Is Alibaba the Most Undervalued Player in AI Right Now?
Alibaba's approach to artificial intelligence has been characterized by a methodical, long-term strategy aimed at building a solid foundation for global leadership. This journey began in earnest in 2017 with the establishment of DAMO Academy, Alibaba's dedicated AI research lab.
The fruits of these early efforts emerged in 2023 with the release of Tongyi Qianwen, Alibaba's first foundation model, and the subsequent Qwen series, marking the company's entry into the upper echelons of AI innovation.
The recent unveiling of Qwen3, the most advanced model in the series, represents a significant milestone in Alibaba's AI journey. More than a technological achievement, Qwen3 is a comprehensive solution designed to integrate AI seamlessly across Alibaba's vast commercial ecosystem.
The Qwen family has already garnered over 40 million open-source downloads, a testament to its growing influence.
Alibaba is doubling down on this momentum with a substantial RMB 380 billion ($53 billion) investment earmarked for cloud and AI over the next three years, eclipsing its total cloud investments from the past decade.
Qwen3: Balancing Scale and Efficiency
Qwen3 is a family of eight models tailored for diverse use cases. This includes six dense models ranging from 0.6 billion to 32 billion parameters and two Mixture-of-Experts (MoE) models.
The flagship model, Qwen3-235B-A22B, employs an efficient MoE architecture that activates only 22 billion parameters per inference, enabling robust performance while optimizing computational resources.
Trained on an extensive dataset spanning 36 trillion tokens across 119 languages, Qwen3 adeptly handles global tasks while navigating local linguistic nuances.
This multilingual proficiency provides Alibaba with a competitive edge in international markets.
Benchmark Performance: Narrowing the Gap
In rigorous benchmarks, Qwen3-235B-A22B demonstrates strong performance, closely trailing industry leaders like Google's Gemini 2.5 Pro in ArenaHard, outperforming OpenAI's o1 in AIME'24, and showcasing robust coding capabilities in LiveCodeBench.
These results underscore Alibaba's rapid advancements in AI, narrowing the gap with leading Western models.
Notably, Qwen3 exhibits hybrid reasoning capability, seamlessly transitioning between quick responses and in-depth analysis, making it well-suited for complex, real-world applications.
From Research to Real-World Impact
Qwen3's impact extends beyond the research lab, with active integration into Alibaba's core businesses, including Alibaba Cloud, Taobao, and Cainiao. This comprehensive integration underscores Qwen3's role in delivering direct business value, distinguishing it from models lacking clear commercial applications.
Alibaba is leveraging Qwen3's open-source adoption through various monetization avenues, such as enterprise API access for high-performance inference, customized fine-tuning for industry-specific adaptations, and optimized cloud instances tailored for Qwen workloads.
These strategies position Alibaba to capture a significant share of enterprise AI spending and reinforce its leadership in China's rapidly evolving cloud market.
The Road Ahead: Scaling and Strategic Partnerships
Looking ahead, Alibaba's substantial investment in Qwen3 ensures ample resources for its continued development and refinement. The recent partnership with Apple to integrate AI features into iPhones in China reflects Qwen3's expanding influence and potential for broad consumer adoption.
As Qwen3 matures, key areas to monitor include its performance in large-scale enterprise deployments and its progression relative to global benchmarks. With China's policy landscape increasingly emphasizing AI as a "new productive force," Qwen3 is poised to benefit from the forthcoming wave of domestic AI investment and adoption.
Conclusion
Qwen3 represents more than just another AI model; it is a foundational element of Alibaba's long-term strategic vision.
Its seamless integration across Alibaba's ecosystem, competitive performance, and growing roster of strategic partnerships underscore its significance in shaping the company's future trajectory. For stakeholders and investors, Qwen3 offers a critical lens through which to assess Alibaba's potential for AI-driven growth and transformation over the coming decade and beyond.
As the AI race continues to unfold on a global scale, Alibaba's deliberate, steady approach, exemplified by Qwen3, may prove to be a winning strategy.​​​​​​​​​​​​​​​​
Thanks for the informative article. However, I'm confused by the title. How are they winning the AI race if you freely admit they're trailing multiple western models?