The Internet Is Becoming Harder to Border

The most consequential conflict on the modern Internet was not designed. It emerged because networks keep making geography cheaper while states still depend on making geography matter.

There is an easy way to misunderstand the technological conflict between open networks and authoritarian states: imagine that somebody planned it.

From Beijing, Washington can look like the architect of an enormous machine for dissolving borders. American companies build global cloud infrastructure. American financial technology makes it easier to transact across jurisdictions. American software companies sell tools that allow teams scattered across continents to work as though they occupied the same office. Crypto networks move assets without asking whether sender and recipient belong to the same banking system. Artificial intelligence increasingly allows one person to perform work that once required an organization.

Put these developments together and a seductive explanation appears: perhaps the United States is constructing an escape system for people living inside closed societies.

It is a good story.

It is also backwards.

The more interesting reality is that nobody had to design the escape routes. They emerged because much of the Internet economy has spent three decades attacking transaction costs, while territorial states continue to derive power from transaction costs.

That contradiction is becoming harder to hide.

The Internet is becoming harder to border.

Geography Used to Be Expensive

For most of modern history, doing something serious required physical proximity.

To start a company, you needed access to lawyers, banks, offices, employees and customers. To publish internationally, you needed publishers or distributors. To move substantial money across borders, you needed banks connected to other banks. To build software, you needed servers somewhere, people to administer them and capital to pay for the machinery. To reach a foreign market, you often needed a local organization.

Geography was not merely where economic activity happened. Geography was infrastructure.

That gave governments enormous leverage.

If economic life depended on institutions physically located inside a jurisdiction, then controlling those institutions meant controlling much of economic life. Governments could license banks, register companies, regulate telecommunications, restrict currencies, monitor organizations and determine which kinds of economic activity were legally recognizable.

The Internet began weakening this arrangement long before anyone understood where it would lead.

Email made correspondence nearly free. Open-source software made technical knowledge portable. Cloud computing separated computation from ownership of physical servers. Online payment systems reduced the importance of local merchant infrastructure. Remote work weakened the relationship between employment and offices. Digital marketplaces connected buyers and sellers who would previously never have found one another.

Each innovation looked narrow.

Together they began separating capability from location.

That distinction matters enormously.

A person may physically remain inside one jurisdiction while parts of his economic life increasingly exist somewhere else: his customers in another country, collaborators across several time zones, software running across a global network, assets represented digitally, intellectual work assisted by models hosted thousands of kilometers away.

The human being remains geographically located.

His capabilities become geographically dispersed.

The Delaware Company Is More Radical Than It Looks

Consider something as boring as company formation.

A century ago, establishing a meaningful commercial presence in another country normally required substantial local relationships and professional infrastructure. Today, services exist specifically to make American corporate infrastructure accessible to founders regardless of where they began.

Stripe originally described Atlas as a way to give entrepreneurs around the world access to the building blocks of an Internet business; its current service automates Delaware incorporation, tax-ID acquisition and founder-equity paperwork.

There is nothing revolutionary in the marketing language. There does not need to be.

The revolution is hidden inside the reduction of friction.

The important unit is no longer simply the American entrepreneur starting an American company. It can be a person somewhere else attaching part of his economic activity to American commercial infrastructure.

This does not erase jurisdiction. Taxes still exist. Securities laws still exist. Banking compliance still exists. Immigration law certainly still exists. A Delaware corporation is not a magical passport out of local law.

But that objection misses the direction of travel.

The question is not whether borders have disappeared.

They plainly have not.

The question is how much economic capability can cross a border without the human body crossing it.

That amount keeps increasing.

Compute Is Leaving the Building

Software infrastructure pushes the same logic further.

A developer no longer necessarily needs to own a server, negotiate for rack space or even choose a conventional server region. Modern serverless systems allow code to be deployed onto globally distributed infrastructure while abstracting away much of the physical machinery underneath it. Cloudflare, for example, describes Workers as running applications across its global network rather than on a developer's individual machine or a single centralized server.

Again, none of this was invented as political dissidence.

Engineers wanted lower latency.

Companies wanted easier deployment.

Customers wanted reliability.

Developers hated managing servers.

Capital wanted scalable software businesses.

The political consequence emerged from the engineering.

Once computation becomes a service purchased across a network, physical possession of computers becomes less closely tied to computational capability. Once storage becomes remote, physical possession of disks becomes less closely tied to information. Once software distribution happens globally, controlling the local software market no longer guarantees control over every tool people can reach.

The state can respond, of course.

It can block services. Demand localization. License networks. Filter traffic. Criminalize particular transactions. Require platforms to identify users. Pressure companies into compliance. Construct national substitutes.

China has demonstrated how far such techniques can go.

But notice what has happened to the nature of the contest.

The state is no longer merely governing activities occurring naturally inside its territory.

It is continuously rebuilding the significance of the border against technologies whose commercial purpose is often to make the border less economically relevant.

That is a very different problem.

Money Discovered the Internet Too

Finance makes the contradiction more obvious.

Traditional international money movement depends heavily on regulated financial institutions, correspondent relationships and national currencies. Governments therefore possess powerful control points.

Crypto did not abolish those control points. The history of the industry has made that painfully obvious: exchanges can be regulated, stablecoin issuers can impose restrictions, fiat on-ramps can require identification, blockchains can be analyzed and users remain subject to law.

But crypto introduced something conceptually difficult to reverse.

Value itself became natively transmissible across an Internet protocol.

Stablecoins have pushed this idea closer to ordinary commercial infrastructure. Circle now explicitly markets USDC infrastructure for global payments, including blockchain transfers and cross-border settlement, rather than treating digital dollars merely as speculative crypto assets.

That does not mean money has escaped government.

It means governments face a new category of money whose network topology is different from that of the twentieth-century banking system.

The distinction is crucial.

The old question was:

Which bank holds your money?

The newer question can involve several layers:

Which asset? Which wallet? Which blockchain? Which issuer? Which exchange? Which jurisdiction? Which bank provides the final conversion?

More layers do not automatically produce freedom. They can produce new surveillance systems and new concentrations of private power.

But they also multiply the interfaces a state must understand and control.

The border has not vanished.

It has become technically expensive to reproduce.

Then AI Arrived

Artificial intelligence changes something deeper than communication, computation or payments.

It changes organizational scale.

Many things people want from organizations are ultimately bundles of cognitive labor: translation, research, software development, graphic production, customer support, analysis, documentation, marketing, tutoring and administration.

Historically, acquiring these capabilities meant hiring people or purchasing services from organizations.

AI compresses some of them into software.

That matters politically because organizations have traditionally been easier to regulate than capabilities.

An office has an address.

A company has employees.

A newspaper has editors.

A school has teachers.

A consulting firm has consultants.

A software company has engineers.

But what happens when increasingly capable software lets one individual temporarily summon pieces of all of them?

Not perfectly, and not without infrastructure controlled by large companies. But enough to alter the minimum viable size of an organization.

This may become one of AI's least appreciated geopolitical effects.

The twentieth-century state became extraordinarily good at managing organizations. Organizations register. Organizations employ. Organizations own property. Organizations open bank accounts. Organizations acquire licenses. Organizations develop hierarchies that can be pressured from the top.

AI does not eliminate organizations.

It lowers the number of humans required to create one.

That makes the smallest economically meaningful unit smaller.

Two Systems Are Naturally Colliding

This is why interpreting the conflict purely as America versus China obscures the more interesting story.

The American technology ecosystem did not need a secret meeting in which somebody decided to weaken Beijing's capital controls, censorship architecture or organizational supervision.

The incentives were already enough.

A payment company wants more transactions.

A cloud company wants more workloads.

A SaaS company wants more customers.

A model provider wants more developers.

A remote-work platform wants more employers and workers.

A crypto network wants more liquidity.

An entrepreneur wants access to more markets.

A programmer wants software to deploy with fewer steps.

Almost every participant is rewarded for reducing some form of friction.

Authoritarian governance frequently runs in the opposite direction.

It needs certain frictions.

It needs identity to remain legible.

It needs organizations to remain identifiable.

It needs financial flows to remain governable.

It needs information channels to remain interruptible.

It needs politically significant groups to remain observable.

It needs jurisdiction to mean something operational rather than merely appearing as a line on a map.

That does not make every friction illegitimate. Liberal democracies also regulate money laundering, taxation, securities, privacy, national security and corporate conduct. Open systems have borders too.

The difference is that highly centralized political systems depend more heavily on the state's ability to make independent organization costly.

And the Internet keeps trying to make organization cheap.

There is the collision.

Not conspiracy.

Architecture.

The Great Firewall Was Only Version One

The first generation of Internet sovereignty was primarily about information.

Governments asked: What can citizens read? What can they publish? Which platforms can they access?

That produced censorship systems, filtering systems and national platform ecosystems.

The next generation is more difficult because the Internet is no longer merely carrying information.

It carries work.

It carries companies.

It carries computation.

It carries money.

Increasingly, it carries intelligence.

Blocking a newspaper and blocking an economic operating environment are not the same problem.

A state can isolate itself technologically. But isolation has a price. The same network connections that create political risk also create commercial value. The same foreign technologies that complicate control may improve productivity. The same international financial connections that permit capital movement also facilitate trade and investment.

The tighter the border becomes, the more economic activity it may obstruct.

The looser it becomes, the more autonomous capability can leak through.

This is not an engineering problem with a clean solution.

It is a structural tradeoff.

The Escape Routes Are Side Effects

This is why the phrase “digital escape route” can be misleading.

An escape route sounds like something constructed by a rescuer.

Most of these systems were built by people trying to sell software.

That is precisely what makes the phenomenon so powerful.

Nobody has to maintain a grand ideological commitment to keep it moving.

A founder does not need to care about political freedom to want global customers.

A developer does not need a theory of authoritarianism to prefer an open API.

A payments company does not need to oppose capital controls philosophically to want faster international settlement.

An AI company does not need a geopolitical manifesto to make one person more productive.

The incentives reproduce themselves.

Every generation of infrastructure asks roughly the same commercial question:

What expensive intermediary can we remove next?

But intermediaries are not merely businesses.

They are often control points.

Remove enough intermediaries and eventually you begin removing the places where authority used to sit.

That is why the Internet keeps accidentally becoming political.

Borders Will Fight Back

None of this guarantees that open networks win.

States are learning too.

Digital identity can strengthen surveillance. AI can automate censorship. Financial networks can make transactions more traceable rather than less. Cloud infrastructure can centralize power into a small number of corporations that governments can pressure. Stablecoins can acquire chokepoints. Platforms can become enforcement partners.

The same technologies that make individuals more portable can make them more visible.

The future is therefore unlikely to be a simple victory of networks over states.

It will be a competition over architecture.

States will attempt to insert jurisdiction into protocols, platforms, identity systems, chips, payment rails and cloud infrastructure.

Networks will continue producing abstractions that make location less important.

Neither side needs to consciously declare war on the other.

Their incentives are enough.

The Internet Is Not Borderless

That old slogan was always too romantic.

The Internet has borders everywhere.

Payment restrictions are borders.

App stores are borders.

Identity checks are borders.

Cloud regions are borders.

Export controls are borders.

Banking compliance is a border.

Censorship is a border.

Corporate law is a border.

And sometimes an actual police officer standing at an actual airport remains the most important border of all.

The transformation is subtler.

The Internet is not becoming borderless.

It is becoming harder to border.

Every year, governments acquire new instruments of digital control. Yet every year the technological economy also invents new ways to separate capability from geography: remote companies, globally deployed software, programmable money, distributed collaboration and increasingly portable intelligence.

That tension will define far more than the future of China.

It may define the political shape of the Internet itself.

The most consequential freedom technologies may never describe themselves as freedom technologies.

They may simply be products that make an old intermediary unnecessary.

And somewhere far from Silicon Valley, somebody will discover that a wall which still looks perfectly solid has acquired another door.