly autonomous governance experiments with long concession terms—were far enough along by 2024 that their Supreme Court ruling them unconstitutional generated genuine political conflict. The question they posed was serious: can a community opt out of a territorial state's political arrangements while still occupying its geography? The question remains open, and its implications are spreading.
Most of this is less theatrical. It is the quiet accumulation of mobile professionals choosing to live near people with shared values, powered by satellite connectivity, organised through digital platforms. But in aggregate, these clusters do not simply detach from the centre. They reshape it. When a critical mass of households exercises exit, the centre must adapt—in service models, in policy, in cultural expectations. The centre-periphery relationship becomes reciprocal: differentiated offshoots act as experiments whose feedback the centre cannot ignore. The city's political dominance is not eliminated, but it is conditioned.
Over this connectivity base, a second-order transformation is unfolding: governance functions themselves are becoming software-like. Identity verification, benefits delivery, health interactions, dispute resolution, compliance reporting—these were once siloed across separate agencies with incompatible data systems and incompatible incentives. They are being abstracted into interoperable digital services. The World Bank has argued that AI can "transform governments" by improving back-end efficiency, cross-domain coordination, and service delivery. The OECD frames AI as capable of enabling the re-imagination of public service, collapsing bureaucratic silos into coordinated capability. These are not speculative claims. The UK government has deployed its "Humphrey" AI tool to review consultation responses, explicitly targeting time and cost savings. GOV.UK has articulated a vision for AI-mediated citizen interaction as the primary interface to government services. Estonia's KrattAI strategy envisions a single, assistant-mediated channel to all public services—a direct instantiation of AI as the orchestration layer across fragmented government functions.
When AI mediation is bundled with connectivity, the resulting system begins to resemble a governing platform: access plus identity plus services plus behavioral coordination. This is where the sovereignty problem becomes unavoidable, and it is why it must be treated as a problem of behavior design rather than merely a problem of technology policy.
Defaults are governance. Prioritisation is governance. Bundling is governance. When a private provider embeds its priorities in the product architecture of an AI-mediated service layer—determining what information a citizen sees first, which pathways are easy and which are friction-laden, how disputes are resolved and by whom—it exercises governance power without electoral accountability. The question is not whether anyone intended this outcome. It is whether democratic societies are equipped to recognise it and respond before the power is consolidated.
The state response, where it exists, is what practitioners have begun calling Sovereign AI: the building of public capacity in models, compute, data governance, and service orchestration to preserve the state's ability to fulfil its functions and maintain democratic legitimacy without full dependence on private platforms. India's IndiaAI Mission, with a ₹10,371 crore five-year outlay and a publicly deployed GPU infrastructure of 10,000 units, is among the most explicit examples of sovereign compute as public infrastructure. The EU's EuroHPC AI Factories programme—expanded to six additional sites in October 2025—and the EU Council's January 2026 framework for AI gigafactories treat compute capacity as a strategic asset on a par with energy. France, having invested over €2.5 billion in AI research and infrastructure since 2018, frames AI leadership as a long-run public investment question. The EU's "Apply AI" strategy, reported by the Financial Times in 2025, explicitly targets reduced reliance on U.S. and Chinese AI platforms within public administrations. This is not defensive nationalism. It is an institutional recognition that the stack through which governance is delivered must be subject to public accountability.
The economic layer undergoes a parallel transformation. Decentralised finance introduces an economic sovereignty layer that can operate across borders and outside conventional intermediaries. The Bank for International Settlements defines DeFi as financial services delivered on permissionless blockchains, and has noted its potential to produce fragmentation in the financial system. The European Central Bank, surveying DeFi's architecture, describes it as cutting out centralised intermediaries through automated protocols—raising questions about systemic risks from a non-bank system outside regulatory perimeters. The Financial Stability Board has treated crypto-asset markets and stablecoins as potential financial stability risks requiring coordinated international response. These are not theoretical concerns. Standard Chartered, reported by Reuters in October 2025, estimated that stablecoin adoption could draw one trillion dollars out of emerging-market banks over the following three years, with stablecoin holdings potentially rising from roughly $173 billion to $1.22 trillion by 2028. Chainalysis found in 2024 that the Central and Southern Asia and Oceania region leads global crypto adoption—not because of libertarian ideology, but because conventional financial systems are costly, constrained, or unstable for many populations there. DeFi, in those contexts, is not an alternative for sophisticates. It is a practical workaround for governance failure.
The composite result is a multi-scale stack: mesh networks provide local continuity and trust when central infrastructure fails or is contested; satellite networks provide global reach and cluster-to-cluster coordination; AI-mediated service orchestration ties governance functions together across domains; DeFi enables portable capital, alternative credit, and cross-jurisdiction value exchange. Each layer individually represents a meaningful shift. Together, they represent a potential reorganisation of the substrate through which political and economic life is conducted.
The implications for behavior design—and for anyone whose work involves understanding how environments shape choices—are acute. Governance has always worked, in part, through the design of default environments. The architecture of a city, the layout of a government office, the sequencing of a benefits application—these are choice environments that shape behavior before individuals consciously engage with them. What changes now is that those default environments are increasingly digital, increasingly private, increasingly global, and increasingly AI-mediated. The designer of the interface is, in a meaningful sense, the designer of the governing environment. When that designer is not accountable to the people being governed, the concept of democratic consent begins to hollow out from the inside.
We do not end with a resolution because there is not one yet. Territory still matters. Cities remain powerful. The stack is still being built and its control is still being contested. What has changed is the question. The decisive question of political life is no longer primarily "who governs this land?" It is "who controls the stack through which life is coordinated?" That question is being answered, largely by default, by a small number of infrastructure providers, platform companies, and AI developers operating outside the architecture of democratic accountability. The path forward is not to reject the stack—it cannot be rejected, and many of its affordances are genuinely emancipatory—but to insist that its construction is a political choice, subject to political contestation, with democratic standards applied to it as seriously as they are applied to territory. We are, in other words, still deciding whether government in the cloud is a public infrastructure or a private one. That decision will not wait indefinitely.
