US-China AI Competition Divides Along 'Brains' and 'Bodies'; Enterprise AI Agents Reshape Auditing While Challenging Economic Paradigms

The global race in artificial intelligence between the United States and China pits American strengths in AI 'brains'—such as large language models (LLMs) and advanced microchips—against China's advantages in AI 'bodies,' including humanoid robots and manufacturing automation, according to Nick Wright, a cognitive neuroscience researcher at University College London (UCL). This framing, articulated by Wright in a BBC report published approximately 11 hours before April 8, 2026, underscores a competition spanning research labs, universities, startups, corporate boardrooms and government levels, with trillions of dollars at stake. Experts like Bloomberg columnist Parmy Olson, author of 'Supremacy: AI, ChatGPT, and the race that will change the world,' note that the US has maintained an edge in LLMs since OpenAI's launch of ChatGPT on November 30, 2022, which dazzled the tech world and prompted rivals like Anthropic, Google and Perplexity to invest billions.

OpenAI claims over 900 million weekly ChatGPT users as of early 2026, per the BBC citing company statements. A senior US official, speaking anonymously to the BBC, attributed America's lead to dominance in high-end microchips, primarily designed by Nvidia, which reached a $5 trillion valuation in October 2025, according to Stephen Witt, author of 'The Thinking Machine.' US export controls, strengthened by President Joe Biden in 2022 and rooted in 1950s policies against Soviet allies, restrict China's access via the 'foreign direct product rule,' affecting Taiwanese firm TSMC and Dutch ASML's ultraviolet printing machines essential for chip production. These measures have largely succeeded in preserving US advantages in AI brains, the senior official told the BBC.

China countered with DeepSeek, an AI chatbot launched in January 2025 coinciding with Donald Trump's second inauguration, which rivals ChatGPT in capabilities like question-answering and code-writing while being free to use. AI journalist Karen Hao told the BBC that DeepSeek's development, using fewer and older Nvidia chips like A100/H800 models amid export curbs, cost 'a fraction' of US counterparts like ChatGPT and Anthropic's Claude—though exact figures remain unverified in public sources. On January 27, 2025, Nvidia shares dropped $600 billion in market value, the largest single-day US stock loss ever, per BBC reports. Selina Xu, a researcher on China AI policy in Eric Schmidt's office, described DeepSeek as a 'positive catalyst' for China's ecosystem, fueled by an open-source approach where firms share code, contrasting US proprietary models, according to Olson and Xu.

In AI bodies, China leads with approximately two million industrial robots—more than the rest of the world combined—bolstered by government subsidies since the 2010s, Olson told the BBC. Xu highlighted robots' integration in daily life, like drone food deliveries in Shenzhen and Shanghai. A Center for Strategic and International Studies (CSIS) report, cited by BBC, described an unverified 'dark factory' in Chongqing with 2,000 robots and autonomous vehicles producing a car per minute—similar to verified sites like Zeekr's Ningbo plant (800+ robots, 800 EVs/day), but specifics unconfirmed. China accounts for 90% of humanoid robot exports, targeting its aging population projected to exceed the US total by 2035, Xu said.

US firms retain an edge in robot 'brains' via agentic AI for complex tasks, Wright told BBC, noting 80% of a robot's value lies there. Boston Dynamics' Spot robot, using agentic AI for warehouse inspections detecting overheating or leaks and integrating with IFS software, exemplifies this. Military applications include Ukraine's Gogol-M drone mothership, deployed summer 2025, releasing AI-guided attack drones without human control, per Wright. Greg Slabaugh, professor of computer vision and AI at Queen Mary University of London, told BBC that victory depends on sustained economic embedding and standards-setting, not a single breakthrough.

Amid this rivalry, AI is transforming enterprise practices. EY launched a global multi-agent AI framework embedded in its EY Canvas assurance platform on April 7, 2026 (Tuesday), targeting 100% audit activity support by 2028 for its 130,000 auditors, per Marc Jeschonneck, EY's global assurance transformation leader, in a Business Insider interview—though public EY releases confirm Canvas and March 2025 agentic platforms without specifying this integration or timeline. Initial agents include a core assistant, document search/summarization tools and administrative automation, with 20 modular capabilities expanding via data combinations; upcoming ones will review work papers and handle reconciliations, Jeschonneck said.

Jeschonneck described a 'steep learning curve' for junior auditors needing experience to review agent outputs, shifting training from repetitive on-the-job tasks to scenario-based learning with adaptive tools and videos. He maintained this simplifies life long-term, freeing skilled graduates from drudgery, and EY plans stable headcount despite reduced legacy audit needs, citing demand for tech-savvy roles amid complex regulations—contrasting PwC's announced US entry-level hiring cut of one-third by mid-2028 (Business Insider, August 2025), KPMG's ~600 UK role cuts (CityAM, March 2026) and graduate salary freezes since 2022, which Jeschonneck did not address.

AI's disruptions extend to economic theory. In 'The Marginal Revolution: Rise and Decline, and the Pending AI Revolution'—a self-published 'generative book' by economist Tyler Cowen available free on his website since early 2026—Cowen argues marginalism, pioneered by Jevons, Walras and Menger in the 1870s amid calculus advances and economics' professionalization, is declining. Financial examples include the Capital Asset Pricing Model (CAPM) undermined by Fama and French's 1992 paper showing Beta's irrelevance, with quants now dominating via machine learning on raw data, per a purported 2024 Journal of Financial Economics paper (unverified in JFE archives) and MIT/Harvard 'fully automated social science' using LLMs as scientists/subjects (unverified specific project).

Cowen, per reviewer Stephen Pimentel in The Federalist (April 7, 2026), posits AI like LLMs matching human forecasting and generating novel hypotheses reveals marginalism's limits, akin to its 1870s emergence from converging preconditions. Mari Sako of Oxford's Said Business School told BBC the side courting wider adoption may prevail, with US favoring unregulated capitalism and China state oversight.