Alibaba's Joe Tsai: Open Source AI Is Europe's Only Tech Sovereignty Path 2026

Open source AI is the only realistic path to technological sovereignty in Europe, according to Joe Tsai, chairman and co-founder of Alibaba. Speaking at the Wave event in Italy, Tsai argued that European companies should not park their critical data inside a closed API likely supplied by an outside vendor. Instead, he said, they should build on open models they can train and adapt themselves.

His claim is blunt and deliberately provocative. Tsai goes further than most industry voices by framing open source as the single viable route to European independence, not merely one option among several.

What Joe Tsai Actually Said About Open Source AI

At the center of Tsai's argument is a simple question about data control. When AI becomes essential to most productive processes, where should a European company keep its data? His answer: not in a closed API controlled by an external provider, but in an open model the company can fine-tune to its own needs.

Tsai also drew a direct contrast between regions. In his view, Chinese AI companies share their innovation and advances, while labs in the United States do not. That framing positions open source as a geopolitical tool as much as a technical choice.

The setting matters. The Wave event in Italy put the message in front of a European audience already anxious about its dependence on foreign technology, which gave the comments extra weight.

Why Open Source AI Matters for Europe's Tech Sovereignty

The backdrop to Tsai's remarks is a widening gap between Europe's AI ambitions and its actual infrastructure. Arthur Mensch, CEO of Mistral, has warned that Europe has roughly two years to build its own AI infrastructure. Despite efforts by European administrators and the business ecosystem, the continent still leans heavily on technology from the United States and China.

Recent events have made that dependence feel less abstract. The international blocking of Anthropic's most powerful models by the United States last summer showed that a government can shut off access on demand. For European firms, that is a supply-chain risk with no easy workaround.

China and the United States together control roughly 90 percent of the world's AI infrastructure, according to the reporting around Tsai's comments. That concentration is the core of the sovereignty problem.

How European Companies Are Responding

Some industrial giants are already hedging. Siemens and Renault now combine models from different origins as a precaution, mixing providers so that no single government or vendor can cut them off.

That strategy buys resilience but not independence. Combining foreign models still leaves the underlying infrastructure in the hands of the same two powers, which is exactly the gap Tsai says open source can close.

Here is what the current landscape looks like for European decision-makers:

• Closed APIs offer convenience but concentrate control with the vendor and its home government.
• Open models let companies train on their own data and keep it in-house.
• Multi-model strategies, as used by Siemens and Renault, reduce single-vendor risk but not foreign dependence.
• Infrastructure concentration remains the hard limit, with China and the US controlling about 90 percent of global capacity.
• Time pressure is real, with Mistral's CEO estimating a two-year window to build European infrastructure.

Closed APIs vs. Open Models: What's at Stake

The choice between closed and open models is not just technical. It determines who can see a company's data, who can change the terms of access, and who can switch off the service.

The table makes the trade-off clear. Open models demand more work, but they shift control back to the organization using them.

Who Is Affected and What to Watch Next

The stakeholders stretch well beyond AI labs. Manufacturers, banks, healthcare providers, and public agencies all face the same question about where their data lives once AI runs core processes.

For European policymakers, Tsai's comments add pressure to fund domestic compute and support open model ecosystems. For businesses, the practical takeaway is to audit which models touch sensitive data and how easily access could be revoked. The background to this story is set out in German Search Engine Ecosia Ditches Mistral for Chinese Open-Source AI Models. A related thread runs through ASML's Zero Europe Sales Show Europe's Chip Industry Gap. Earlier coverage in Europeans Want United Europe but Doubt Path to Achieve shows how the argument developed.

Watch three signals in the coming months: whether European firms shift budget toward open model training, how US export policy on advanced models evolves, and whether infrastructure investment in Europe closes any of the gap with China and the United States. A closer look at the same subject appears in AI Future Debate: Open Source vs Safety in Silicon Valley. Anyone following the wider picture can start with China Deploys Thousands of Delivery Robots Across Europe. The same theme surfaces in Meta's Superintelligent AI Vision: Open-Source Model for All, which takes a different angle.

Tsai's position will not settle the debate, but it sharpens it. The question for Europe is no longer whether to reduce dependence on foreign AI, but how quickly it can build alternatives it actually controls. Context on how this played out before sits in Meta's Open-Source Superintelligent AI: Zuckerberg's Vision for All.