By Vanguard Enterprise Intelligence Unit with the work of Ian Bremmer, Pankaj Ghemawat, Michael Porter, Anu Bradford, and Henry Farrell.

Artificial intelligence is no longer only a technology race. It is now a contest over national power, industrial capacity, supply-chain control, energy access, data governance, talent concentration, and standards-setting influence.

For executives, this changes the strategic question. AI adoption can no longer be managed only as an internal productivity initiative or a digital-transformation project. It must be understood as part of a broader geopolitical operating environment. The firms that build, buy, deploy, or depend on AI are increasingly exposed to national policy choices, export restrictions, compute constraints, data-localization rules, industrial subsidies, cybersecurity concerns, and shifting alliances.

This does not mean every company needs to become a geopolitical institution. It does mean that companies can no longer treat AI as politically neutral infrastructure. The AI stack is becoming strategic terrain.

The most advanced models depend on chips, cloud capacity, data centers, specialized talent, energy, software frameworks, training data, and regulatory permission. Each of those inputs is becoming more contested. Nations want domestic capability. Alliances want trusted supply chains. Companies want scale, speed, and access. Regulators want accountability. Security agencies want control over dual-use risk. These priorities do not always align.

The result is a more fragmented competitive landscape. In one scenario, AI remains broadly global, with interoperable standards, cross-border research, shared cloud infrastructure, and commercial diffusion. In another, AI becomes increasingly sovereign, with national or regional ecosystems separated by regulation, export controls, security restrictions, and political mistrust. The real world is likely to contain elements of both.

Executives need strategies that can operate in that mixed environment.

AI Sovereignty as a Strategic Force

AI sovereignty refers to the ability of a nation, region, or institution to control the critical inputs, infrastructure, rules, and capabilities that determine how AI is developed and used. It is not only about owning a domestic large language model. It includes compute access, data governance, cloud infrastructure, semiconductor supply, skilled labor, cybersecurity, model evaluation, standards, and deployment rules.

Governments are pursuing AI sovereignty for several reasons.

The first is economic competitiveness. AI is expected to influence productivity, scientific discovery, defense capability, industrial automation, financial services, healthcare, education, logistics, media, and software development. Countries do not want to depend entirely on foreign systems for capabilities that may shape future growth.

The second is national security. Frontier AI systems can have dual-use applications in cyber operations, intelligence analysis, weapons development, influence campaigns, and critical infrastructure management. Governments are therefore treating advanced AI as a strategic technology rather than a normal commercial product.

The third is regulatory control. Different political systems have different views on privacy, speech, transparency, safety, liability, and state access to data. AI sovereignty gives governments more authority over the rules by which AI systems operate within their borders.

The fourth is industrial policy. AI infrastructure requires large investment in chips, data centers, electricity, networking, cooling systems, and cloud services. Governments increasingly view these investments as part of national industrial strategy.

For companies, AI sovereignty creates both opportunity and constraint. Sovereign investment may produce new markets, incentives, public-private partnerships, and demand for trusted providers. But it may also create duplicated systems, compliance complexity, restricted data flows, and limits on cross-border AI deployment.

The strategic challenge is to participate in national AI buildouts without becoming trapped by one jurisdiction’s assumptions.

The U.S.-China AI Contest

The U.S.-China rivalry is the central axis of AI geopolitics, but it is not a simple two-player race. It is a competition across multiple layers: compute, advanced chips, model quality, talent, deployment, open-source ecosystems, enterprise adoption, defense applications, data access, and global influence.

The United States retains advantages in frontier model development, leading AI labs, cloud infrastructure, venture capital, semiconductor design, and deep research ecosystems. Its technology firms operate at global scale and often set the pace for commercial AI adoption.

China has different strengths. It has a large domestic market, strong state direction, rapid industrial deployment, major platform companies, abundant engineering talent, and a policy system that can mobilize resources quickly. Export controls have limited China’s access to the most advanced AI chips, but they have also accelerated domestic substitution efforts and efficiency-focused innovation.

This creates a strategic paradox. Restrictions can slow a competitor’s progress, but they can also strengthen its determination to build independent capability. The longer the rivalry persists, the more both sides will seek to reduce dependence on the other.

For multinational firms, the implication is direct. AI strategy must be designed for a world in which technology flows are less predictable. A model, chip, cloud provider, data transfer, or software dependency that is available today may be restricted tomorrow. Business continuity will depend on architectural flexibility and geopolitical awareness.

Compute Becomes a Geopolitical Bottleneck

In the AI economy, compute is not merely technical capacity. It is strategic capacity.

Advanced AI systems require specialized chips, data-center infrastructure, networking, electricity, cooling, and software optimization. Access to compute affects who can train frontier models, who can deploy AI at scale, who can serve enterprise customers, and who can participate in the most advanced AI markets.

This makes compute a central point of geopolitical control. Export controls on advanced chips are one form of leverage. Restrictions on semiconductor manufacturing equipment are another. Critical minerals, photonics, high-bandwidth memory, advanced packaging, data-center components, and energy infrastructure are also becoming strategically important.

Executives should avoid thinking of compute as a commodity input. In many sectors, compute availability will shape competitive timing. A company with assured access to scalable, compliant, cost-effective compute may be able to deploy AI faster than competitors. A company dependent on fragile or restricted infrastructure may face delays, higher costs, or forced redesign.

The compute question should therefore become part of enterprise risk management. Where is the company’s AI capacity hosted? Which cloud regions matter? Which providers are exposed to regulatory or geopolitical restrictions? What happens if access is constrained? Can workloads be shifted? Are there jurisdictional rules around data, models, or inference? Is the company overdependent on one supplier or geography?

AI strategy without compute resilience is incomplete.

Energy as the Hidden Constraint

The AI race is also an energy race. Data centers require large and reliable electricity supply. As AI workloads grow, energy availability will influence where infrastructure is built, how fast it scales, and what it costs to operate.

This creates a new strategic link between digital ambition and physical infrastructure. AI may appear intangible at the user interface, but the underlying system is highly material. It depends on power grids, land, cooling, water, transmission capacity, permitting, and long-term energy contracts.

Countries with reliable, scalable, and affordable energy may gain an advantage in attracting AI infrastructure. Regions with grid constraints may find that AI investment becomes harder to support. Companies may need to evaluate not only cloud pricing, but also the energy resilience and regulatory environment behind that cloud capacity.

Energy also introduces reputational and sustainability questions. AI expansion will increase scrutiny of emissions, water use, power demand, and local infrastructure impact. Firms that deploy AI heavily may face pressure to explain not only what their AI systems do, but how the infrastructure behind them is powered.

Executives should treat energy as part of AI strategy, not as a facilities issue. In many industries, the availability and cost of AI infrastructure will become a competitive variable.

Standards, Governance, and the Race to Define Trust

AI competition is not only about who builds the strongest model. It is also about who defines the rules of acceptable AI.

Standards influence model evaluation, safety testing, transparency, interoperability, cybersecurity, data handling, intellectual property, procurement, and auditability. The countries and institutions that shape these standards will influence how AI markets develop.

For companies, standards matter because they determine market access. A model or AI-enabled product that is acceptable in one jurisdiction may need modification in another. A company that builds compliance into its AI architecture early may move faster than competitors that treat compliance as an afterthought.

The issue is especially important for multinational firms. They may need to navigate U.S. rules, European requirements, Chinese regulations, sector-specific obligations, and local data-governance regimes. Fragmented rules can create operational complexity, but they can also create strategic advantage for firms capable of managing them.

Trust will become a differentiator. Customers, governments, and partners will increasingly ask where AI systems are hosted, how data is protected, how outputs are validated, whether models can be audited, and whether suppliers are exposed to geopolitical restrictions.

In this environment, trusted AI is not only an ethical position. It is a commercial capability.

Middle Powers and the Strategic Space Between Giants

The AI geopolitical landscape is not limited to the United States and China. Middle powers have strategic options of their own. Countries such as Canada, the United Kingdom, France, Germany, Japan, South Korea, Singapore, India, the United Arab Emirates, and others are pursuing different versions of AI capability, sovereignty, regulation, and industrial positioning.

Middle powers generally face a difficult balance. Few can replicate the full AI stack at frontier scale. Building advanced chips, cloud infrastructure, national models, large talent pools, and global platforms requires enormous capital. But middle powers can still create influence by specializing.

There are several possible strategies.

One is the trusted infrastructure strategy. A country builds secure, compliant, energy-backed AI infrastructure and positions itself as a trusted location for data, cloud, and regulated AI deployment.

A second is the vertical specialization strategy. A country focuses on sectors where it has existing strength, such as healthcare, finance, defense, advanced manufacturing, energy, logistics, or public services, and builds AI capability around those domains.

A third is the standards-and-governance strategy. A country or region becomes influential by shaping rules, certification, safety evaluation, and responsible AI frameworks.

A fourth is the talent-and-research strategy. A country invests in universities, research labs, immigration pathways, and public-private partnerships to become a source of high-value AI expertise.

A fifth is the alliance strategy. A country does not attempt full independence, but joins trusted networks for chips, cloud, cybersecurity, data governance, and model evaluation.

For executives, middle-power strategies matter because they create new operating environments. Some markets will become attractive because they offer regulatory clarity, energy capacity, public incentives, or trusted data regimes. Others may become harder to operate in because of localization rules, procurement restrictions, or political alignment pressures.

Three Scenarios for the AI Geopolitical Environment

Executives should prepare for multiple possible futures rather than assume one dominant outcome.

The first scenario is managed interdependence. In this world, geopolitical tension remains high, but countries preserve selective collaboration. Export controls continue, but they are targeted. Standards remain partially interoperable. Companies can operate globally, but must comply with more complex rules. This is the most commercially favorable scenario, but it requires disciplined compliance and flexible architecture.

The second scenario is competitive fragmentation. In this world, AI ecosystems separate more sharply. Data, models, chips, cloud services, and standards become aligned with political blocs. Cross-border collaboration declines. Companies must duplicate systems across regions, localize AI operations, and manage higher costs. This scenario favors firms with strong regional strategies and modular technology systems.

The third scenario is strategic shock. In this world, a geopolitical event, cyber incident, military crisis, supply-chain disruption, or regulatory move suddenly changes access to critical AI inputs. A chip supply route closes. A cloud service becomes restricted. A model is subject to export controls. A data-transfer rule changes. A critical material becomes unavailable. This scenario rewards firms that have contingency plans before they are needed.

The purpose of scenario planning is not prediction. It is preparation. Leaders should ask what assumptions their AI strategy depends on and which of those assumptions are vulnerable to geopolitical disruption.

An Executive Playbook for Geopolitically Aware AI Strategy

Companies can build resilience through five practical actions.

First, map the AI exposure stack. Leaders should identify the critical dependencies behind their AI systems: chips, cloud providers, data centers, software tools, models, data sources, vendors, talent pools, and jurisdictions. Most firms cannot manage risks they have not mapped.

Second, classify AI use cases by geopolitical sensitivity. Not every use case carries the same risk. Internal productivity tools may be less exposed than defense applications, cross-border customer analytics, regulated financial models, healthcare data systems, or critical infrastructure AI. Sensitive use cases require stronger governance and contingency planning.

Third, diversify where it matters. Redundancy is expensive, so it should be targeted. Companies should identify which AI dependencies are mission-critical and where alternatives are needed. This may include multi-cloud capacity, regional hosting options, supplier diversification, or fallback models.

Fourth, design for jurisdictional flexibility. AI architecture should be modular enough to comply with different regulatory environments. Data localization, model governance, audit trails, human oversight, and access controls should be built into systems early.

Fifth, create a geopolitical review function for AI. This does not need to become bureaucratic. It can be a cross-functional group involving strategy, legal, technology, risk, security, procurement, and business leadership. Its role is to review major AI investments through a geopolitical lens.

The core question should be simple: if the external environment changes, can this AI strategy still operate?

Collaboration Without Naivety

Even in a fragmented world, collaboration will remain necessary. AI progress depends on research, talent mobility, open-source tools, standards, safety evaluation, commercial partnerships, and global markets. Complete technological isolation would be costly for most countries and companies.

But collaboration now requires more discipline. Firms must know which collaborations are low risk, which require safeguards, and which could become restricted. They must understand how intellectual property, model access, data sharing, and technical knowledge move across borders.

The best posture is neither naïve openness nor reflexive isolation. It is selective, governed collaboration.

Executives should build partnerships that expand capability while protecting critical assets. They should participate in standards conversations, engage regulators early, monitor export-control developments, and maintain visibility into supplier and vendor exposure.

AI advantage will increasingly depend on the ability to collaborate across borders while remaining resilient if those borders harden.

The Strategic Imperative

The geopolitics of AI will not remain outside the firm. It will enter through supply chains, cloud contracts, data policies, vendor relationships, talent markets, investment decisions, cybersecurity requirements, customer expectations, and regulatory approvals.

Companies that recognize this early will make better strategic choices. They will understand where their AI ambitions depend on fragile external conditions. They will build flexibility into systems. They will choose partners more carefully. They will manage compliance as a strategic capability. They will treat trust, infrastructure, and sovereignty as part of competitive advantage.

Companies that ignore AI geopolitics may still move quickly in the short term. But speed without resilience can become exposure. A strategy built on assumptions of open access, stable rules, unlimited compute, and frictionless global deployment may not survive the next phase of competition.

The central lesson is straightforward: AI is becoming both an economic engine and a geopolitical instrument. Executives should therefore manage it with the same seriousness they apply to capital, supply chains, regulation, and market entry.

The firms that win will not be those that simply adopt AI fastest. They will be those that build AI strategies capable of operating across uncertainty, fragmentation, and strategic competition.

In the AI era, competitive advantage will belong to organizations that understand not only the technology, but the world in which the technology must operate.