Alibaba chairman Joe Tsai has challenged conventional thinking about the global AI competition, arguing that the race between the United States and China isn’t about building the most powerful models—it’s about who can deploy and adopt AI technology faster. Speaking at the All-In Summit 2025 in September, Tsai emphasized that “when it comes to AI, there’s no such thing as winning the race” in the traditional sense.
Tsai’s perspective comes as US tech giants pour hundreds of billions of dollars into developing increasingly large AI models and infrastructure. Meta expects to spend $600 billion on AI infrastructure through 2028, while OpenAI and Oracle announced the $500 billion Stargate data center project. However, Tsai questions whether this massive capital expenditure represents the optimal strategy.
“My definition of winning is not who comes up with the strongest AI model, but who can adopt it faster,” Tsai stated, noting that model leadership changes weekly as different systems leapfrog each other in capabilities. Instead, he advocates for the US to focus more on adoption and diffusion of AI technology rather than exclusively pursuing larger, more expensive models.
Tsai highlighted China’s approach as a contrasting strategy that prioritizes practical implementation over raw power. Chinese companies are embracing open-source and smaller models optimized for real-world applications, including mobile devices and laptops. This lean, efficient approach was exemplified by DeepSeek’s R1 model, which disrupted the AI industry and US stock markets earlier this year by rivaling top competitors while reportedly being built at a fraction of the typical cost.
“I’m not saying China technologically is winning in the model war,” Tsai clarified, “but in terms of the actual application and also people benefiting from AI, it has made a lot of development.” He emphasized that widespread AI proliferation should be the ultimate goal.
According to Ray Wang, research director for semiconductors and emerging technology at Futurum Group, China’s strategy focuses on rolling out AI across everyday technology at breakneck speed rather than trying to outbuild leading players like OpenAI. Wang told Business Insider that this rapid integration could prove just as crucial as model quality in determining a country’s overall competitiveness in the AI landscape, suggesting that practical deployment may ultimately matter more than theoretical capabilities in the global AI competition.
Key Quotes
When it comes to AI, there’s no such thing as winning the race. It’s a long marathon.
Alibaba chairman Joe Tsai reframed the AI competition at the All-In Summit 2025, challenging the notion that AI development is a winner-take-all contest and emphasizing the long-term nature of technological advancement.
My definition of winning, you know, is not who comes up with the strongest AI model, but who can adopt it faster.
Tsai articulated his core thesis that practical deployment and adoption speed matter more than raw model capabilities, suggesting a fundamental shift in how AI success should be measured.
I’m not saying China technologically is winning in the model war. But in terms of the actual application and also people benefiting from AI, it has made a lot of development.
Tsai acknowledged China’s technical limitations while highlighting its advantages in practical AI implementation and real-world benefits, drawing a distinction between theoretical capabilities and tangible impact.
Every week there’s a model that’s leading, but then the next week another model overtakes them.
Tsai pointed to the rapid pace of AI model development to argue that pursuing model supremacy is a moving target, suggesting that focusing on deployment infrastructure may be more strategically sound than chasing temporary technical leads.
Our Take
Tsai’s argument exposes a critical tension in AI strategy that deserves serious consideration. While US companies pursue computational supremacy through massive capital expenditure, they may be overlooking the “last mile” problem of AI deployment. China’s approach—exemplified by DeepSeek’s cost-efficient models—suggests that practical engineering and rapid integration could matter more than benchmark performance. This mirrors historical technology competitions where adoption and ecosystem development often trumped raw technical superiority. The VHS versus Betamax battle and Android’s market dominance despite iOS’s technical polish offer cautionary tales. However, Tsai’s perspective may underestimate how frontier model capabilities enable new applications that smaller models cannot replicate. The reality likely requires both approaches: breakthrough research to expand AI’s possibilities and aggressive deployment to capture value. The question isn’t which strategy is correct, but which balance will prove optimal—and that answer may differ across industries and use cases.
Why This Matters
Tsai’s perspective represents a fundamental challenge to the prevailing AI development paradigm in Silicon Valley, where bigger models and massive infrastructure investments have been viewed as the path to dominance. His argument that adoption speed matters more than model power could reshape how companies and governments approach AI strategy and investment.
This debate has significant implications for global competitiveness and resource allocation. If Tsai is correct, the hundreds of billions being spent on ever-larger models may yield diminishing returns compared to investments in deployment infrastructure and integration capabilities. For businesses, this suggests that competitive advantage may come from implementation excellence rather than access to the most powerful models.
The contrast between US and Chinese strategies also highlights different philosophical approaches to technology development—capital-intensive moonshots versus pragmatic, distributed deployment. As AI becomes increasingly central to economic competitiveness, which strategy proves more effective could determine technological leadership for decades to come. For workers and society, faster adoption could mean more immediate AI benefits but also more rapid disruption across industries.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- OpenAI’s $157B Valuation: Can It Win the Brutal AI Race?
- Google’s Gemini: A Potential Game-Changer in the AI Race
- The Artificial Intelligence Race: Rivalry Bathing the World in Data
- The AI Hype Cycle: Reality Check and Future Expectations
- Mistral AI Launches Le Chat Assistant for Consumers and Enterprise