The global competition in artificial intelligence between China and the United States has evolved into a comprehensive contest of education systems, industrial ecosystems, computing infrastructure, and application-driven innovation. A revealing window into China’s strategic thinking emerged from the minutes of a meeting between Ren Zhengfei and the ICPC chairman, coaches, and award-winning contestants, published on the official website of the ICPC Beijing headquarters. This exchange offers a panoramic view of how technological ambition is being translated into national capability through operational discipline rather than abstract theorizing.
In his November 2025 speech, Ren Zhengfei articulated not a futuristic vision but a pragmatic blueprint for China’s AI development. By stressing that “education is education, business is business,” that “AI focuses on application, not invention,” and that “computing power without a network is an information island,” he cut through prevailing global myths that equate AI leadership solely with breakthroughs in algorithms or raw computing power. His remarks underscore a strategic emphasis on systemic coordination, industrial deployment, and the integration of infrastructure as the true foundations of competitive AI advantage.
A Powerful Engine of Innovation, but a Hollower Body
The United States stands as the world’s most powerful engine of artificial intelligence innovation, yet its physical industrial foundation increasingly appears hollowed out. This dual structure—overwhelming strength in theoretical breakthroughs and capital mobilization alongside growing weaknesses in manufacturing, infrastructure, and social cohesion—defines the paradox of American AI leadership today.
At the level of original innovation and fundamental theory, U.S. dominance remains overwhelming. The founders of deep learning—Geoffrey Hinton, Yann LeCun, and Yoshua Bengio—were all trained and based in North America, and core paradigm shifts such as Transformers, diffusion models, and modern reinforcement learning emerged from American and Canadian research institutions. Elite universities like Stanford, MIT, Berkeley, and Harvard form a self-reinforcing talent pipeline, producing the majority of highly cited AI research globally. Ren Zhengfei has repeatedly acknowledged that the West’s advantage lies precisely in this capacity for abstract, long-term scientific imagination—what he calls the “fantasies” of mathematics and physics that later reshape civilization.
This innovation engine is reinforced by unmatched capital mobilization and global talent absorption. In 2024 alone, U.S. AI startups raised over $80 billion—more than half of global funding—with OpenAI reaching a valuation near $40 billion in a single round. A powerful “positive power-law” loop is formed: extreme capital concentrates elite talent, elite talent generates outsized breakthroughs, and those breakthroughs attract even more capital. As Ren Zhengfei has observed, the United States functions as fertile soil for global talent, including many from China, and the technological civilization created there ultimately diffuses benefits worldwide.
The United States also wields enormous standard-setting and ecosystem-defining power. Nvidia’s CUDA has become a near-monopoly in GPU computing; PyTorch and TensorFlow dominate development frameworks; and AWS, Azure, and Google Cloud control roughly 70% of global cloud AI services. More subtly, the U.S. controls global narrative authority through the concept of AGI itself—binding “the future of humanity” to its own technological pathway and forcing others into its agenda-setting framework.
Yet beneath this intellectual and financial supremacy lies a growing structural hollowing-out of physical industry. Manufacturing contributes barely over one-tenth of U.S. GDP, while advanced chip fabrication is entirely dependent on Taiwan’s TSMC. Automation penetration in heavy industry lags far behind China and Germany. This produces a fundamental contradiction: if AI cannot be deeply embedded into steel, mining, logistics, ports, railways, and energy systems, it risks becoming a “cloud illusion”—powerful in computation but detached from material transformation.
This fragility is increasingly exposed through energy and infrastructure bottlenecks. A single large AI computing center can consume electricity equivalent to that used by 150,000 households annually, yet America’s power grid is aging and increasingly unstable. In 2024, major outages in Texas and California were already linked to data center loads. In December 2025, Nvidia CEO Jensen Huang publicly warned on The Joe Rogan Experience that the future bottleneck of AI will not be chips, but electricity, predicting that major technology firms may eventually need to operate their own nuclear reactors to sustain AI growth. Meanwhile, U.S. 5G coverage remains far behind China’s, limiting real-time industrial AI deployment.
Finally, the United States faces a fragile social contract under AI acceleration. Labor unions resist autonomous trucking, Hollywood writers strike against AI-generated content, and there is growing bipartisan support for tighter regulation of tech giants. Without an effective redistribution mechanism, AI primarily amplifies returns to capital rather than to labor, undermining the traditional “American Dream” narrative of broad-based upward mobility. This social tension increasingly constrains the political legitimacy of large-scale AI deployment.
In sum, the United States remains the unrivaled source of foundational AI innovation, capital concentration, talent aggregation, and global standard-setting power. Yet it simultaneously suffers from a thinning industrial base, rising energy constraints, infrastructure decay, and political resistance. This is the central contradiction of American AI: a towering cognitive superstructure resting on an increasingly hollow physical and social body.
China’s AI: Unmatched Application Scale vs. Foundational R&D and Infrastructure Gaps
China’s artificial intelligence trajectory is distinguished by exceptionally rapid application deployment, enabled by unparalleled industrial depth and scenario density. As the world’s largest manufacturing base across steel, coal, cement, shipbuilding, textiles, and electronics, China provides an unmatched physical foundation for large-scale AI integration. By 2025, roughly half of the world’s industrial robots are projected to operate within China. On the digital side, massive real-world platforms—such as 12306, the world’s largest real-time ticketing system, nationwide mobile payments with over one billion users, and more than 500 smart city pilots—supply continuous data streams for training, validation, and iterative optimization. This combination allows AI to be rapidly transformed into measurable productivity gains rather than remaining confined to experimental settings.
A defining characteristic of China’s model is the translation of AI into incremental yet massive efficiency improvements. Ren Zhengfei’s examples are emblematic: a 1% reduction in energy consumption in blast-furnace ironmaking, unmanned coal mining for gas-explosion prediction, and even a 0.1% improvement in coal-washing accuracy can each yield national-scale resource savings measured in tens of millions of tons. Such cases illustrate how China’s advantage lies not primarily in symbolic technological leadership, but in the systematic conversion of algorithms into concrete industrial output.
This application strength is further reinforced by a distinctive state–market collaboration framework. Artificial intelligence has been elevated into national strategic planning through the “Artificial Intelligence+” initiative, tightly linked to industrial upgrading and the cultivation of new productive forces. China has prioritized infrastructure first, building the world’s largest 5G and fiber-optic network, advancing the East–West Data Computing Project, and deploying BeiDou’s centimeter-level positioning. State-owned enterprises anchor large-scale implementation, private firms such as Huawei and XPeng lead in technological breakthroughs, and top universities provide sustained research support—forming an unusually tight research–engineering–manufacturing loop. Huawei’s Lianqiu Lake R&D Center exemplifies this integrated approach, along with its efforts to reallocate human capital by training rural youth into highly skilled technical workers.
China also possesses a degree of institutional resilience that serves as a social buffer against the disruptive effects of automation. As AI accelerates labor displacement, China’s collective rural land-ownership system provides a fallback safety net for migrant workers in a way that many highly financialized economies cannot replicate. Policy discourse increasingly emphasizes vocational retraining, education vouchers, and “common prosperity” to prevent the extreme polarization often associated with automation shocks. In this sense, AI in China is still largely treated as a tool of national development, rather than as a new driver of rigid class stratification.
Yet beneath these strengths lies a persistent structural vulnerability: dependence on the foundational computing layer. Under restrictions on advanced GPUs such as the H100 and B100, domestic alternatives like Ascend and Cambricon continue to lag in software ecosystems, cluster interconnection, and advanced numerical formats such as FP8. Critical links in the semiconductor value chain—including EDA tools, advanced packaging technologies, and high-end photoresists—remain reliant on imports. Even industry leaders acknowledge that full technological self-reliance is a long and uncertain process, making the infrastructure layer the principal bottleneck in China’s AI ascent.
A second constraint is the relative weakness of original, theory-driven innovation. Compared with the United States, China produces fewer breakthroughs at the frontier of basic AI research. The academic system remains partially misaligned with industrial needs, while venture capital favors fast-return application layers—such as live streaming, recommendation engines, and e-commerce AI—over long-cycle fundamental research. There is growing awareness that over-commercialization of education risks hollowing out the scientific foundations needed for long-term leadership.
China also faces a global trust deficit that complicates its efforts to export AI ecosystems. Western narratives framing Chinese digital systems as instruments of “digital authoritarianism” restrict the overseas adoption of platforms such as smart cities and health-code infrastructures. Although Chinese open-source models have expanded rapidly and achieved enormous domestic uptake, their international influence still trails leading Western counterparts, underscoring the geopolitical and normative barriers confronting China’s outward AI expansion.
Nevertheless, China’s practical strength in applied AI is vividly illustrated by telemedicine and remote diagnostics. Portable ultrasound probes in remote Tibetan pastoral regions can transmit real-time medical data over thousands of kilometers to coastal diagnostic centers, enabling early detection of serious diseases through 5G and fiber-optic networks. When combined with AI-assisted diagnosis, such systems dramatically compress medical inequality between urban cores and remote peripheries. Similar application-driven models are now being extended to ports in Southeast Asia, photovoltaic infrastructure in the Middle East, and healthcare across Africa.
In sum, China’s AI development is defined by a powerful contradiction: extraordinary speed and scale in application, anchored in industrial depth, infrastructure, and coordinated mobilization, alongside persistent fragility in foundational computing, original research, and global trust. Its comparative advantage lies in turning algorithms into immediate productivity across vast material systems; its core disadvantage lies in the upstream technologies and institutional conditions that ultimately determine long-term autonomy. The future of China’s AI power will be decided not by how fast it deploys models, but by whether it can overcome these foundational bottlenecks.
Global Competition and Cooperation in Third-Party Markets: How Development Models Shape the Outcome of the U.S.–China AI Rivalry
The strategic center of gravity in the U.S.–China rivalry in artificial intelligence is no longer confined to bilateral competition or frontier-model capability alone. Instead, it is increasingly determined in third-party markets—especially across the Global South—where the contest is not merely about who owns the most advanced algorithms, but about which national model of technological diffusion, industrial integration, and development partnership proves more durable and attractive. In these regions, AI is not an abstract scientific frontier but a concrete instrument for infrastructure upgrading, productivity growth, and state capacity building.
The United States has largely pursued a model based on point-to-point technology export embedded in a geopolitically entangled ecosystem. Its typical offerings are high value-added yet structurally isolated products: advanced GPU systems, closed-source large-model APIs, and cloud-based AI services delivered through hyperscalers. This approach prioritizes short-term efficiency and rapid deployment, as seen in hardware and cloud service exports to partners such as India. Yet beneath this commercial logic lies a deeper structural objective: to lock recipient countries into a U.S.-centered technological hierarchy through standards, software toolchains, and data interfaces, thereby consolidating long-term ecosystem dependence.
China, by contrast, has advanced a fundamentally different path centered on end-to-end ecosystem co-construction and development empowerment. Rather than exporting discrete technological components, it emphasizes integrated delivery: open-source large models, localized hardware–software stacks, communications and power infrastructure, vocational training systems, and long-term financing mechanisms. The objective is not merely market access but the cultivation of indigenous digital production capacity and sovereign adaptability, enabling partner countries to internalize, reproduce, and evolve the systems they adopt.
The intelligent transformation of Tianjin Port illustrates this systemic approach with exceptional clarity. Huawei enabled the construction of a fully integrated 5G private network and F5G all-optical communications backbone, achieving millisecond-level latency and carrier-grade reliability across the entire port. On this foundation, AI-driven scheduling systems dynamically optimize vessel berthing, crane operations, and truck routing via real-time digital twins, while unmanned cranes and autonomous electric container trucks operate through centimeter-level positioning and collaborative obstacle avoidance. Predictive maintenance based on AI and IoT further enhances safety and resilience, pushing the system toward self-optimizing operation.
The strategic significance of Tianjin Port lies not merely in automation itself, but in the transformation of industrial organization and labor structure. Functions that once required hundreds of on-site workers have been reorganized into small teams of remote operators and system engineers. More importantly, the technological architecture is modular and replicable. Under the Port IntelliVerse framework, the same 5G+AI model has been transferred to major Chinese ports and extended to overseas projects such as Chancay Port in Peru—where unmanned loading, intelligent stacking, automated customs clearance, energy optimization, and local workforce training are delivered as a unified system rather than as fragmented products.
This pattern reveals the deeper geopolitical contrast between the two models. The U.S. system delivers “precision instruments”—powerful but dependent tools embedded in a tightly controlled ecosystem. China, in contrast, offers “the entire factory”—a co-built industrial platform that integrates infrastructure, software, hardware, human capital, and institutional learning. One emphasizes transactional access; the other emphasizes long-term co-production. One consolidates structural dependence; the other seeks to build systemic autonomy within partner states.
In Global South markets marked by weak industrial foundations and urgent demands for large-scale productivity upgrades, this distinction is decisive. When a country is offered not only GPUs or cloud APIs, but a complete solution combining satellite positioning, AI-enabled port logistics, smart energy management, and localized engineer training, the choice becomes a matter of national development trajectory rather than technical preference. It is in these third-party arenas—ports, power grids, transport corridors, industrial parks—that the ultimate balance of technological influence between China and the United States is being shaped.
The outcome of the AI rivalry, therefore, will not be determined solely by which country leads in frontier model parameters or compute scale. It will be decided by which model proves capable of embedding artificial intelligence into the real economy of the developing world in a way that is sustainable, sovereign, and scalable. In this sense, the competition is no longer just about intelligence itself—it is about whose vision of technological modernity becomes structurally dominant across the next generation of industrializing economies.
Final Thoughts
The US–China competition in artificial intelligence is no longer a narrow race over computational speed or technical breakthroughs, but a deeper contest over who defines the meaning of “winning” in the transformation of civilization itself. The United States, driven by free markets, limited government, and exceptional individuals, excels at igniting innovation but often struggles to sustain it at scale. China, by contrast, integrates an effective market with a proactive government and collective resilience, combining the invisible and visible hands to convert technological potential into enduring social and institutional capacity—not to be the first to touch the moon, but to ensure that its light reaches every home.
As Ren Zhengfei urged young scientists to “dare to move forward amidst doubt,” true leadership in this era lies not in solitary brilliance, but in turning uncertainty into direction and progress into shared benefit. In the coming decade, while the world debates AGI in theory, China may quietly normalize AI-powered disease screening in remote plateaus and millisecond-level coordination of high-speed heavy-haul rail. These advances may not appear dramatic, but they are real, structural, and transformative—redefining national destiny not through spectacle, but through sustained, inclusive change.
References
- Minutes of the meeting between Ren Zhengfei and the Chairman, coaches and award-winning athletes of the ICPC, November 2025