The US AI leadership is driven more by infrastructure control than by model innovation, according to expert Alex Capri.
The United States continues to lead the artificial intelligence sector, not merely through advanced AI models but by dominating essential infrastructure such as chips, cloud services, and data centers. Alex Capri, author of "Techno-Nationalism," emphasized this in a recent Fortune column.
Capri critiques the common focus on comparing AI models from companies such as OpenAI and Anthropic, asserting that the success of AI systems hinges on a more complex and costly infrastructure. This includes graphics processors, data centers, and extensive data transmission networks.
Nvidia is highlighted as a critical player in this landscape, controlling around 85% of the global graphics processor market essential for training AI models. Capri points out that Nvidia's competitive edge comes from both its chips and its CUDA software ecosystem, which has become the go-to platform for developers, creating significant barriers for switching to other options.
He envisions an "American AI stack" that integrates hardware manufacturers like Broadcom and cloud giants such as Amazon Web Services, Google Cloud, and Microsoft Azure, alongside model developers including OpenAI and Meta.
Collaborative Defense Initiatives
Capri connects the US's advantages in AI to the collaboration between tech companies and defense entities like the Pentagon. Major AI infrastructure providers are not only serving commercial needs but also fulfilling defense requirements. For instance, the US Department of Defense has earmarked $54.6 billion for its Defense Autonomous Warfare Group (DAWG), which focuses on autonomous military technologies.
Federal contracts in AI are also on the rise, with projections indicating a potential value of $90.7 billion by 2026, and current allocations amounting to $7.2 billion, according to Brookings.
Strategic Data Transmission
Capri also discusses the significance of submarine fiber optic cables, noting that American tech firms are constructing their own networks to gain control over data transmission routes. He cites the Pacific Light Cable Network, which received government backing for its US-Taiwan segment, highlighting geopolitical sensitivities regarding connections to Hong Kong.
This control over data routes is crucial for both the United States and China, as even AI models trained in China rely on global internet access for data and APIs. The CLOUD Act, enacted in 2018, further allows US authorities to access data from American cloud providers, regardless of where that data is stored.
US AI Landscape Challenges
Despite these strengths, Capri notes that US dominance is not without its limitations. Key components of the AI infrastructure are still reliant on Asian manufacturers like TSMC and Samsung. However, he believes that countries like Taiwan and South Korea are unlikely to exploit this dependence against the US, given their integration into a supportive tech supply chain.
Looking ahead, Capri foresees the emergence of two contrasting AI ecosystems—one American and one Chinese—each with distinct standards and governance frameworks. In a related note, Nvidia, Meta, Microsoft, and 22 other firms have urged US regulators to keep the AI ecosystem open, arguing this is essential for maintaining national leadership. Additionally, Anthropic has announced plans for a $50 billion investment in US AI infrastructure by November 2025.
