And What Now? The Time for Choice

In my previous articles, The Split Stack and The World It Makes, I explored how the technological stack is fragmenting and what that fragmentation may mean for countries that increasingly depend on systems built elsewhere. The question that follows is simple: what now?

There was a period when technological dependence could be treated almost entirely as an economic question.

If another country could manufacture something more efficiently, you bought it. If an American company built better software, you licensed it. If Asian manufacturers produced electronics more cheaply, you imported them. If another country launched satellites at a lower price, you paid for the launch.

The logic of globalization was straightforward: specialization made everyone richer.

There was little economic reason for every country to manufacture every semiconductor, operate its own cloud infrastructure, build its own search engine, design its own aircraft engines or launch its own satellites. Technology was something you could purchase.

Increasingly, that assumption is becoming harder to maintain.

Cloud platforms now sit underneath governments, banks and hospitals. Semiconductors have become objects of export controls. Satellite networks can become strategically important during conflict. Artificial-intelligence systems are moving from optional software tools into the machinery through which companies write code, governments process information, scientists conduct research and consumers make decisions.

The question is therefore changing.

It is no longer simply: Who makes the best technology?

It is becoming: What happens if the technology on which your society depends is controlled somewhere else?

That is where the discussion becomes uncomfortable. Because for most countries, the obvious answer—build everything yourself—is no longer realistic.

The Fantasy of Complete Sovereignty

It is easy to speak about technological sovereignty as if it meant national self-sufficiency: build the chips, build the cloud, train the models, own the satellites, control the telecommunications infrastructure, manufacture the servers, produce the energy, write the software, and protect the data.

In theory, that is sovereignty.

In practice, even the world’s largest technological powers depend on supply chains crossing multiple countries. Modern semiconductors alone involve extraordinarily specialized ecosystems spanning design software, lithography equipment, manufacturing, packaging, materials, memory and precision components.

AI adds another layer of interdependence rather than removing it.

Frontier models require vast computing infrastructure. Computing infrastructure requires chips. Data centers require enormous amounts of electricity, cooling and network capacity. AI companies depend on cloud providers, chip designers, equipment manufacturers, energy systems and increasingly specialized suppliers.

There may therefore be no meaningful path toward complete technological autarky.

Even very powerful states must decide what they can safely depend on others to provide. For everyone else, the problem is more immediate. A medium-sized country cannot realistically decide tomorrow to recreate the entire technological stack. It cannot reproduce decades of semiconductor expertise, build a hyperscale cloud ecosystem, manufacture leading AI accelerators, develop competitive foundation models, construct a satellite constellation and create globally competitive software platforms merely by announcing a sovereignty strategy.

Money helps, but money does not instantly create industrial ecosystems.

So the real strategic choice is not between dependence and independence.

It is between different kinds of dependence.

The central questions become:

Where are we prepared to depend on others?

Where do we insist on retaining control?

And what can we build that makes others depend on us as well?

From those questions, three broad strategies are emerging.

Strategy One: The “Digital Colony”

The phrase is intentionally uncomfortable. It should not be taken literally. A country using American cloud services or Chinese telecommunications equipment has obviously not become a colony in the historical meaning of the word.

But the metaphor identifies a real strategic condition.

Imagine a country that decides not to build significant indigenous capabilities in frontier AI. Its government uses foreign cloud infrastructure. Its banks run on foreign software. Its universities access foreign AI models. Its companies build applications on foreign APIs. Its telecommunications systems depend on foreign vendors. Its citizens search, communicate, work and increasingly make decisions through platforms headquartered elsewhere.

Economically, this can be entirely rational.

Building alternatives might cost tens of billions of dollars and still produce inferior systems. Why spend enormous national resources recreating technology that can be rented cheaply from companies that already possess it?

For many countries, this will probably remain the practical answer. They will rent the frontier.

And in ordinary times, this may work extremely well. The country’s businesses gain access to sophisticated AI without financing foundation models themselves. Its universities can use world-class tools. Its government can modernize faster. Its entrepreneurs build on existing platforms instead of reinventing infrastructure.

Technological dependence can therefore increase capability. That is the paradox. The country becomes technologically more advanced while simultaneously becoming technologically more dependent.

This distinction matters. Dependence by itself is not evidence of failure. No modern economy produces everything it consumes. International specialization is one of the foundations of prosperity.

The real question is whether a dependency remains replaceable.

If one foreign cloud provider becomes too expensive, can the country move to another?

If one AI provider changes its terms, can companies switch models?

If geopolitical tensions interrupt one semiconductor supplier, are alternatives available?

If a foreign government imposes restrictions, can critical domestic systems continue operating?

A dependency that can be replaced relatively easily is very different from a dependency that has become structurally embedded.

The danger begins when switching becomes extraordinarily expensive—or impossible.

Once thousands of companies have built their systems around a particular model provider, cloud architecture or software ecosystem, changing supplier involves far more than signing a different contract. Applications need rewriting. Employees need retraining. Data must move. Security systems must be recertified. Business processes have adapted to the original platform.

What began as convenient outsourcing can become infrastructure. And infrastructure has political consequences.

The most extreme version is not that a foreign power presses a mysterious button and turns another country off. That image is dramatic but too simplistic.

The more realistic danger is gradual loss of bargaining power.

Prices change somewhere else. Access policies change somewhere else. Content rules change somewhere else. Export regulations are written somewhere else. Model capabilities are released according to someone else’s priorities. Data architecture follows somebody else’s technical standards.

Your country remains sovereign politically. But an increasing portion of its economic operating system is governed by decisions over which it has limited influence.

There is nothing inevitable about this outcome. And for many states, accepting considerable technological dependence may still be the most rational strategy.

The important thing is to recognize that a choice is being made.

Not building is also a technology policy.

Strategy Two: Strategic Regulatory Defense

Europe provides the clearest example of another response.

If you cannot—or do not want to—replicate every major technology company, you can still control the conditions under which those companies operate inside your market.

Rules concerning privacy, competition law, consumer protection, cybersecurity requirements, data rules, procurement standards, interoperability obligations, and AI regulation become instruments of strategic power.

The idea is straightforward.

A foreign technology company may own the model, but it does not own Europe. If it wants access to European citizens, businesses and institutions, it must operate within European rules.

This strategy is sometimes dismissed as “regulating what others invent.” That criticism contains some truth, but it misses why regulation can itself become a source of power.

A market of hundreds of millions of relatively wealthy consumers is difficult to ignore. Rules adopted inside such a market can influence how global technology companies design products everywhere.

Economists sometimes describe this phenomenon through the “Brussels effect”: companies comply with European standards because maintaining entirely separate systems for different markets may be inefficient. In that sense, regulation can project influence far beyond the jurisdiction that created it.

But regulation has limits.

A law can establish safety requirements for an AI model; it cannot train one. Competition regulation can constrain a cloud provider; it cannot construct a data center. A privacy regime can define how information should be handled; it cannot manufacture advanced processors.

That limitation appears increasingly well understood in Europe itself.

The European strategy is no longer simply regulatory. The EU is simultaneously expanding computing infrastructure, developing AI Factories and preparing much larger AI Gigafactories. As of 2026, the European Commission says 19 AI Factories are being deployed and has launched a program for up to seven Gigafactories, backed by public financing intended to mobilize more than €30 billion of combined investment.

The EU has also explicitly framed newer initiatives across chips, cloud, open source and AI in terms of technological sovereignty and reducing strategic dependency.

That evolution is revealing.

Europe appears to be discovering that regulatory sovereignty and productive capability cannot be separated indefinitely.

Rules can protect autonomy. But autonomy ultimately requires some capacity to build.

This makes “Strategic Regulatory Defense” a better description than simply “the regulatory strategy.”

The objective is not to stop technology at the border. It is to prevent technological dependence from eliminating domestic political authority.

A government may decide that foreign AI models are welcome—but that sensitive government information must remain within national jurisdiction. It may permit foreign cloud providers while requiring local resilience. It may allow foreign platforms while enforcing competition rules designed to prevent irreversible lock-in. It may require transparency or auditing in areas where automated decisions affect fundamental rights.

None of this produces technological independence.

It attempts something different: preserving room to govern inside technological dependence.

The difficulty is finding the point where protection does not become paralysis.

Regulate too little and domestic institutions may become dependent on systems they cannot meaningfully influence. Regulate too aggressively and companies may hesitate to invest, domestic experimentation may slow and emerging firms may carry compliance costs that established giants can absorb more easily.

There is no simple equation that solves this.

But it is a genuine strategic choice.

Strategy Three: Sovereign Alliances and Niche Specialization

The third strategy begins by accepting something that politicians often dislike saying publicly:

We cannot build everything.

Instead of treating that as defeat, a country asks a different question:

What can we become exceptionally good at?

Perhaps advanced memory, robotics, semiconductor materials, energy, particular forms of manufacturing, satellite systems, cybersecurity, AI optimized for a national language, specialized scientific models, secure cloud infrastructure, or industrial datasets unavailable elsewhere.

Then sovereignty stops meaning: We depend on nobody.

It begins to mean: Others have a reason to depend on us too.

That is a profound difference.

Consider two countries. Both depend on foreign semiconductors, cloud services and AI models.

The first contributes almost nothing technologically indispensable to the countries supplying them. The relationship is asymmetric. The suppliers can impose conditions with relatively little fear of losing something essential in return.

The second country also imports technology. But perhaps it dominates an important semiconductor material, or produces advanced memory required for AI systems, or controls critical manufacturing expertise, or supplies energy, or possesses unique scientific capabilities, or operates strategically important infrastructure.

That country is still dependent, but its dependence is embedded in a network of reciprocal dependence. It has bargaining power.

This may be one of the most important distinctions in the emerging technological order.

The objective for many countries should perhaps not be self-sufficiency. It should be indispensability.

South Korea: Specialization with Depth

South Korea demonstrates this logic particularly clearly.

It already possesses major capabilities in semiconductors, memory, electronics and industrial manufacturing. Yet it is also investing directly in sovereign AI models.

Its Ministry of Science and ICT has been running a Sovereign AI Foundation Model project supporting domestic teams with compute, data and government backing. In 2026, Korean teams continued advancing through the program, while the government explicitly described sovereign models as part of its effort to retain control over critical AI capabilities.

Korea is not attempting to recreate every component of the global technology economy. Nor is it simply accepting dependence.

It is strengthening layers where it already possesses industrial depth while attempting to prevent the model layer from becoming entirely external.

That is niche specialization evolving toward strategic depth.

India: Sovereignty Through Optionality

India starts from a different position.

It possesses enormous software talent, a vast domestic market, a rapidly expanding digital infrastructure and linguistic diversity that global AI systems do not always serve well.

Rather than attempting to reproduce the American technology ecosystem exactly, India has been building shared compute infrastructure while supporting domestic foundation models and multilingual systems.

By its 2025–26 annual reporting, the IndiaAI Mission had established access to more than 38,000 GPUs through 14 cloud partners and was supporting multiple indigenous foundation-model initiatives.

The logic is again different from total independence.

India can continue using foreign technology while ensuring that domestic companies and researchers have enough infrastructure to build alternatives where national scale, language or strategic need makes them valuable.

It is sovereignty through optionality.

The Gulf States: The Negotiated Hybrid

The Gulf states make the framework still more interesting because they have something technology companies desperately need: capital, energy and strategic geography.

The United Arab Emirates in particular has been combining domestic infrastructure ambitions with deep partnerships involving major foreign technology companies.

Abu Dhabi’s government has committed to sovereign cloud infrastructure while simultaneously working with Microsoft; in 2026 it expanded Microsoft Copilot across government while requiring AI processing for that deployment to remain within UAE borders.

That is neither complete domestic independence nor simple technological outsourcing.

It is a negotiated hybrid.

Foreign technology is accepted, but deployment conditions attempt to preserve selected elements of sovereignty.

This may become a common model.

Collective Technological Sovereignty

There is another consequence to this approach.

If no medium-sized country can realistically reproduce the entire technology stack, several countries can combine complementary capabilities.

One possesses semiconductor expertise, another contributes energy, another contributes models, another possesses manufacturing depth, another controls important datasets, another contributes capital, and another provides secure cloud infrastructure.

Together, they can create something no member could construct alone.

This is not traditional sovereignty.

It is collective technological sovereignty.

Military alliances have worked this way for decades. Not every NATO country manufactures every weapon system. The alliance creates security partly through interoperability and mutual commitments.

The same logic may increasingly apply to technology.

A sovereign AI strategy need not require every country to train its own trillion-parameter model. It may require credible access to compute, models, data and infrastructure through relationships that cannot be withdrawn casually.

The distinction is important.

Buying a foreign service is a commercial transaction.

Building mutual technological dependency is a strategic relationship.

Managed Interdependence

There appears to be another choice: do nothing.

Let private companies decide. Import whatever works. Avoid expensive industrial strategies. Assume markets will provide substitutes if problems arise.

For some technologies, this remains perfectly reasonable. Nobody needs a sovereign national alternative to every accounting application or video-editing tool.

The mistake is treating all technologies equally.

Some become foundational.

Electricity is not simply another product. Telecommunications networks are not simply another service. Payment infrastructure is not merely an application. And once AI becomes deeply embedded in administrative, economic, scientific and military systems, parts of AI infrastructure may acquire similar characteristics.

The challenge is identifying which dependencies are strategic before replacing them becomes prohibitively expensive.

That decision will inevitably vary by country.

A small trading state has different requirements from a continental power.

A country with abundant energy has different options from one that imports almost all of it.

A country with a major semiconductor industry has leverage another country does not.

A multilingual state may place greater value on domestic models.

There is therefore no universal sovereignty checklist.

There is only strategy.

This is where the categories become useful—but only if we resist turning them into rigid camps.

Real countries will mix them.

Europe will continue using American cloud providers while regulating them and simultaneously constructing domestic compute. South Korea will use foreign chips and models while developing sovereign models and strengthening areas where its companies are already globally important. India will import advanced technology while expanding domestic compute and models suited to its own languages and market. The UAE will partner with foreign technology leaders while negotiating domestic control over selected infrastructure and data.

Even the United States and China combine domestic capability with dependencies they would prefer to reduce.

The three strategies are therefore not destinations.

They are sliders.

Every country will position them differently. The strategic question is the mix.

What to Own, What to Regulate, What to Build

Some technologies are simply not worth reproducing domestically.

If five reliable suppliers exist, switching costs are low and the service is not strategically sensitive, dependence may be economically rational. Trying to nationalize every technological layer would waste enormous amounts of capital.

Sovereignty pursued without economic discipline can become expensive isolation.

So countries will need the confidence to say: We do not need to own this.

That may be just as important as deciding what they must control.

Other technologies may remain foreign but operate so deeply inside society that domestic rules become essential:

Healthcare AI

Government cloud systems

Financial infrastructure

Identity systems

Large communication platforms

Critical algorithms affecting access to employment, credit or public services

In those areas, allowing technological dependence does not require surrendering regulatory authority.

The strategic principle becomes: You may provide the technology, but we determine the rules under which it operates here.

That is the logic of regulatory sovereignty.

Then there are capabilities for which dependence becomes strategically dangerous:

Compute

Energy infrastructure

Certain communications systems

Defense technologies

Models trained for sensitive government or military applications

Here the calculation changes.

Efficiency is no longer the only objective. Redundancy has value. Domestic capacity has value. An alternative supplier has value even if it costs more.

This is familiar in national defense. Nobody chooses military capability solely by asking which foreign supplier is cheapest.

Some technologies may now require similar thinking.

Finally comes the most ambitious component.

A country should ask: What can we build that would be difficult for others to replace?

This is where technological strategy becomes something more than protection.

The goal is not simply to reduce vulnerability. It is to create leverage.

The Netherlands does not need to manufacture every semiconductor to matter enormously to the semiconductor ecosystem.

Taiwan does not need to own every layer of computing to occupy a critical position in advanced manufacturing.

South Korea does not need to reproduce the whole American digital economy for its memory and electronics capabilities to matter.

Strategic specialization can therefore produce sovereignty through interdependence.

And perhaps that is the most realistic form of sovereignty available in a deeply integrated technological world.

The Larger Transition

This may ultimately be the larger transition.

The globalization of the late twentieth century often treated dependency as economically benign. If another country could produce something better or cheaper, buying it was rational.

The geopolitical shocks of the last decade have made that assumption harder to sustain.

Semiconductor restrictions, pandemic supply disruptions, sanctions, energy crises and wars have all reminded governments that efficiency and resilience are different objectives.

AI intensifies the issue because digital infrastructure can become deeply embedded extremely quickly.

Once thousands of institutions depend on a particular technological architecture, changing it becomes progressively harder.

The future may therefore not be deglobalization.

It may be managed interdependence.

Countries will still trade. Technology will still cross borders. Supply chains will remain international. American companies will operate in Europe, European equipment will enter Asian factories, Korean memory will sit beside American processors, Gulf capital will finance global infrastructure, and Indian engineers will build systems used around the world.

But governments will increasingly distinguish between:

Ordinary economic dependence

Dependence capable of becoming strategic leverage

That distinction changes policy.

The Time for Choice

None of these strategies guarantees technological success.

The digital-colony model can provide extraordinary technology cheaply and quickly, but it can create deep dependency.

Regulatory defense can preserve political autonomy, but regulation alone cannot create industrial capability.

Sovereign alliances and specialization can produce strategic leverage, but they require governments to choose areas of concentration and accept dependence elsewhere.

There is no clean solution because the objective itself has changed.

The old objective was technological independence. For most countries, that is probably unattainable.

A more realistic objective is something else:

Negotiable dependency.

Dependence in which alternatives remain possible. Dependence in which suppliers need something from you too. Dependence governed by institutions rather than merely accepted by default. Dependence that can be renegotiated when circumstances change.

That may be what technological sovereignty actually means in the AI era.

Not isolation. Not self-sufficiency. Not building a national copy of everything invented elsewhere.

But ensuring that no single external actor occupies a position from which your essential technological choices cease to be choices.

And what now?

The first age of globalization taught countries to ask: Where can this be produced most efficiently?

The digital age taught them to ask: Who builds the best technology?

The AI age adds another question: Who controls the system after we depend on it?

Some countries will decide that renting the frontier is worth the risk. Others will use law and market power to constrain technologies they cannot fully control. Others will build sovereign capabilities in selected fields and assemble the rest through alliances.

Most will do all three.

The important thing is that these choices are increasingly being made now, whether governments describe them this way or not.

Every foreign cloud contract, national AI program, semiconductor subsidy, data-localization requirement, sovereign-compute project, and technology alliance answers, in a small way, the same question.

Where are we willing to depend?

Where must we retain control?

And where can we become indispensable?

The age of complete technological independence may never have truly existed.

But an age in which dependency itself becomes a strategic choice is clearly arriving.

And that is why the time for choice is now.

Karim Spring