Global discourse on artificial intelligence frequently centers on models: which systems are fastest, most capable, or most cost-effective. The emergence of inexpensive, high-performing models from China—developed by companies such as DeepSeek, z.ai, and Moonshot—has intensified this focus, fueling suggestions that the gap with U.S. industry leaders is narrowing.

Yet the rise of these capable, low-cost models obscures a more fundamental reality: as long as the United States controls the underlying infrastructure that AI ecosystems depend on, it will remain dominant.

Beyond Models and Chips

It is tempting to frame the AI race as a head-to-head contest between models—Anthropic's Fable versus OpenAI's GPT, or against DeepSeek V4 and Moonshot's Kimi K3—or as a hardware story centered on the specialized chips used to train and run them. Frontier AI, however, rests on a far broader and more capital-intensive foundation: hyperscale data centers, cloud computing platforms, AI servers, and the submarine fiber-optic cables that interconnect them. These layers of physical infrastructure remain deeply exposed to American export controls, extraterritorial data-extraction laws, and the spillover effects of an expansive commercial-military partnership. Since 2022, U.S. export controls have repeatedly tightened restrictions on the sale of advanced AI chips to China, demonstrating how control over physical chokepoints translates directly into strategic leverage—regardless of which lab produces the most capable model.

Nvidia's AI Ecosystem and American Tech Hegemony

Nvidia's business model illustrates how the U.S. wields control over the AI ecosystem. The world's most valuable company derives its strength from its graphics processing units—the chips essential for training frontier AI models—which command roughly 85% of the global market (Motley Fool).

Nvidia's ecosystem extends to hardware manufacturers such as Broadcom, as well as the cloud hyperscalers—Amazon Web Services, Google Cloud, and Microsoft Azure—that supply the infrastructure powering foundational AI developers including Anthropic, OpenAI, Meta, and Alphabet. Nvidia's software platform, CUDA, binds this ecosystem together. Launched in 2007, CUDA has accumulated over a decade and a half of optimized libraries, developer tooling, and community knowledge, making it the default development environment for AI and imposing prohibitive switching costs on any organization seeking to migrate to a competitor.

This constellation of tech giants constitutes the core of a new U.S. AI industrial complex, one with deep ties to the country's defense and intelligence apparatus through large-scale binding contracts. The Department of Defense has signed standard operational agreements enabling major providers—including Google, OpenAI, Microsoft, Amazon Web Services, Oracle, and Nvidia—to deploy their frontier AI tools on classified military networks (DoD).

The Pentagon's FY2027 budget earmarks more than $54 billion for autonomous warfare and drone systems, funding a newly established Defense Autonomous Warfare Group (DAWG) (The Cipher Brief). Despite earlier objections about how their technologies might be deployed, OpenAI, xAI, and Google have each signed binding contracts granting the Pentagon "all lawful use" for defense-related purposes, including autonomous weapons and mass surveillance.

Even as Anthropic pursues litigation against the U.S. government over objections regarding how its AI models may be used, the company has reportedly embedded engineers inside the National Security Agency to adapt its Mythos model for offensive cyber operations, potentially targeting networks in China and Iran (Axios).

The scale of the commercial-military alignment between Washington and Silicon Valley is extraordinary. Ten of the world's largest companies by market capitalization are the very tech firms that comprise this ecosystem—nearly all American—representing a commercial concentration measured in the trillions of dollars. Federal AI contracting surged by $90.7 billion in 2026 alone, according to the Brookings Institution (Brookings).

Hyperscalers and Hard Infrastructure

As American cloud hyperscalers expand data center capacity worldwide, they are increasingly constructing their own privately owned subsea fiber-optic networks. Meta plans to build a globe-spanning subsea cable covering 40,000 kilometers at a projected cost of $10 billion (TechCrunch). Google, through its Pacific Connect Initiative, will invest over $1 billion to extend connectivity from Japan to the South Pacific (Submarine Networks).

These critical data pipelines—owned and operated by Silicon Valley giants—account for 70% of usable undersea cables in 2026 (Submarine Networks). Washington retains the authority to restrict where these networks terminate and who may access them. In 2020, U.S. regulators blocked the Hong Kong segment of the Pacific Light Cable Network—a project backed by Google and Meta—citing Chinese espionage concerns and forcing the companies to abandon the direct U.S.-Hong Kong link. Approximately 13,000 kilometers of already-laid cable was left unused on the ocean floor (Silicon.co.uk).

Even China's frontier AI labs, which train models on domestically hosted infrastructure, remain dependent on American-owned undersea cables whenever their systems interact with the global internet—whether sourcing training data, serving overseas users, or running APIs outside China. China is now racing to construct parallel subsea fiber-optic networks along its so-called digital silk road (JuanCole.com), seeking to reduce its reliance on and exposure to American dominance over the physical infrastructure underpinning the global AI landscape.

AI hyperscalers also remain subject to a range of U.S. extraterritorial data laws. The U.S. CLOUD Act, for instance, compels U.S.-headquartered cloud providers to produce data within their possession, custody, or control—even when that data is stored outside the United States (Wikipedia). Subject to legal process, U.S. authorities may obtain not only stored data but also a detailed digital trail of AI activity: users' prompts, model responses, access logs detailing who used a system and when and where, and technical records revealing behavioral patterns. This authority, if exercised, grants Washington extraordinary leverage over the AI ecosystem.

Limitations of U.S. Dominance

U.S. dominance of the AI stack is not absolute. Manufacturers in Asia produce the semiconductors, AI servers, and other components essential to the industry. Nvidia's supply chain runs through TSMC, SK Hynix, and Samsung, alongside system integrators such as Foxconn, Quanta, and Wistron.

In the past month, Nvidia CEO Jensen Huang announced billions of dollars in new investments and contracts spanning Taiwan and South Korea. In Taiwan, Nvidia secured orders for advanced chips and packaging from TSMC, as well as servers and networking hardware from Quanta Computer and other suppliers. In Korea, the company signed billion-dollar agreements with memory-chip manufacturers SK Hynix and Samsung, along with new partnerships involving LG, Hyundai Motor Group, and Doosan Robotics.

Taiwan and South Korea, however, are unlikely to weaponize their positions in the AI value chain. Both countries have a strong interest in preserving their earnings within what is effectively a friend-shored U.S. technology stack. The 2022 CHIPS and Science Act, which committed $52 billion in subsidies to domestic semiconductor manufacturing, further anchored this alliance by giving allied chipmakers financial incentives to expand production in the United States rather than pursue independent strategies.

The world may ultimately see competing AI ecosystems emerge—one led by the United States and another by China—each with distinct standards, infrastructure, and governance frameworks. But until such an alternative materializes, the United States will continue to maintain a firm grip on the broader AI ecosystem.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune. This story was originally featured on Fortune.com.