TL;DR — AI has become the defining axis of geopolitical competition. The US leads in frontier models and chip design but depends on TSMC for manufacturing. China trains competitive models at a fraction of the cost despite export controls. The EU regulates comprehensively while its AI industry falls further behind. The outcome will reshape global economic power for decades.
Intelligence as National Interest
For most of its history, artificial intelligence was a research curiosity debated in academic conferences. That era is over. AI has become a matter of national security, economic competitiveness, and geopolitical positioning on par with nuclear technology or space capability. Governments now treat AI leadership the way they once treated steel production or semiconductor fabrication—as a strategic asset too important to leave to market forces alone.
The stakes are not abstract. The nation or bloc that leads in AI will likely dominate the industries of the next several decades: autonomous systems, drug discovery, materials science, financial services, defense. Falling behind in AI risks falling behind in everything AI touches, which increasingly means everything.
The American Position
The United States enters this competition from a position of significant but precarious strength. American companies—OpenAI, Anthropic, Google, Meta—build the world’s most capable frontier models. American chip designers, primarily NVIDIA and AMD, dominate the architectures that power AI training and inference. American venture capital funds the majority of global AI investment, with 79% of AI funding flowing to US companies in 2025 — fueling the winner-take-all wealth concentration that defines this era.
The CHIPS and Science Act, signed in 2022 and now deploying its $52.7 billion in semiconductor subsidies, represents the most aggressive industrial policy the US has pursued in decades. TSMC is building fabrication plants in Arizona. Intel is expanding domestic capacity. Samsung has committed to facilities in Texas. The goal is reducing dependence on Taiwanese manufacturing for the chips that underpin AI capability.
But the US position has structural vulnerabilities. Despite the CHIPS Act investment, the US still manufactures less than 12% of the world’s semiconductors. TSMC’s most advanced fabrication processes remain in Taiwan, geographically proximate to China and vulnerable to disruption. American AI companies depend on a global talent pipeline—roughly 60% of AI PhD researchers in the US are foreign-born, and immigration policy directly affects research capacity.
The Numbers — The US captures 79% of global AI funding and leads in frontier model development. But it manufactures less than 12% of the world’s semiconductors, and roughly 60% of AI PhD researchers in the US are foreign-born. Leadership is real but structurally fragile.
China’s Asymmetric Response
The prevailing assumption in Washington has been that export controls would cripple Chinese AI development. The evidence tells a more complicated story.
The Bureau of Industry and Security’s October 2022 controls, expanded in 2023 and again in 2024, restricted China’s access to advanced AI chips (NVIDIA’s A100, H100, and subsequent generations) and the equipment to manufacture them. These controls were the most significant technology restrictions imposed on a major economy since the Cold War.
China’s response has been instructive. DeepSeek-R1, released in early 2025, briefly surpassed ChatGPT as the top iOS app and caused an 18% single-day drop in NVIDIA’s stock. The model, reportedly trained for roughly $6 million compared to over $100 million for GPT-4, demonstrated that algorithmic innovation can partially offset hardware disadvantage. Chinese researchers, constrained on compute, focused relentlessly on training efficiency—and the results startled the industry.
China’s broader strategy operates on multiple fronts. Domestically, Huawei’s Ascend chips provide an alternative AI hardware ecosystem, though still generations behind NVIDIA’s performance. State-backed investment channels billions into AI champions. The civil-military fusion doctrine ensures commercial AI advances flow immediately into defense applications. And critically, China generates enormous quantities of training data from its massive digital economy—a resource that export controls cannot restrict.
Key Takeaway — Export controls forced China to innovate on efficiency rather than scale. DeepSeek trained a competitive model for roughly $6M versus $100M+ for GPT-4. Constraints bred adaptation, not surrender. The controls slowed China’s frontier capability but did not stop it.
Europe’s Regulatory Gambit
The European Union has chosen a fundamentally different approach to AI competition. Rather than racing to build frontier models or cornering chip supply, Europe has positioned itself as the world’s AI regulator. The EU AI Act, which began phased enforcement in 2025, represents the most comprehensive AI regulation anywhere in the world.
The Act categorizes AI systems by risk level—unacceptable, high, limited, and minimal—and imposes corresponding requirements. High-risk systems face mandatory conformity assessments, transparency obligations, and human oversight requirements. Foundation model providers must disclose training data summaries, demonstrate copyright compliance, and publish energy consumption metrics.
Whether this regulatory leadership translates into economic advantage or competitive handicap remains genuinely uncertain. Proponents argue that clear rules create market certainty, build consumer trust, and establish standards that other jurisdictions eventually adopt—the “Brussels Effect” that previously shaped global privacy law through GDPR. The EU’s Single Market of 450 million consumers gives its regulations gravitational pull regardless of where AI companies are headquartered.
Critics counter that Europe is regulating an industry it doesn’t have. No European company ranks among the top ten AI model developers. European AI startups raised roughly 10% of global AI venture capital in 2025. Mistral, France’s most prominent AI company, remains small relative to American and Chinese competitors. The risk is that regulation raises barriers to entry for European challengers while merely adding compliance costs for American incumbents with the resources to absorb them.
The Numbers — European AI startups raised roughly 10% of global AI venture capital in 2025. No European company ranks in the top ten AI model developers. The EU AI Act is the world’s most comprehensive AI regulation—but it governs an industry Europe largely does not control.
The Semiconductor Chokepoint
Beneath the model-level competition lies a more fundamental contest over semiconductor manufacturing. Advanced AI chips require fabrication processes at the cutting edge of physics—currently 3-nanometer and moving toward 2-nanometer nodes. Only three companies worldwide can manufacture at these scales: TSMC, Samsung, and Intel. TSMC alone commands over 60% of the global foundry market and an estimated 90% of the most advanced process nodes.
This concentration creates a single point of fragility for the entire global AI ecosystem. TSMC’s most advanced fabs sit in Hsinchu, Taiwan—roughly 100 miles from mainland China. Any disruption to Taiwanese semiconductor production, whether from conflict, natural disaster, or political pressure, would immediately constrain global AI capability.
The scramble for semiconductor sovereignty reflects this vulnerability. The US CHIPS Act, Europe’s European Chips Act (committing 43 billion euros), Japan’s semiconductor investment program, and South Korea’s $471 billion chip industry plan all aim to reduce dependence on concentrated manufacturing. But building advanced fabrication capacity takes years, costs tens of billions per facility, and requires specialized expertise that cannot simply be purchased.
NVIDIA’s position within the chip ecosystem adds another layer of concentration — one we examine in detail in our chip wars analysis and AI chip landscape overview. The company shipped an estimated 4 to 5 million AI chips in 2025, consuming 63% of TSMC’s advanced CoWoS packaging capacity. NVIDIA’s CUDA software ecosystem creates switching costs that make alternatives difficult even when they offer competitive hardware. This combination of hardware dominance and software lock-in gives a single American company extraordinary influence over global AI capability.
Middle Powers and Emerging Players
The AI competition extends well beyond the three major blocs. Nations across the Gulf, South Asia, and Southeast Asia are staking positions with varying strategies and resources.
The United Arab Emirates and Saudi Arabia have emerged as significant AI investors, leveraging sovereign wealth to fund both domestic capability and partnerships with American and Chinese AI companies. The UAE’s Technology Innovation Institute developed Falcon, one of the more capable open-source models to emerge from outside the US-China axis. Gulf states view AI as central to their economic diversification strategies beyond oil dependence.
India occupies a distinctive position. Its massive talent pipeline—Indian nationals represent a disproportionate share of AI researchers globally—contrasts with relatively modest domestic AI investment. India’s Digital Public Infrastructure approach, which built population-scale systems for identity, payments, and data exchange, could provide a foundation for AI deployment at scale that few other nations can match.
Japan and South Korea, despite their technological sophistication, have found themselves competing from behind in AI. Both nations are channeling substantial investment into catching up, with particular focus on semiconductor manufacturing capacity and applied AI for their aging populations.
Key Takeaway — AI competition is not a two-player game. Gulf states invest sovereign wealth in AI diversification. India contributes disproportionate talent but underinvests domestically. Japan and South Korea leverage hardware expertise while racing to close the model gap.
The Talent Dimension
Perhaps the most overlooked dimension of AI competition is human capital. The world contains a limited number of researchers capable of advancing frontier AI—estimates range from roughly ten thousand to fifty thousand, depending on how the boundary is drawn. Where these people live and work matters enormously.
The United States has historically attracted a disproportionate share of global AI talent, with leading universities and well-funded labs serving as magnets. But this advantage depends on immigration policy, compensation competitiveness, and the intangible appeal of working at the frontier. Restrictive visa policies, geopolitical tensions, and rising compensation in competing hubs all threaten the talent pipeline.
China has invested heavily in domestic AI education, graduating more STEM PhDs annually than any other country. But top Chinese researchers frequently cite limited academic freedom and compute access as factors driving them toward foreign institutions. The net flow of elite AI talent still favors the United States, though the margin has narrowed.
What the Race Means
The global AI competition carries economic implications that extend far beyond the technology sector. The nation that leads in AI capability will likely set standards, control chokepoints, and capture disproportionate value from the industries AI transforms. The semiconductor restrictions already demonstrate how technology leadership translates into geopolitical leverage—the ability to constrain a rival’s economic development through export controls. Meanwhile, AI regulation approaches diverge across jurisdictions, adding another dimension to the competition.
For businesses, the geopolitical dimension of AI creates practical considerations. Supply chain resilience now requires accounting for potential chip access disruptions. Regulatory compliance differs dramatically across jurisdictions. Talent strategies must navigate immigration uncertainty. The assumption that AI technology flows freely across borders no longer holds.
For the broader economy, the outcome of this competition will shape whether AI’s economic benefits distribute broadly or concentrate in the nations and companies that control the technology’s critical infrastructure. The arms race is not just about who builds the best model. It is about who controls the supply chains, sets the rules, and captures the value of the most transformative technology of the century.
Sources
- Bureau of Industry and Security export control regulations
- CHIPS and Science Act implementation data
- European Commission AI Act documentation
- Semiconductor Industry Association reports
- Stanford HAI AI Index 2025
- Company filings and public statements
- Brookings Institution geopolitical AI research
