US vs China in AI: How Competing Rules, Chips, and Ecosystems Are Shaping a New Digital Cold War (Part One)
A business in Europe seeking to implement AI at scale faces a challenging strategic decision: to align with either the US or the Chinese ecosystems.
Europe does not possess a complete AI infrastructure—there is no equivalent to Nvidia or TSMC for advanced chips, no hyperscale providers with comparable GPU capabilities, and only a few advanced language models, such as Mistral, DeepL, and Aleph Alpha, that operate at a significantly smaller scale than their US and Chinese counterparts. As a result, many European AI initiatives, particularly in the enterprise sector, tend to depend on US-based clouds and models (such as OpenAI, Anthropic, Google, Meta, and Microsoft) or on Chinese platforms, where concerns over geopolitics and data sovereignty are even more pronounced.
For a German or French company, this issue transcends mere procurement; it has become a strategic choice regarding which standards, export-control frameworks, and regulatory environments to adopt. The ecosystem selected will determine technical dependencies, compliance with the EU AI Act, and the flexibility available in an increasingly fragmented world of competing “sovereign AI” blocs. When I teach AI literacy to MBA students, I talk about how Obama, in the final year of his presidency, “helped” shape China’s national AI strategy (without intending it), and how a single comment from Merkel encapsulated the future competitiveness of the German economy for the decade following her departure from government.
In 2016, the Obama administration published three significant documents: Preparing for the Future of Artificial Intelligence, the National Artificial Intelligence Research and Development Strategic Plan, and a subsequent report addressing AI, automation, and the economy. These reports positioned AI as a vital technology, advocated for sustained federal research and development funding, encouraged public-private collaborations, emphasised talent cultivation, promoted the use of open datasets and benchmarks, highlighted the importance of ethics and safety, and called for the incorporation of AI into government operations.
In July 2017—less than a year later—China’s State Council released the New Generation Artificial Intelligence Development Plan, marking its first comprehensive national strategy for AI. The goal was to catch up with leading nations by 2020 and to become the global leader in AI by 2030.
Analyses from US and Chinese policy experts indicate that this plan mirrors the structure and priorities of the Obama-era AI strategy, focusing on long-term investments in AI research and development, fostering collaboration among government, industry, and academia, creating talent pipelines, establishing open platforms and datasets, considering societal impacts, and leveraging AI for both economic growth and national security.
The timing, structural resemblances, and various expert analyses indicate that Chinese strategists carefully examined and modified the US framework while creating the 2017 New Generation AI Development Plan.
Since 2017, China has emerged as a serious leader in the AI landscape, establishing itself as the second, if not equal, AI ecosystem and economy. Meanwhile, Europe began to address emerging AI and robotics technologies while focusing on risks, despite a lack of comprehensive understanding. The regulation of AI in Europe has been crafted by bureaucrats, lawyers, and consultants. Let us examine recent developments that influence how AI research and development, along with application development, are shaping two distinct AI ecosystems in the US and China.
The White House published its AI Action Plan, which lays out a whole-of-government strategy to reindustrialise America around sovereign compute.
The initiative seeks to create a National AI Research Resource along with new national AI institutes, offering centralised access to computing and data resources for U.S. researchers. It mandates that federal agencies implement the NIST’s AI Risk Management Framework to ensure consistent evaluation and oversight across various departments.
Additionally, the plan underscores the importance of thorough safety testing and transparency requirements for frontier models, aiming for proactive governance of dual-use technologies.
On the hardware front, it relies on the coordination of the CHIPS Act to enhance local advanced semiconductor manufacturing and suggests targeted incentives for energy-efficient data centres. Importantly, it also encompasses immigration reform and federal fellowships aimed at revitalising the U.S. AI talent pool.
Central to this strategy is the goal of re-industrialising the U.S. economy around sovereign computing, both as a defensive measure against adversarial nations and as an offensive strategy to lead in the development of frontier models. The U.S. aspires for global reliance on its standards, even as NVIDIA promotes AI factories as a form of sovereignty, which many world leaders seem to embrace.
China has responded with its own vision. The official policy paper from the Chinese Ministry of Foreign Affairs lays out priorities for international AI governance, including a new multilateral institution to balance US dominance. Beijing presents this as a collaborative, open-source-first strategy, but the true intention is evident: to establish dominance over global AI standards before the West can.
Eric Schmidt, the former CEO of Google and a passionate supporter and investor in AI, believes that China may achieve this goal.
During a 2025 episode of the Moonshots podcast, Schmidt expressed his concern that “most governments might ultimately rely on Chinese AI models simply because they are free,” whereas the leading US models remain closed and costly. He called this a “bizarre outcome”: the largest US models (OpenAI, Anthropic, Google, etc.) are closed-source, whereas many of the largest Chinese models (e.g., DeepSeek, Alibaba’s Qwen) are open-weight or effectively open, making them far easier for budget-constrained governments to adopt.
Schmidt presents this as a geopolitical divide: on one side, there are affordable and open Chinese systems; on the other, costly and closed Western systems, with developing nations being drawn towards the former due to economic reasons rather than ideological ones. Schmidt connects this to the discussion on “sovereign AI,” suggesting that if the US and its allies fail to provide affordable, value-aligned AI solutions (such as certain open-source or subsidised models), many countries’ “sovereign AI” will effectively rely on Chinese technology and governance standards.