Power, Chips, Sovereignty: The Geopolitical Price Shock Reshaping AI

A new study published in the MDPI journal World argues that the real contest in artificial intelligence (AI) is moving beneath the surface: into power grids, semiconductor supply chains, cloud regions, data centres, and sovereign infrastructure. Authored by Victor Frimpong and Ortopah Kojo Botchey of SBS Swiss Business School, the paper introduces a timely concept, geopolitical repricing, to explain how instability is changing the economic value, risk profile, and strategic importance of AI infrastructure.

AI infrastructure is becoming too energy-intensive, capital-heavy, and geopolitically exposed to be treated as ordinary digital plumbing. As conflict, sanctions, energy volatility, chip restrictions, and infrastructure concentration reshape global markets, firms and governments are being forced to reassess where AI systems are built, how they are powered, who controls them, and what risks are priced into their future.

AI's physical backbone is becoming strategic territory

The study challenges the convenient fiction that AI is a frictionless digital technology. Large models and cloud services depend on massive computing capacity, specialized hardware, reliable electricity, cooling systems, data centres, and global networks. The paper notes that data centres have become significant electricity consumers, and AI workloads are expected to intensify that demand. Training large models can also generate substantial carbon emissions, underlining AI's dependence on physical resources rather than software alone.

AI is no longer only a matter of innovation strategy, ethics, or productivity. It is also an infrastructure question. Countries that lack stable power systems, advanced chips, resilient cloud capacity, or secure data-centre ecosystems may find themselves structurally disadvantaged in the next phase of AI development.

The authors define it as the process by which firms, investors, governments, and infrastructure operators reassess the economic value, risk profile, and strategic significance of AI infrastructure in response to geopolitical instability. In other words, the value of an AI facility is no longer determined only by compute output or operating cost. It is also shaped by exposure to energy shocks, supplier concentration, jurisdictional risk, sanctions, infrastructure vulnerability, and national-security priorities.

Energy volatility is turning compute into a political-economy problem

The first transmission channel is energy. AI systems require large and dependable electricity supplies, and the expansion of AI data centres is already influencing grid planning and infrastructure investment. The paper argues that geopolitical events in energy-producing regions can raise electricity uncertainty and operating costs for cloud providers and data-centre operators.

The study uses the Strait of Hormuz as an illustrative case. Around 20 million barrels of oil per day passed through the strait in 2024, representing about 20% of global petroleum consumption. Disruptions in such a chokepoint do not have to target AI directly to affect AI economics; they can alter energy costs, insurance costs, market expectations, and long-term infrastructure planning.

For governments, this creates a new link between AI competitiveness and energy policy. Stable, affordable, and increasingly low-carbon electricity will become a strategic advantage. For companies, data-centre location decisions will be judged not only by tax incentives, connectivity, and land availability, but also by grid reliability, energy security, and long-term power-purchase arrangements.

AI ambition without energy resilience risks becoming an expensive promise. Countries seeking to build AI capacity will need to think as much about transmission lines, renewable power, storage, and grid governance as they do about coding talent or AI startups.

Chip chokepoints and cloud concentration are raising the risk premium

The second channel is supply-chain disruption. AI infrastructure depends heavily on advanced semiconductors and specialized computing hardware, whose production and inputs are geographically concentrated. Export controls, sanctions, strategic rivalry, and conflict can restrict access, extend lead times, and raise costs, slowing the expansion of AI capacity.

The policy consequence is already visible in how states and firms are thinking about sovereign AI, domestic compute capacity, hardware access, and supply-chain resilience. The paper argues that companies may respond through diversification, stockpiling, vertical integration, and supplier-risk management. These strategies can improve resilience, but they also raise capital intensity and reduce the efficiency gains that once defined globalized digital infrastructure.

The third channel is infrastructure vulnerability. AI infrastructure is concentrated in data centres, cloud regions, and network ecosystems. Concentration brings efficiency, but it also creates exposure: grid overload, cyberattacks, physical disruption, regional dependency, and cloud-region failure.

The study's key insight is that these risks reinforce one another. Energy insecurity can make data centres more fragile; semiconductor dependence can limit recovery options; infrastructure concentration can magnify the effects of both. The result is a higher risk premium around AI infrastructure, where investors and policymakers begin to value redundancy, control, and resilience alongside speed and scale.

The AI race is shifting from efficiency to resilience

The framework maps a four-stage process: geopolitical shock, transmission through energy volatility, supply-chain disruption and infrastructure vulnerability, revaluation of AI infrastructure, and strategic reallocation. The repricing layer includes increased compute costs, higher risk premiums, higher capital expenditure, and reduced reliability expectations. The final stage points toward sovereign AI investment, infrastructure diversification, resilience, and security-aligned development.

The study explains why states and firms may accept higher costs to gain control. In a stable global system, the cheapest and most efficient cloud or chip supply may win. In a fractured geopolitical environment, the more resilient and controllable system may become more valuable.

The shift has major governance implications. AI governance cannot remain focused only on model behavior, transparency, bias, and accountability. Those questions still matter, but the paper argues that governance must also include infrastructure-level risk assessment, geopolitical stress testing, energy dependence, supply-chain concentration, and cloud exposure.

The authors note that the study is a conceptual paper, not an empirical test. The framework has not been validated through case studies or quantitative analysis, and its propositions should be treated as theoretical expectations rather than established facts. They also caution that not every geopolitical event leads to repricing; repricing occurs when instability becomes significant enough to affect investment decisions and long-term valuation.

The concept gives researchers, policymakers, and investors a language for something already becoming visible: AI infrastructure is being valued through a geopolitical lens. The next phase of AI competition will not be decided only in research labs or product launches. It will also be decided in energy ministries, chip supply chains, investment committees, cloud contracts, and national-security strategies.