What China’s “diffusion-forward” strategy reveals about technology, production, and who wins in the AI economy
Listen to this article:
The New York Times recently reported that in Washington, the AI question focuses on whether the government should review powerful models before they are released. But on the other side of the globe, China is asking how quickly AI can be embedded into the machinery of production. This difference marks a consequential strategic divergence, one that may have broad implications for where the global economy is headed. In the new working paper “China’s Diffusion-Forward AI Strategy: Chatbots, Robots, and Political Economic Possibilities,” HBS AI Institute Associate Meg Rithmire and her co-author Hao Chen explain how China is treating AI as an input into production that can reshape factories, robotics, logistics networks, and the entire future economy. The result is a race not only over model capability and power, but national coordination and the speed of practical deployment.
Key Insight: The “Diffusion-Forward” Strategy
“The priority has been to deploy AI in a diverse set of sectors and production stages in the physical economy.” [1]
The authors argue that the US and China have arrived at fundamentally different visions of what AI is for. The American approach, shaped by a service economy, private capital, and a fixation on Artificial General Intelligence (AGI), treats AI primarily as a tool to amplify cognitive work. China’s wager is that AI will matter most when it is absorbed into the systems where the country already has scale. Rather than treating model performance or AGI as the finish line, China is turning AI into the operational baseline. The authors call this a “diffusion-forward” strategy, and argue that it could amplify what China already does well: manufacture, commercialize, and scale physical goods quickly. It doesn’t mean that model advancement is irrelevant, but the larger strategic logic is downstream: cost efficiencies, sector-specific applications, and enterprises that use AI from inception rather than bolting it onto existing workflows.
Key Insight: The Rise of the “Investor-State”
“AI was not imagined as an end in itself.” [2]
China’s approach is enabled by what the authors call an “investor-state” model that doesn’t simply regulate, but invests directly through state-backed capital. The state functions as an active equity investor through over 2,000 government guidance funds, channeling trillions of RMB into strategic sectors like semiconductors and robots. This capital is deployed through “campaign-style” industrial policies, such as the “AI+ Initiative” introduced in 2024, which calls for the deep integration of AI with the real economy. This allows for what the authors call a “cascading” implementation where central targets are rapidly translated into local mandates. For instance, by early 2026, the Cyberspace Administration of China (CAC) had recorded over 1,177 generative AI service registrations, with state-linked entities representing nearly 25% of the total. This state footprint ensures that AI development aligns with industrial upgrading rather than mere consumer entertainment.
Key Insight: AI Walks Onto the Factory Floor
“While Silicon Valley builds chatbots, Shenzhen builds robots.” [3]
The authors’ case study of Shenzhen-based UBTECH Robotics shows the diffusion-forward strategy in action. The Liuzhou municipal government invested 200 million RMB in UBTECH to facilitate its integration into local automotive supply chains. UBTECH’s 4,091 patent filings cover a full range of robotics, including power systems, motion control, and AI perception, and its commitment to innovation is staggering, maintaining an average R&D spend of 62.4% of its total revenue. Combined with land grants, supply-chain proximity, and industrial customers, UBTECH’s Liuzhou plant reportedly scaled from its first Walker S2 robot in June 2025 to its 1,000th unit in December that same year.
Key Insight: The Global Imbalance
“The success of either country’s strategy will also depend on the accompanying economic and political strengths and fragilities and how well the institutional structures adapt.” [4]
The authors are careful to note that China still faces real constraints: AI deployment remains concentrated in Beijing and Shanghai, fiscal strain from industrial overinvestment is real, and youth unemployment may worsen as automation displaces workers. There is also the risk that AI strengthens patterns that have already created trade tensions: enormous productive capacity, weak domestic demand, falling prices, and pressure to export (with products like solar panels, batteries, and EVs). It remains to be seen whether institutional systems can adapt as quickly as the technologies they are trying to harness.
Why This Matters
If China’s diffusion-forward strategy continues, competitors will face AI-enabled firms that move faster, produce cheaper, and integrate software intelligence directly into physical processes. As a business leader, will you become one of those lagging competitors, focusing just on upgrading to the latest AI model, or are you ready and prepared for the challenge of integrating AI into your organization’s workflows, products, supply chains, and service models? The lesson from Beijing is that the most important AI question today is no longer “What can AI say?” but “What can AI do?”
Bonus
If AI systems are being deployed at scale across the physical economy and the information ecosystem, what happens if the wrong actors target public trust and political participation? For a look that extends this conversation from economics to geopolitics and the resilience of the public sphere itself, check out The New Influence War: How AI Could Hack Democracy.
References
[1] Chen, Hao and Meg Rithmire, “China’s Diffusion-Forward AI Strategy: Chatbots, Robots, and Political Economic Possibilities,” Working Paper (April 2026): 2.
[2] Chen and Rithmire, “China’s Diffusion-Forward AI Strategy,” 6.
[3] Chen and Rithmire, “China’s Diffusion-Forward AI Strategy,” 18.
[4] Chen and Rithmire, “China’s Diffusion-Forward AI Strategy,” 30.
Meet the Authors
Hao Chen is a research fellow at Harvard Kennedy School.
Meg Rithmire is the James E. Robison Professor of Business Administration at Harvard Business School and an Associate at the HBS AI Institute.
Watch a video version of the Insight Article here.