The global debate over artificial intelligence often centers on which AI models are fastest, most capable, or least expensive. Recent advances from Chinese companies like DeepSeek, z.ai, and Moonshot have intensified this focus, creating the impression that the gap between U.S. and Chinese AI development is closing. However, the true foundation of AI dominance lies not in the models themselves but in the underlying infrastructure that supports them.

The United States maintains a commanding lead in this infrastructure, which includes hyperscale data centers, cloud computing networks, AI servers, and underwater fiber-optic cables. These components are critical to training and deploying advanced AI systems, and they remain heavily influenced by U.S. policies. American export controls, extraterritorial data laws, and deep ties between the tech industry and the military create a system that reinforces U.S. control over global AI ecosystems.

A key example of this dominance is Nvidia, the world’s most valuable company, which produces the graphics processing units essential for training AI models. Nvidia holds roughly 85% of the global market for these chips. Its influence extends beyond hardware to include a software platform called CUDA, which has become the default environment for AI development worldwide. This creates high barriers for competitors attempting to enter the market.

Nvidia’s ecosystem includes major cloud providers like Amazon Web Services, Google Cloud, and Microsoft Azure, which supply the infrastructure for leading AI developers such as Anthropic, OpenAI, Meta, and Alphabet. These companies, along with Nvidia, form the core of what has been described as a new U.S. AI industrial complex. Many of them have deep ties to the U.S. defense and intelligence communities through large, binding contracts. The Pentagon’s 2027 budget allocates over 54 billion dollars (approximately 49 billion euros) for autonomous warfare and drone systems, funding a new Defense Autonomous Warfare Group. Major AI companies, including OpenAI, xAI, and Google, have signed agreements allowing the Pentagon broad use of their technologies for defense purposes, including autonomous weapons and mass surveillance.

Despite ongoing litigation from Anthropic over how its AI models might be used, reports indicate the company has embedded engineers within the National Security Agency to adapt its systems for offensive cyber operations, potentially targeting networks in China and Iran. The scale of collaboration between Washington and Silicon Valley is vast, with the ten largest companies by market capitalization nearly all being American tech firms. Federal AI contracting surged by 90.7 billion dollars (approximately 82 billion euros) in 2026 alone.

American tech companies are also expanding their control over global data infrastructure. Cloud providers are building privately owned underwater fiber-optic networks to connect data centers worldwide. Meta plans to construct a 40,000-kilometer (approximately 24855 miles) subsea cable at an estimated cost of 10 billion dollars (approximately 9 billion euros), while Google’s Pacific Connect Initiative will invest over 1 billion dollars (approximately 907 million euros) to enhance connectivity between Japan and the South Pacific. These networks, owned by U.S. companies, account for 70% of usable undersea cables in 2026. The U.S. government retains the authority to restrict access to these networks, as demonstrated in 2020 when regulators blocked a segment of the Pacific Light Cable Network connecting the U.S. to Hong Kong due to espionage concerns. As a result, thousands of kilometers of cable were abandoned on the ocean floor.

Even Chinese AI labs, which train models on domestic infrastructure, remain dependent on American-controlled undersea cables for global internet access, including sourcing training data and serving international users. U.S. laws like the CLOUD Act further extend this control by requiring American cloud providers to hand over data stored abroad if requested by authorities. This allows U.S. agencies to access detailed records of AI activity worldwide, including user prompts, model responses, and behavioral patterns.

While the U.S. dominates the AI infrastructure landscape, its control is not absolute. Critical components like semiconductors and AI servers are manufactured in Asia, with companies like TSMC, SK Hynix, and Samsung playing key roles in Nvidia’s supply chain. Recent investments by Nvidia include billions of dollars in contracts with Taiwanese and South Korean firms for advanced chips, memory, and hardware. However, these countries are unlikely to challenge U.S. dominance, as they benefit economically from their role in the American-led tech ecosystem.

The future may see the emergence of competing AI ecosystems, with one led by the U.S. and another by China, each with distinct standards, infrastructure, and governance. Until such an alternative fully develops, the United States will continue to hold a firm grip on the global AI landscape.

fortune.com, (deepseek), (nvidia), (anthropic), (openai), (meta), (alphabet), (xai), (pentagon), (tsmc), (samsung), (china), (japan), (taiwan), (iran)

This article does not provide real, usable help to a normal person. Here is the evaluation point by point.

The article offers no actionable information. It describes the U.S. dominance in AI infrastructure, the role of companies like Nvidia and Meta, and the geopolitical dynamics between the U.S. and China. However, it does not give readers any clear steps, choices, or tools they can use in their daily lives. There are no instructions on how to navigate AI services, assess risks in using AI tools, or make informed decisions about technology adoption. A reader concerned about AI governance, data privacy, or infrastructure dependence has no practical path to follow. The article is purely descriptive, leaving the audience with no way to act on the information.

The educational depth is uneven. While the article explains the components of AI infrastructure—such as data centers, cloud networks, and undersea cables—it does so at a high level without delving into how these systems work or why they matter to an ordinary person. It mentions Nvidia’s market share and the Pentagon’s budget but does not clarify how these factors translate into real-world consequences for users, businesses, or governments outside the U.S. The discussion of export controls and data laws is presented as a fait accompli, with no explanation of how these policies are enforced or what alternatives might exist. The reader is left with a broad understanding of U.S. dominance but no deeper insight into the mechanics or implications of that dominance.

Personal relevance is limited for most readers. The article focuses on large-scale geopolitical and corporate dynamics, which may feel distant to an average person. It does not connect the information to everyday concerns like data privacy, the cost of AI services, or the reliability of cloud-based tools. For example, it mentions the CLOUD Act but does not explain how it might affect someone using a U.S.-based cloud service for personal or business needs. It does not discuss how individuals or small businesses can mitigate risks, such as data exposure or dependence on U.S.-controlled infrastructure. The relevance is tied to abstract power structures rather than tangible impacts on daily life.

The public service function is weak. The article recounts the U.S. government’s control over global AI infrastructure and its collaboration with tech companies but does not provide any warnings, safety guidance, or emergency information. It does not explain how individuals or organizations can protect their data, avoid surveillance risks, or navigate legal restrictions. There is no discussion of how to assess the trustworthiness of AI services, how to diversify technology dependencies, or how to advocate for more transparent or equitable AI governance. The focus is on describing the status quo, not helping the public understand or respond to it.

Practical advice is nonexistent. While the article highlights the risks of U.S. dominance, it does not offer any steps for readers to take action. There is no guidance on how to evaluate AI services, how to choose between cloud providers, or how to secure data against potential access by U.S. authorities. The advice is limited to passive awareness, leaving the reader with nothing to apply beyond understanding that the U.S. holds significant control over global AI infrastructure.

Long-term impact is minimal. The article focuses on the current state of AI infrastructure and geopolitical competition but does not help readers plan for the future. It does not discuss how to build resilience against infrastructure dependencies, how to prepare for potential disruptions in AI services, or how to adapt to evolving regulations. There is no discussion of how to stay informed about changes in AI governance or how to make long-term decisions about technology adoption. The information is tied to a snapshot in time and provides no tools for forward-thinking or risk management.

Emotional impact is neutral but unconstructive. The tone is factual and detached, neither alarming nor reassuring. While the article does not create fear or helplessness, it also does not offer clarity or constructive thinking. It describes a complex and potentially concerning situation without helping readers process the information in a way that empowers them. There is no guidance on how to cope with the implications of U.S. dominance, how to advocate for change, or how to make informed choices in a world where AI infrastructure is heavily controlled by a single country.

The article does not use clickbait or ad-driven language. It avoids exaggerated claims and presents the information straightforwardly. However, this does not compensate for the lack of substance or practical value. The focus is on describing the power dynamics of AI infrastructure rather than helping the reader navigate or respond to them.

Missed chances to teach or guide are significant. The article presents a critical issue—the concentration of AI infrastructure control in the hands of the U.S. and its corporate allies—but fails to provide tools for the reader. It could have explained how individuals and businesses can assess their exposure to U.S.-controlled infrastructure, such as identifying which cloud services they rely on and what data might be vulnerable. It could have offered simple tips for diversifying technology dependencies, such as using non-U.S. cloud providers or open-source alternatives where possible. It could have discussed how to evaluate the transparency and accountability of AI services, including understanding terms of service and data handling policies. Instead, it leaves the reader with no way to apply the information beyond passive awareness of the issue.

To add real value that the article failed to provide, here is concrete guidance on how individuals and small organizations can assess their exposure to AI infrastructure risks and make more informed technology choices.

Start by identifying your current dependencies. Take inventory of the AI tools, cloud services, and digital platforms you use regularly. Note which ones are provided by U.S.-based companies, as these are subject to U.S. laws like the CLOUD Act. For example, if you use Google Cloud, Amazon Web Services, or Microsoft Azure, your data may be accessible to U.S authorities under certain conditions. Similarly, if you rely on AI models from companies like OpenAI, Anthropic, or Meta, understand that these models are trained and deployed on U.S.-controlled infrastructure. Recognizing these dependencies is the first step toward managing them.

Evaluate the sensitivity of your data. Not all data is equally valuable or risky to expose. Ask yourself what information you store or process using AI and cloud services. For example, personal emails, financial records, or business strategies may be more sensitive than public-facing content. If you handle sensitive data, consider whether U.S.-based services are the best choice or if alternatives with stronger privacy protections might be more appropriate. This evaluation will help you prioritize which dependencies to address first.

Explore alternatives to U.S.-controlled services. While the U.S. dominates the AI and cloud infrastructure landscape, alternatives do exist. For example, some European cloud providers offer services that comply with stricter privacy laws, such as the General Data Protection Regulation. Open-source AI models and tools can also provide more control over data and infrastructure. Research non-U.S. providers and compare their offerings to your current services. Keep in mind that switching providers may require adjustments to workflows or compatibility, so weigh the benefits against the effort involved.

Understand the legal and regulatory landscape. Familiarize yourself with the laws that govern the services you use. For example, if you use a U.S.-based cloud provider, learn about the CLOUD Act and how it might affect your data. Similarly, if you operate in a region with strict data sovereignty laws, such as the European Union, ensure that your technology choices comply with local regulations. Understanding these legal frameworks will help you make informed decisions and avoid potential pitfalls.

Create a contingency plan for disruptions. Even if you cannot fully eliminate dependencies on U.S-controlled infrastructure, you can prepare for potential disruptions. For example, back up critical data to multiple locations, including offline storage or non-U.S. servers. Develop a plan for how to continue operations if a key service becomes unavailable or compromised. This might include identifying alternative providers, training staff on backup procedures, or maintaining manual processes as a fallback. A well-prepared contingency plan can minimize the impact of unexpected disruptions.

Advocate for transparency from service providers. Demand clarity from the companies you rely on regarding their data handling practices. Ask questions about where your data is stored, who has access to it, and under what conditions it might be shared with governments or third parties. Look for providers that publish transparency reports or offer detailed explanations of their compliance with data protection laws. By holding companies accountable, you can encourage better practices and make more informed choices about which services to trust.

Stay informed about changes in the AI landscape. The field of AI is evolving rapidly, and new developments can have significant implications for infrastructure and governance. Follow reputable sources of information, such as technology news outlets, academic research, and policy analyses, to stay up to date on trends and shifts in the industry. Join communities or forums where these topics are discussed, and engage with others who share your concerns. Staying informed will help you adapt to changes and make proactive decisions about your technology use.

By following these steps—identifying dependencies, evaluating data sensitivity, exploring alternatives, and creating contingency plans—you can reduce your exposure to risks associated with U.S.-controlled AI infrastructure. While no approach can eliminate all risks, being proactive, informed, and prepared will help you navigate this complex landscape more effectively. Stay aware of your choices, ask questions, and prioritize transparency and control over your data. A thoughtful approach will help you make better decisions and build resilience against potential disruptions.

The text says "the true foundation of AI dominance lies not in the models themselves but in the underlying infrastructure." This picks one part of AI power and hides others. It makes readers think only infrastructure matters. It helps the U.S. by ignoring how good Chinese models might be. The words make the U.S. look stronger than it might really be.

The text says "American export controls, extraterritorial data laws, and deep ties between the tech industry and the military create a system that reinforces U.S. control." This order puts the laws first and the military last. It makes control sound like a natural system, not a choice. It hides that the U.S. government and companies work together on purpose. The words help the idea that U.S. power is fair and strong.

The text says "Nvidia holds roughly 85% of the global market for these chips." The word "roughly" is soft. It makes 85% sound like a guess, not a real number. It hides how big Nvidia's control really is. The word helps Nvidia look less like a monopoly.

The text says "the Pentagon’s 2027 budget allocates over 54 billion dollars (approximately 49 billion euros) for autonomous warfare and drone systems." The word "approximately" is soft. It makes the number sound less exact. It hides how much money the U.S. spends on military AI. The word helps the spending look smaller or less important.

The text says "reports indicate the company has embedded engineers within the National Security Agency to adapt its systems for offensive cyber operations." The word "reports indicate" hides who said it. It makes the claim sound like a guess, not a fact. It helps Anthropic avoid blame if the claim is wrong. The words hide who really knows the truth.

The text says "the U.S. government retains the authority to restrict access to these networks, as demonstrated in 2020 when regulators blocked a segment of the Pacific Light Cable Network connecting the U.S. to Hong Kong due to espionage concerns." The words "due to espionage concerns" hide who said it was espionage. It makes the block sound like a fact, not an accusation. It helps the U.S. look right without proof. The words hide if the block was fair or not.

The text says "U.S. laws like the CLOUD Act further extend this control by requiring American cloud providers to hand over data stored abroad if requested by authorities." The word "further" makes the control sound normal. It hides that the law might be unfair or too strong. It helps the U.S.'s power look like a natural part of the world. The words hide how the law affects other countries.

The text says "these countries are unlikely to challenge U.S. dominance, as they benefit economically from their role in the American-led tech ecosystem." The word "unlikely" is soft. It makes resistance sound like a guess, not a choice. It hides that other countries might want to challenge the U.S. but can't. The words help the U.S.'s power look safe and fair.

The text says "the future may see the emergence of competing AI ecosystems, with one led by the U.S. and another by China." This order puts the U.S first and China second. It makes China sound like a follower, not an equal. It hides that China might already be strong. The words help the idea that the U.S. will stay on top.

The text says "Until such an alternative fully develops, the United States will continue to hold a firm grip on the global AI landscape." The word "firm" is strong. It makes U.S. control sound safe and good. It hides that the grip might be unfair or too tight. The words help the idea that U.S. power is natural and right.

The text conveys a range of emotions, some overt and others subtly embedded in its language, all serving to shape the reader’s perception of U.S. dominance in AI infrastructure. One of the most prominent emotions is **pride**, which appears in descriptions of American technological and corporate strength. Phrases like "commanding lead," "most valuable company," and "deep ties between the tech industry and the military" evoke a sense of superiority and confidence. This pride is not just celebratory but strategic, reinforcing the idea that U.S. control is natural, inevitable, and even beneficial. The emotion is strong, as it permeates the text’s framing of American achievements, from Nvidia’s market dominance to the Pentagon’s budget allocations. The purpose here is to build trust in U.S. capabilities while subtly dismissing concerns about fairness or competition, making the reader more likely to accept this dominance as a given rather than a point of debate.

Another key emotion is **concern**, which surfaces in discussions of U.S. control over global infrastructure. Words like "restrict access," "blocked," and "abandoned" create unease, particularly when describing the Pacific Light Cable Network incident, where thousands of kilometers of cable were left unused. The text also invokes **fear** through phrases like "offensive cyber operations" and "mass surveillance," which suggest potential misuse of power. These emotions are not accidental; they serve to highlight the risks of unchecked control, even as the text ultimately downplays them. The concern is moderate in intensity, as it is often followed by reassurances that U.S. dominance remains unchallenged or that other countries benefit from the system. The effect is to create a tension in the reader’s mind—acknowledging the dangers of U.S. power while simultaneously reinforcing its inevitability.

The text also employs **urgency**, particularly when discussing the future of AI ecosystems. Phrases like "the future may see" and "until such an alternative fully develops" suggest that time is running out for competitors to catch up. This urgency is paired with a subtle **dismissiveness** toward Chinese advancements, framing them as secondary or dependent on U.S. infrastructure. For example, the mention of Chinese AI labs relying on American undersea cables undermines their independence, while the focus on U.S. companies like Meta and Google building new networks reinforces their leadership. The purpose of this urgency is to steer the reader toward accepting U.S. dominance as a long-term reality, discouraging hope for alternative systems.

The writer uses several rhetorical tools to amplify these emotions. **Repetition** is one such tool, with phrases like "U.S. control" appearing frequently to reinforce the idea that this dominance is pervasive and unshakable. **Contrast** is another, as the text pits U.S. strength against the implied weakness or dependency of other nations, particularly China. For instance, the description of Nvidia’s 85% market share is followed by the observation that even Chinese AI labs depend on U.S. infrastructure—a comparison that makes U.S. power seem even more absolute. **Hyperbole** also plays a role, as seen in the exaggerated language around the Pentagon’s budget ("surged by 90.7 billion dollars") or Meta’s cable project ("40,000-kilometer subsea cable"). These choices make the stakes feel higher, ensuring the reader perceives U.S. dominance as both monumental and unassailable.

The emotional tone of the text ultimately serves a persuasive purpose: to frame U.S. control over AI infrastructure as both natural and necessary. Pride in American achievements builds confidence in the system, while concern and fear about its risks are tempered by reassurances that no viable alternative exists. Urgency and dismissiveness work together to discourage skepticism, making the reader more likely to accept this dominance as an unchangeable fact. The text does not invite debate but instead guides the reader toward a specific conclusion—that the U.S. will maintain its grip on global AI, and resistance is futile. By blending these emotions, the writer shapes the reader’s reaction to align with this perspective, whether through admiration, resignation, or reluctant acceptance.