Cruz says US leads AI race but faces cost and trust hurdles

Sen. Ted Cruz states that the United States is winning the artificial intelligence race against China, citing superior chip technology, models, and economic systems. However, he argues that winning requires more than technical superiority; it depends on whether American technology becomes the global standard. Cruz describes the competition as a triathlon with three simultaneous legs: innovation, cost, and trust. He asserts that the outcome will be determined by which AI stacks are adopted worldwide for productivity and decision-making.

In the innovation leg, Cruz claims the US holds a significant advantage. Experts estimate a two-year lead in chip technology, driven by lithography advances and contributions from Nvidia and its partner TSMC. On AI models, he notes that American frontier labs lead by two to three generations, a gap he characterizes as eight to 12 months. While acknowledging that this margin is narrowing, Cruz maintains it remains substantial in the context of frontier AI development. He points to cybersecurity as an example where US models from Anthropic and OpenAI outscore competitors, noting these companies are working with the US government on responsible releases.

The cost leg presents a different dynamic. Cruz reports that top American models cost five to ten times more per token than their Chinese counterparts. Although Chinese models lack the full capabilities of US frontier labs, they are often sufficient for many tasks. This price difference drives adoption among cost-sensitive users, including large global corporations, startups, and developing-country governments. According to data from the model marketplace OpenRouter, approximately 50% of tokens used in the last year were consumed on Chinese models. Cruz warns that many US startups use these Chinese models as code assistants or application substrates, sometimes without disclosing their provenance. He cites Booz Allen research indicating that Chinese models introduce more vulnerabilities when coding for American applications compared to Chinese ones, posing a potential risk to the US software supply chain.

The third leg involves trust. Cruz argues that users are less likely to adopt technology they cannot control or that they believe operates against their interests. He contends that America has a structural advantage in this area due to its free-market system and values, comparing the global adoption of the internet to the lack of interest in China’s state-controlled internet behind the Great Firewall. However, he notes the trust race is closer than it should be, with both nations eroding confidence. China trains its models to refuse questions contrary to Communist Party dogma, while in the US, polls suggest citizens are turning negative on issues like data center construction and the pace of AI advances due to disinformation and a lack of clear frameworks.

Cruz urges simultaneous action on all three fronts to secure national and economic security. He references the President’s America’s AI Action Plan as a roadmap and suggests strengthening it through critical infrastructure designation and communication mechanisms by the end of 2026. This timeline is driven by the observation that AI model capabilities double every four months. The goal is to establish these frameworks before models reach recursive self-improvement or before Chinese ecosystems achieve widespread global adoption.