Geopolitics of AI: Why nations are fighting over AI (and what's actually at stake)

EDITOR’S NOTE

Dear future-proof humans,

Welcome to another edition of Nanobits. This week, we're zooming out.

In most editions, we focus on AI tools, workflows, and practical applications. But something fundamental is shifting beneath all of that. The global competition over AI is no longer just about which company builds the best model. It has become a contest between nations over power itself.

When governments treat advanced chips like weapons. When cloud providers must report foreign computing activity to their home countries. When companies split their operations across borders to serve competing masters. These are not normal market dynamics. This is geopolitics.

The decisions being made in Washington, Beijing, and Brussels today will determine which AI tools you can access tomorrow, what data they're trained on, and under what terms you can use them. The fight over compute, data, models, and talent is shaping the infrastructure of the AI era.

So this week, we're asking: how did AI become the new axis of global power? What are nations actually fighting over? And what does this mean for countries trying to navigate between competing tech blocs?

Let's break it down.

Why technology is the new center of global power

Geography used to define global power. The Himalayan mountains shielded India from the north. Ocean access determined trade routes. Rivers split territories. Today, power flows through fiber optic cables and data centers. The new battleground is not land or sea. It is technology.

Governments across the world now operate from a shared belief: technological leadership equals national power. This conviction is not academic. It shapes policy, drives investment, and redraws alliances. AI sits at the center of this transformation.

This shift is creating a new kind of global competition. Trade wars have become tech wars. Companies align with governments. International cooperation happens only when nations share strategic interests. And the rules are still being written.

Five trends shaping high-tech geopolitics

Trade wars are likely to be tech competitions at their core

Large-scale military conflict between nuclear powers is too risky. Economic warfare has limits when supply chains are interconnected. Technology offers a third path. Nations can compete here without triggering the clear consequences of bombs or total trade embargoes.

The backstops are not well-known. No one is certain what happens when you deny advanced chips to a rival for five years. This uncertainty encourages experimentation. The US restricted Nvidia chip sales to China. China invested billions to build alternatives. Both countries are testing boundaries in real time.

This competition happens across four levels: materials, machines, humans, and values. In semiconductors, the fight focuses on materials and machines. With AI, contestation occurs at all four levels at once.

Governments are accepting the costs of decoupling

Separating tech ecosystems is expensive. Nvidia loses revenue when it cannot sell to China. Intel faces higher costs when building fabs only in allied nations. Governments know this. They accept the trade-off.

Take Nvidia, for instance. About 50% of AI developers are in China, according to CEO Jensen Huang. Cutting off this market hurts growth. But the US government decided that national security priorities outweigh those lost sales. This calculation shows how seriously nations take technological advantage.

Aggressive national competition over high technology might produce some non-linear breakthroughs this decade

A concept called creative insecurity explains a historical pattern. When a nation faces a credible external threat from a technologically advanced adversary, its innovation rate accelerates. The threat cannot be overwhelming. Internal divisions cannot be too deep. But in that sweet spot, nations innovate faster.

The Sputnik moment is the classic example. Americans saw a Soviet satellite pass overhead with their own eyes. Panic followed. That panic created NASA and DARPA. Both organizations produced breakthroughs that defined the next half-century.

Today's tech competition might produce similar results. Nations are racing to develop new chip architectures, novel AI models, and alternative computing platforms. Some of these efforts will fail. A few might succeed spectacularly.

There is likely to be higher alignment between private high-technology players and their national governments

Intel wanted to invest in China a few years ago. It planned new chip design partnerships and manufacturing plants. Then geopolitics intervened. Now the US government owns a stake in Intel. The company builds fabs on American soil with government support.

This pattern repeats across the tech sector. Tech CEOs dine with presidents. Governments fund domestic chip production. Export controls restrict what companies can sell and to whom.

Some large companies try to hedge. Nvidia proposed isolating its China operations from US divisions. TikTok was separated from ByteDance and moved its headquarters. These "geopolitical enterprises" split governance across jurisdictions to serve both home governments and global markets. But this balancing act is fragile.

The Spin-On Economy from consumers to the military

Old tech followed a spin-off model. The government funded research for military use. Decades later, that technology reached consumers. The internet started as DARPANET. GPS was a military system first.

New tech follows a spin-on model. The most advanced chips are in your iPhone, not in an F-35 fighter jet. Consumer devices get cutting-edge technology first. Military applications come later.

Why? Economics. A semiconductor fab costs $12 billion to $15 billion. Only massive consumer sales justify that investment. Defense orders alone cannot cover the cost. This reality drives the core tension in AI geopolitics: companies need global markets, but governments want control.

These trends show how technology has become the new arena for global competition. Now, let's explore how AI, specifically, is changing the fundamental definition of national power.

How AI changes the power equation

Major innovations are rewriting the rules of national strength. AI is doing this in three ways.

It introduces new elements of power. Access to computing resources now matters as much as access to oil did after the internal combustion engine. A nation's ability to run large-scale AI training jobs directly affects its economic and military capacity.

It changes the importance of existing factors. Human capital matters more when AI depends on skilled engineers. Reliable electricity becomes critical when data centers consume gigawatts. Nuclear power gains strategic importance as the only reliable baseload energy source for AI infrastructure.

It alters a nation’s intermediate goals. Information warfare is now achievable at scale. Social control becomes technically feasible. These were not realistic objectives before. AI makes them possible, so nations pursue them.

With this new power equation, a fierce competition has emerged. The next section breaks down the specific "battlegrounds" where this contest over AI is taking place.

Four battlegrounds in the AI stack

The geopolitical fight over AI happens in layers. Each layer has different characteristics. Each requires different strategies.

Compute: The Hardware War

Compute is the most contested layer right now. It is detectable. You can count chips. It is excludable. Someone must buy them. It is quantifiable. Performance can be measured. And it comes from a concentrated supply chain.

The US uses export controls to restrict advanced chip sales. Executive Order 14110 requires cloud providers to report when foreign entities use large amounts of computing power. These measures aim to slow rival progress.

The most controversial proposal is on-chip governance. The idea is to build usage limits or remote shutoff switches directly into chips. A chip could report its location and usage. A manufacturer could disable it remotely if the terms are violated. This raises sovereignty concerns. Do you really own hardware that someone else can control from abroad?

Nations and companies are racing to develop alternatives. Nvidia dominates not because of GPU hardware alone. Its CUDA software platform connects high-level code to chips with high efficiency. Breaking this dependency requires open-source alternatives. Some governments fund these efforts. Companies invest too, hoping to capture market share.

ASML in the Netherlands holds a near-monopoly on extreme ultraviolet lithography machines. These machines are necessary for advanced chip production. China is investing heavily to build alternatives. The timeline is five to ten years. Until then, ASML's technology remains a critical bottleneck.

Data: The Fuel War

Data trains AI models. Controlling data means controlling what models learn.

Data localization laws require data to stay within national borders. Governments justify this as protecting a national advantage. In practice, it also gives them more control over what data companies can access and share.

Data poisoning is an offensive tactic. If you can manipulate an adversary's training data before they run expensive training jobs, their models will produce unreliable outputs. This is not theoretical. Cybersecurity experts warn about it. Governments consider it a real threat.

Some nations are building strategic datasets. The US AI Action Plan includes provisions for creating national data assets. These might cover scientific research, economic data, or other strategic domains.

Protecting these strategic data assets from theft or manipulation through industrial espionage and cyberattacks has become a national security priority.

Models: The Intelligence War

AI models are trained algorithms that produce useful outputs. As models grow more powerful, they become targets.

Governments debate restricting model weight distribution. Should powerful model weights be treated like controlled technology? The answer is not settled. Some argue open-weight models increase security through transparency. Others worry about adversaries gaining capabilities too easily.

Model corruption is another concern. If someone can sabotage a trained model directly, they degrade an adversary's AI capabilities without touching hardware or data.

The race for new architectures continues. Deep learning dominates now. Transformers are the leading model type. But a breakthrough in model design could shift the entire balance. Nations fund research into alternatives, hoping to find the next paradigm.

Open-weight models serve a strategic purpose. They reduce vendor lock-in. They build trust through transparency. They accelerate developer adoption. For nations worried about supply chain resilience, open models offer a hedge against dependency on any single company.

Talent: The Human War

AI depends on people. Skilled researchers and engineers are scarce. Nations compete to attract and retain them.

Visa restrictions limit the movement of AI scientists. Some countries bar foreign researchers from sensitive projects. Australia and Japan restrict Chinese participation in certain academic work. China imposes similar limits on foreign engineers.

Talent poaching works in the opposite direction. Nations create incentives to attract top researchers from rivals. Competitive salaries, research funding, and academic freedom all play a role.

The most important metric might not be top researchers. It might be diffusion to average engineers. China has many strong AI researchers. But the density of researchers publishing at top conferences is lower than in the US. China's "diffusion deficit" means fewer engineers can apply cutting-edge techniques in businesses and products.

The competition across these four layers is not random. It is driven by deeply held, and often conflicting, strategic visions of what AI is and what it means for the future of international relations.

Four competing visions, four strategies

How a country views the fundamental nature of AI determines its national strategy, its policies, and its posture on the world stage. Four main visions are currently competing to define the geopolitical era of AI.

It is important to note that these visions are not all viewed equally. The "AI as Nuclear Weapons" analogy, in particular, is considered controversial by many experts, who worry its framing could lead to overly restrictive and potentially harmful policies. These competing visions are actively shaping the policies and international relations of countries around the world today.

India’s strategic balancing act, navigating the new order

India shows how these forces play out in practice, plotting a technological landscape increasingly defined by US-China competition. Its approach is a practical blend of alignment and autonomy.

India is not non-aligned in technology. It has committed $10 billion to semiconductor fabs. It engages in selective international cooperation (e.g., iCET, Quad), and ISRO continues its collaboration with NASA on satellite projects despite tariff restrictions. It participates in Western technology ecosystems. Defense spending as a share of the budget is falling. Technology investment is rising.

But India seeks leverage, not total independence. Full indigenization is unrealistic given global supply chains. The goal is to achieve "leverage over critical supply chain segments" that give India some form of strategic deterrence. If India has capabilities others need, they cannot easily deny India access to what it needs.

In pursuing this strategy, India faces clear strategic trade-offs. India accepts the "opportunity costs in tech decoupling" from certain actors as a necessary price for building long-term national power and avoiding dangerous dependencies. It invests domestically. It builds partnerships with aligned nations. And it monitors emerging governance models, like on-chip controls, that could undermine strategic autonomy.

India's balancing act, aligning with partners while building independent leverage, offers a model for how other nations may navigate the emerging AI-driven geopolitical landscape, bridging the gap between national ambition and global realities.

End note

The game has changed.

Technology, with AI at its core, has moved from the periphery to the very center of global power competition. It is no longer just a tool for nations; it is a primary arena where their futures are being decided.

We've seen how this shift has created new rules of engagement, fundamentally altered the definition of national power, and opened up new battlegrounds across the entire AI stack.

Selective cooperation will happen among aligned nations. But comprehensive multilateral development is unlikely. The world is entering a geopolitical innovation race. It blends zero-sum security concerns with the potential for positive-sum gains among partners.

The rules are still being written. The outcomes are uncertain. And the questions we face are not easy ones:

Can the economics of AI and the geopolitics of AI coexist? AI needs massive global markets to justify the capital required. But nations want control over strategic technology. One demands openness. The other demands restrictions. Which force wins?

Will verification regimes emerge for AI the way they did for nuclear technology? Some are already proposing personnel checks, usage monitoring, and international inspections for AI development. Will nations accept this level of intrusion? And who decides what counts as acceptable use?

What happens to nations caught in the middle? The world is splitting into technological blocs. Some countries have the scale to lead. Others must choose sides or find leverage in specific supply chain segments. For most of the world, strategic autonomy is not an option. So what is?

The answers to these questions will shape the next decade. And whoever leads in AI will help write the rules for everyone else.