Alibaba Chairman Joe Tsai recently shifted the narrative on the global AI race, declaring it a “long marathon” where victory isn’t about building the biggest models, but about achieving the fastest and most widespread adoption. His comments, delivered at the All-In Summit, highlight a stark contrast in strategy between the US, focused on massive spending for large models, and China, which prioritizes lean, open-source, and rapid integration into everyday technology.
In the high-stakes world of artificial intelligence, the prevailing narrative often centers on an intense “race” to build the most powerful, largest-scale models. However, Alibaba Chairman Joe Tsai recently offered a refreshing counter-perspective, asserting that the AI competition between the US and China is not a winner-take-all sprint, but rather a “long marathon.” Tsai’s key insight, shared at the All-In Summit 2025, is that true success in AI will be defined by the speed of adoption and diffusion, not by who develops the technically strongest model.
Rethinking the Definition of ‘Winning’ in AI
Tsai articulated his viewpoint clearly: “When it comes to AI, there’s no such thing as winning the race… it’s a long marathon.” He elaborated that the constantly evolving nature of AI models means that a leading model one week can be surpassed the next. For him, the real measure of success is not raw power but practical implementation. “My definition of winning,” he stated, “is not who comes up with the strongest AI model, but who can adopt it faster.” This perspective challenges the current trend of pouring billions into building ever-larger models, suggesting a more pragmatic path forward.
He urged the US to recalibrate its focus from massive spending on development to prioritizing the broad adoption and integration of AI technologies across various sectors. This emphasis resonates deeply within the tech community, especially among developers and businesses looking for accessible and deployable AI solutions.
Divergent AI Strategies: US vs. China
The strategies currently employed by the US and China illustrate Tsai’s point vividly. US tech giants are investing colossal sums in AI infrastructure and model development. For instance, executives at Meta anticipate spending a staggering $600 billion on AI infrastructure, including massive data centers, through 2028. Similarly, OpenAI and Oracle have ambitious plans for a $500 billion data center project dubbed Stargate, as reported by Business Insider.
In contrast, China’s approach to AI is characterized by its lean, efficient, and rapid deployment model. Chinese companies are embracing open-source and smaller models that are optimized for real-world applications on everyday devices like mobile phones and laptops. This strategy allows for quicker integration into existing technological ecosystems, making AI more accessible to a wider user base at a fraction of the cost.
A prime example is DeepSeek’s R1 model, which gained significant attention for rivaling top competitors while being developed at a considerably lower cost, as highlighted by Business Insider. This focus on cost-effectiveness and rapid integration is propelling AI proliferation across China. Ray Wang, research director at Futurum Group, observed that China prioritizes rolling out AI across everyday tech at “breakneck speed,” a strategy that could be as crucial as model quality for overall AI competitiveness.
The Power of Proliferation and Practical Application
Tsai pointed to China as a compelling example of faster AI adoption, noting that companies are leveraging open-source and smaller models designed for optimal real-world performance. While acknowledging that China may not be “winning in the model war” technologically, he stressed its significant advancements in “actual application and also people benefiting from AI.” This practical, user-centric approach has seen a remarkable surge in AI utilization: a survey indicated that the percentage of Chinese businesses using AI dramatically increased from 8% last year to 50% this year.
The emphasis on “proliferation” means making AI ubiquitous, seamlessly integrated into daily tools and workflows. For the fan community, this translates to more accessible, efficient, and affordable AI experiences. Smaller, open-source models empower developers to innovate more freely, tailor solutions to specific needs, and avoid the prohibitive costs associated with proprietary large models and their immense computational demands.
Long-Term Impact and the Developer’s Perspective
Tsai’s vision underscores the long-term impact of AI on society. Focusing on adoption fosters a more inclusive AI ecosystem where benefits are widely distributed, rather than concentrated among a few tech giants with massive budgets. This approach encourages innovation at the grassroots level, enabling startups and individual developers to contribute to the AI landscape without needing colossal resources.
For those in the developer community, the appeal of China’s strategy is clear: it champions practicality and broad utility. While groundbreaking large models capture headlines, it’s the efficient, deployable models that truly democratize AI and integrate it into countless applications, from smart devices to enterprise solutions. As Tsai concluded, “You want AI to proliferate,” and that proliferation depends on ease of access, cost-effectiveness, and practical utility, not just raw, unattainable power.
