Artificial Intelligence and the New Great Power Struggle: The Geopolitics of Data, Algorithms, Semiconductors, and Energy
# Prem Sagar Poudel
The power politics of the 21st century has entered a phase of fundamental transformation. In the previous century, the great power struggle primarily revolved around territory, sea lanes, energy resources, military alliances, trade routes, and the financial system. Today, all these elements remain relevant, but the decisive center of the balance of power is becoming more subtle, invisible, and technology-dependent. The time for viewing artificial intelligence (AI) as a singular technology is over. Today’s question is not merely which nation has a large army; it is also which nation possesses cutting-edge chips, who controls data, who builds algorithms, who runs cloud infrastructure, who secures energy supply, and who determines global digital standards.
To understand this transformation, the historical lessons of Cold War-era technological competition are instructive. Just as the space race and nuclear technology defined the power structure of the 20th century, AI, semiconductors, and data are laying the foundation of a new world order in the 21st century. However, there is one crucial difference. During the Cold War, technological competition was largely confined to two superpowers; in today’s digital age, alongside states, multinational technology companies such as Google, Microsoft, Meta, OpenAI, ByteDance, Tencent, and Huawei have emerged as independent centers of power. These non-state actors, through their algorithms, cloud services, and platforms, are challenging the traditional sovereignty of the state. For this very reason, AI is no longer just a subject for the tech industry; it has become a core question of national security, economic competitiveness, ideological influence, the labor market, governance systems, and the international balance of power.
There was a time when oil was called the lifeblood of the global economy. Today, data is increasingly seen in that same role. But this analogy is incomplete and misleading. Oil is extracted, refined, and consumed; its use is a one-time affair. Data is continuously produced, collected, analyzed, reused, and employed to shape political, economic, and social behavior. Oil becomes increasingly scarce; data expands with every passing moment. The value of oil ends upon consumption; the value of data multiplies through reuse and recombination. These fundamental differences underscore the need to view data beyond the oil analogy. Data is not merely the new oil; it is the new sovereignty. Whoever controls data related to citizens, markets, geolocation, health, education, financial behavior, communication, and security can exert a profound influence over future policy, the economy, and society.
In this context, understanding the internal differentiation of AI is essential. There is a vast difference in the geopolitical risks and opportunities presented by Narrow AI (designed for specific tasks), Generative AI (which creates new content), Foundation Models (massive pre-trained systems), and AGI (human-level general intelligence, which remains a concept for now). Narrow AI primarily impacts economic productivity and military capability. Generative AI and Foundation Models are reshaping the information ecology, cultural influence, and the power balance in knowledge production. The potential of AGI could challenge all traditional assumptions of strategic stability. These different types of AI have distinct impacts on geopolitics, and this differentiation must be considered in policymaking.
In the AI era, the definition of sovereignty has expanded. Traditional sovereignty was tied to territory, borders, the military, currency, and law. The new digital sovereignty is tied to data storage, cloud infrastructure, AI models, cybersecurity, chip supply, digital payments, citizen identification systems, social media, information flows, and algorithmic decision-making processes. The most serious aspect is that a country can be politically independent yet entirely dependent technologically. If its government data sits on foreign clouds, its financial transactions depend on foreign platforms, its communications are shaped by foreign algorithms, its security systems rely on foreign equipment, and its policymaking depends on foreign data analysis, then its real sovereignty is incomplete. This is why the concept of “Strategic Autonomy” is becoming increasingly important globally. It seeks a path that avoids both complete isolation and blind dependence, maintaining decision-making capacity while building selective interdependence.
The global AI power struggle is currently unfolding on three interconnected levels. The first level is the struggle for chips and computing power. Developing advanced AI models requires cutting-edge semiconductors, Graphics Processing Units (GPUs), data centers, electricity supply, and trained engineers. This is why the United States, China, Europe, Japan, South Korea, Taiwan, and India are investing heavily in semiconductor and AI infrastructure. In reality, however, the semiconductor supply chain is so complex and interdependent that no single country can fully control it. Dutch lithography (ASML), Japanese chemicals, Taiwanese manufacturing (TSMC), American design software (Cadence, Synopsys), and British ARM architecture are all interwoven. The objective of US export controls is to limit China’s access to advanced AI chips and semiconductor equipment. China has emphasized building a self-reliant chip industry, domestic AI models, and a national technology infrastructure. But complete self-reliance in the chip industry is nearly impossible. This is not merely a commercial competition; it is an attempt to control the critical chokepoints of the supply chain. It is a battle to control the foundation of future military, industrial, financial, and knowledge power.
The second level is the struggle for data and algorithmic dominance. The power of an AI model does not depend solely on chips; it requires vast, diverse, high-quality, and legally usable data. Data available in the English language disproportionately dominates the global digital knowledge structure. This has a profound impact on the linguistic, cultural, and ideological biases of AI models. Minor languages, local knowledge systems, indigenous histories, the experiences of the Global South, Eastern philosophies, and non-Western political concepts are not adequately represented in AI training. Consequently, future AI systems not only risk understanding the world from an unequal linguistic and cultural perspective but also threaten to automatically reproduce and reinforce that bias. This is not just a technical problem; it is a core question of epistemic justice.
The third level is the struggle for governance and standard-setting. Differing global perspectives exist on how to regulate AI. Europe has prioritized risk-based regulation through the EU AI Act, emphasizing human rights and transparency. The United States is seeking a balance between innovation, the private sector, and national security. China has attempted to simultaneously advance AI development, social stability, state capacity, and governance control. India has presented itself with a distinct identity, emphasizing public digital infrastructure, local innovation, language technology, and digital inclusion. The United Nations, OECD, G20, and the Global Partnership on Artificial Intelligence (GPAI) are searching for common principles of global AI governance. The real problem, however, is that the pace of technology far outstrips the pace of regulation, and these multilateral forums themselves have become arenas of power competition.
The greatest paradox of the AI era is this: technology is global, but governance is national. Data flows across borders, but law is written within them. AI models are built by multinational companies, but their social impact is borne by local communities. The chip supply chain is global, but export controls are imposed based on national security. The source of a cyberattack can be ambiguous, but its political fallout manifests in tension between states. For this reason, AI governance cannot be resolved by domestic law alone; a new coordination between multilateral, regional, and national levels is needed. In this debate, the power of multinational technology companies cannot be ignored. The market capitalization of some companies is larger than the GDP of many nations. They are giving concrete form to concepts like algorithmic sovereignty, data-sovereignty, and platform sovereignty. The conflict between state law and corporate terms of service has created a new governance crisis.
AI can offer great opportunities to the global economy in productivity, health, education, agriculture, finance, energy management, and public services. But along with opportunity, inequality may also grow. Those with AI infrastructure, capital, computing power, data, and skilled human resources will quickly reap the benefits. Those without these assets could become mere consumers of technology. This risk is most significant for the Global South. In this context, the danger of AI colonialism is serious. In the past, colonialism operated through territory, trade, minerals, and military control. Today, digital colonialism can evolve through data collection, platform dependency, algorithmic influence, cloud control, technology licenses, intellectual property, and digital financial structures. However, this threat is multidimensional. A small country can be just as dependent on its larger neighbor’s digital infrastructure as it is on a Western tech company. Both types of dependency limit policy autonomy.
But it is also wrong to view AI solely as a threat. It is also a historic opportunity for developing nations. With the right policy, public investment, open data structures, local language models, university-industry collaboration, and ethical regulation, AI can boost education, health, agriculture, disaster management, and economic productivity. In this regard, open-source AI (such as Meta’s Llama, Mistral) and decentralized technologies (like Federated Learning) have provided smaller states with important alternatives to bypass centralized power structures. These technologies decentralize infrastructure ownership and increase the potential for local adaptation.
The struggle for AI and semiconductors is reorganizing the global supply chain. Semiconductor production is excessively concentrated in a few countries and companies. Design, equipment, manufacturing, packaging, chemicals, rare earth minerals, and intellectual property are distributed across different nations. This has turned semiconductors into a new geopolitical weapon. But this weapon is a double-edged sword. Supply chain disruption can cause serious economic damage even to the side wielding it. When access to AI chips is restricted, it is not merely a commercial problem for a company; it is linked to a nation’s research capacity, defense technology, industrial modernization, and scientific competitiveness.
Energy is another decisive factor. Training and operating AI models require massive data centers. Data centers demand enormous amounts of electricity, water, cooling systems, and grid stability. If AI expansion is not coordinated with energy infrastructure, it can lead to pressure on power grids, water resource disputes, and increased carbon emissions. In the AI era, technology policy and energy policy cannot remain separate. Whoever possesses cheap, stable, clean, and sufficient energy can become a leader in AI infrastructure. The future AI competition will depend not only on “who has the chips” but also on “who has the energy.”
Cybersecurity is another sensitive front of the AI era. AI can strengthen cyber defenses, but it also makes attacks faster, more sophisticated, and automated. Disinformation, deepfakes, election interference, financial fraud, social division, automated surveillance, and psychological influence campaigns can be made more complex by AI. The entry of AI into the military domain has added new risks to strategic stability. AI-assisted surveillance, target identification, autonomous systems, drone swarms, cyber warfare, and decision-support systems can accelerate the pace of military competition. The biggest risk is the compression of decision-making time. When leaders have less time to decide during war or a crisis, the probability of miscalculation increases. If nuclear-armed states build excessive dependence on AI-based early warning or military decision systems, the risk of accidents, false signals, or unintended escalation can increase dramatically. Compared to the Cold War, the role of human control in today’s technology-driven military structures is diminishing, making crisis management even more unstable.
Three potential paths for AI governance are visible in the world. The first is techno-nationalism, where each power seeks to isolate its own AI systems, data, and chips as a matter of security. This can boost innovation but carries the risk of global fragmentation and the creation of digital blocs. The second is the market-led path, where private companies are the main drivers and the state follows. This delivers rapid innovation but can weaken public interest, transparency, and labor rights. The third is a human-centered multilateral path, which seeks a balance between innovation, security, public interest, national autonomy, and global coordination. This path may be more stable and just in the long run, but it requires a real restructuring of multilateral institutions, which is impossible without the consensus of major powers.
The AI era represents a historic opportunity for Asia. China possesses manufacturing, data, and technology implementation capabilities. India has vast talent, digital public infrastructure, and a democratic market. Japan, South Korea, and Taiwan have specialized capabilities in semiconductors and high-tech supply chains. ASEAN countries are important for manufacturing, data services, and regional connectivity. If Asia can identify minimal areas of cooperation without allowing internal competition to turn into outright hostility, it can become not just a user, but a manufacturer and standard-setter in the AI era. But the path to Asian cooperation is not smooth. China-India mistrust, US-China technological rivalry, the Taiwan question, South China Sea disputes, and political system differences limit cooperation. Therefore, AI cooperation must begin in practical and technical areas rather than as a broad political slogan. Cooperation is possible in areas such as health AI, climate risk forecasting, language technology, agricultural technology, disaster management, cybercrime control, and AI safety evaluation. Such areas can lay the foundation for trust-building.
The most critical question for the Global South is this: in the AI era, will we become mere consumers of models built by others, or will we become masters of our own knowledge, languages, data, and policy structures? This question is even more acute for small nations. For a country like Nepal, AI is simultaneously an opportunity and a risk. The opportunity lies in AI’s potential to provide great assistance in education, health, tourism, agriculture, disaster preparedness, translation, access to justice, administrative services, and foreign employment management. The risk lies in the fact that if national data, citizen identification, the financial system, media information, language resources, and government services become dependent on foreign platforms, long-term digital dependence can increase. Even if small countries cannot build large AI models, they can safeguard their digital sovereignty through their data policy, language funds, public digital infrastructure, cybersecurity, university research, use of open-source technology, and regional partnerships.
In this context, policymakers must clarify five priorities.
First, a national data strategy. A clear legal framework is needed to determine which data is public, which is sensitive, which is available for research, which can be shared with foreign companies, and which must be kept secure within the country. This requires the establishment of a data classification system, data trusts, and a data protection authority.
Second, local language AI. AI cannot be inclusive without building digital resources in languages like Nepali, Maithili, Bhojpuri, Tamang, Newari, Tharu, Gurung, Limbu, Magar, and others. This requires building language data funds through government-community partnerships, adapting open-source language models, and establishing language technology research centers in universities.
Third, cybersecurity and data protection. In the AI era, protecting citizens’ privacy, financial security, and the state’s sensitive information is mandatory. A national cybersecurity strategy, regular security audits, cyber insurance, and active participation in international cyber cooperation are essential.
Fourth, education and human resources. From school to university, AI literacy, mathematics, computing, ethics, and critical thinking must be integrated into the curriculum. Priority must be given to AI specialization in technical education, research scholarships, and industry-academic collaboration.
Fifth, regional cooperation. Small countries cannot build all infrastructure alone, but they can cooperate regionally on shared research, cloud services, language models, regulation, and cybersecurity practices. The possibility of shared digital infrastructure and policy coordination should be explored through forums like the South Asian Association for Regional Cooperation (SAARC) or the Bay of Bengal Initiative for Multi-Sectoral Technical and Economic Cooperation (BIMSTEC).
To implement these five priorities, an institutional structure is equally necessary. Every country must seriously consider establishing structures such as a National AI Security Council, a Data Policy Commission, an AI Ethics Board, public computing infrastructure, and a local language technology program.
In-depth study on this subject is necessary. Five research priorities can be proposed for think tanks and scholars.
First, the development of an AI Sovereignty Index. A comprehensive indicator to measure a country’s digital autonomy based on chips, cloud, data, language, human resources, policy, cybersecurity, and energy capacity should be built.
Second, the formulation of a Global South AI development model. An alternative development framework must be prepared, focused not on replicating the Western private-company-dominated model, but on public-private-community partnerships, open-source technology, linguistic diversity, and social justice.
Third, the creation of an AI and geopolitical risk map. An analytical map should be developed to assess conflict potential, dependency indicators, and risk assessment of AI power competition on a regional basis.
Fourth, the preparation of a minimum digital sovereignty package. A minimum standard document must be created detailing the data policy, language fund, cybersecurity, public cloud, and legal framework that small nations must mandatorily build.
Fifth, the integration of local knowledge into AI governance. In-depth research is needed on how to incorporate local communities, languages, cultures, indigenous knowledge systems, and social justice into AI governance.
Artificial intelligence is not just a new industrial revolution; it is a new geopolitical revolution. It is a rewriting of global economic productivity, military power, social structures, governance systems, language, culture, energy, and sovereignty. Whoever understands this in time will play an active role in the future balance of power. Whoever sees it merely as software or a commercial tool will become only a consumer of a digital structure built by others in the future. Cold War history teaches us that the future of nations left behind in the technology race is determined by others. Therefore, an AI policy is not a luxury for any nation; it is a strategic imperative.
At the international level, the necessary approach is a balance of competition and cooperation. Complete isolation is not possible; complete dependence is also unsafe. The wise path is a balance between strategic autonomy and multilateral cooperation. Power has always shifted in history, but in every era, not everyone has the capacity to recognize it. Today’s power is invisible, but it is not weak. It is hidden in data, algorithms, semiconductors, energy, and regulations. It resides in the server farms of Google, Microsoft, and ByteDance. It is concretized in ASML’s lithography machines and TSMC’s fabs. It is woven into the neural networks of OpenAI’s GPT and Meta’s Llama models. Whoever understands this, understands the 21st century. Whoever ignores it will lag behind in the future world order.
In this era, the greatest challenge for any nation is to write its own digital destiny. Benefit can be derived from others’ technology, but long-term freedom cannot be preserved by being wholly dependent on others’ algorithms. Openness is needed, but not blind dependence. Innovation is needed, but with social responsibility. Security is needed, but not by destroying civil liberties. Global cooperation is needed, but not by abolishing national autonomy. The freedom of the future must be safeguarded not only at border outposts but also in servers, chips, code, the cloud, language models, and algorithms. Therefore, the debate must now center on the capacity to understand and control technology. The decision of whether we make AI our servant or become the market for AI lies in the hands of today’s policymakers, scholars, tech entrepreneurs, and political leadership.
About the Author: Prem Sagar Poudel is a senior journalist and international relations analyst from Nepal. He has conducted in-depth studies on Nepal-China relations, the geopolitics of the Himalayan region, and Asian security issues.
