Alibaba's Joe Tsai: Open-Source AI Is Europe's Only Route to True AI Sovereignty

Alibaba Chairman Calls Open-Source AI Europe's Independence Strategy

Alibaba co-founder and chairman Joe Tsai delivered a direct message to European leaders and investors at Wave by Vento in Turin on Wednesday: open-source AI represents the continent's only viable path to AI sovereignty. Speaking in a session titled "The Other Side of the Story," Joe Tsai emphasized that Europe must embrace open models, train them on European data, and run them on local infrastructure to achieve true technological independence1

1

. "Open source is the opportunity for Europe... That's the approach Europe can take to truly be independent. We talk about technology independence, we talk about AI sovereignty in Europe. I think the open source approach is the only way to go," Tsai stated during the interview led by James Anderson, managing partner of Lingotto Innovation1

1

.

Why Europe Needs Its Own AI Infrastructure

Tsai warned that no country should fully trust another nation's technology, as political shifts or changing circumstances could result in access being cut off. This applies equally to American and Chinese technology, making multiple options essential for European firms' competitive edge1

1

. The ECB's Christine Lagarde echoed similar concerns in September, warning that Europe risks being isolated without developing its own AI capabilities. The missing piece, according to Tsai, is computing power. Europe needs to "get serious" about building data centers with the capacity to train models and run inference on European infrastructure2

2

. Alibaba Cloud demonstrated this commitment by opening data centers in France in June, showing how companies can establish local infrastructure to support regional AI ambitions1

1

.

Proprietary Industrial Data as Europe's Competitive Advantage

Tsai identified proprietary industrial data from European factories as an untapped resource that could give the region a significant advantage. Industrial companies around Turin and across Europe hold valuable manufacturing data they would be reluctant to send to closed models controlled by foreign entities1

1

. "In many countries in Europe, we still have large industrial bases here," Tsai noted, drawing parallels to how Chinese factories leverage manufacturing data to train models2

2

. By taking open-weight models and training them further on this proprietary industrial data, European companies can build specialized AI systems that run in their own data centers, avoiding dependency on closed systems while maintaining control over sensitive information1

1

.

Talent Exists But Investment Infrastructure Lags

Addressing concerns about Europe's ability to compete with the US and China, Tsai emphasized that talent is not the problem. Europeans make up 20% to 30% of staff at some US frontier labs, demonstrating the region's strong AI research talent pool1

1

. However, Anderson pointed to a critical gap: Europe lacks companies that reinvest cash flows into AI infrastructure the way US hyperscalers, Tencent, or ByteDance do. Tsai revealed that Alibaba uses about $25bn annually in free cash flow from e-commerce to fund its AI push, doubling capital spending on computing infrastructure each year for the past three years1

1

. By comparison, US hyperscalers together will invest around $1tn this year, a scale that neither China nor Europe currently matches1

1

.

How Scarcity Drives Innovation in AI Development

When asked whether US chip restrictions had pushed Chinese companies to innovate, Tsai confirmed they had, comparing the situation to growing up without wealthy parents. "If you are scraping all along, that drives you," he said1

1

. Despite US curbs on chip supplies, Chinese companies continue to open-source their work on model design and inference efficiency while publishing research papers. Tsai contrasted this with closed US labs that no longer publish because they want to protect their competitive advantages1

1

. China's complete supply chains for electric vehicles, batteries, and robots, combined with accounting for about 30% of global industrial production, provide additional edges in AI development1

1

.

Open Models Enable True Independence and Cost Savings

Tsai compared open-weight models to recipes that anyone can take and improve, offering a path to avoid dependency on closed systems. Companies can run these models in their own data centers, train them on proprietary data, and no longer depend on the original developer1

1

. This approach also eliminates ongoing costs associated with accessing closed models, making it financially attractive for European firms. Mistral, the Paris-based AI company, demonstrated this approach by releasing its Large 4 model with open weights on Tuesday, just before Tsai's speech1

1

. Tsai predicted that within five years, AI will be embedded in every business just as the internet is today, making current conversations about AI as a separate technology obsolete1

1

.