Alibaba Group delivered its most comprehensive defense yet of an aggressive, diversified artificial intelligence investment strategy on Thursday, with chairman Joe Tsai arguing that the company’s “all in” approach reflects a fundamental belief that AI’s total addressable market will eventually dwarf current valuations of frontier model companies.
Speaking at the VivaTech technology conference in Paris on June 18, Tsai articulated a vision of AI as a transformative force reshaping human productivity itself. His framing positioned Alibaba’s sprawling investments, from semiconductor design to cloud infrastructure to foundation models to consumer applications, not as hedging bets, but as a rational response to the scale and uncertainty of AI’s economic impact.
The $50 Trillion Productivity Thesis
Tsai’s core argument rested on a deceptively simple calculation. “If you look at global GDP, over US$100 trillion of GDP, at least half of that, US$50 trillion, is about human productivity and human intelligence,” he stated. “And that is the TAM (total addressable market) of AI. And that’s why we’re all in on AI.”
This framing redefines how investors should think about AI’s economic opportunity. Rather than focusing on the current market capitalizations of leading AI model companies—which have attracted billions in venture funding and corporate investment—Tsai argued that the real prize lies in capturing value across the entire AI value chain. If AI genuinely augments or replaces human cognitive work across $50 trillion in economic activity, then companies positioned at multiple points in that chain could capture disproportionate value.
Full-Stack Strategy Across Chips, Infrastructure, Models, and Applications
South China Morning Post reported that Alibaba’s strategy spans four critical layers of the AI stack. At the foundation, the company is developing custom semiconductors optimized for AI workloads. Above that, Alibaba Cloud provides the infrastructure—data centers, networking, and compute resources, that power large-scale AI training and inference. The company also develops foundation models, competing directly with OpenAI, Anthropic, and Chinese rivals like DeepSeek and Zhipu AI. Finally, Alibaba is building consumer and enterprise applications that embed these models into products users and businesses actually interact with.
This full-stack approach contrasts sharply with the strategies of pure-play AI companies that focus narrowly on model development or infrastructure. Alibaba’s willingness to invest across all four layers reflects both its financial resources and its conviction that no single layer will prove to be the primary source of AI-driven value creation.
Challenging the “Model Company” Narrative
Tsai directly challenged the prevailing narrative that model companies, the organizations developing frontier large language models, will capture the lion’s share of AI’s economic value. “Right now, the model companies, pure model companies, are very hot. They seem to accrue a lot of the value. But over time, that may not be the case,” he said during a panel discussion at VivaTech.
This statement carries particular weight given the current market environment, where companies like OpenAI, Anthropic, and Zhipu AI have attracted extraordinary valuations and investment rounds. Tsai’s skepticism suggests that Alibaba’s leadership views the current valuation multiples for pure-play model companies as potentially unsustainable and that the real winners in AI will be companies that control multiple layers of the value chain.
Implications for the US-China AI Competition
Alibaba’s full-stack strategy also carries geopolitical significance. As the United States has tightened export controls on advanced semiconductors and AI models, Chinese technology companies face mounting pressure to develop domestic alternatives across the entire AI stack. Alibaba’s investments in semiconductor design, cloud infrastructure, and foundation models position the company to operate independently of US technology, reducing its vulnerability to future export restrictions.
(Related: Five AI Models, Zero Nvidia: China’s Labs Are Proving Domestic Chips Can Train Frontier AI)
The company’s willingness to invest across all four layers also reflects a broader shift in how Chinese technology giants approach AI development. Rather than waiting for US companies to develop technologies and then reverse-engineering or licensing them, companies like Alibaba are building complete, vertically integrated AI ecosystems from the ground up.
The Uncertainty Premium
Tsai’s “all in” framing also implicitly acknowledges the profound uncertainty surrounding which layers of the AI stack will ultimately prove most valuable. By investing across chips, infrastructure, models, and applications, Alibaba is essentially hedging against multiple possible futures. If chips become the primary bottleneck, Alibaba’s semiconductor investments will pay off. If infrastructure becomes the critical differentiator, Alibaba Cloud’s investments will prove decisive. If applications prove to be where value concentrates, Alibaba’s consumer and enterprise products will drive returns.
This diversified approach contrasts with the strategies of venture-backed AI startups, which typically focus on a single layer, usually foundation models, and rely on larger technology companies to provide the infrastructure and distribution channels. Alibaba’s scale and capital resources allow it to pursue a more comprehensive strategy.