Alibaba’s Pitch for Europe: Can Open Models Deliver AI Independence?

AI-generated illustration of an office workstation beside a local server cabinet, representing an organisation running its own AI systems.
AI-generated illustration for Routes, Bytes & Budgets: an organisation operating local computing infrastructure. This is a conceptual scene, not an Alibaba office or a tested installation.

I like the idea of software that remains useful even if its supplier changes direction. It is one reason open models interest me: keeping a working copy can offer a different kind of control from depending entirely on a remote service.

That is the appealing part of Alibaba chairman Joe Tsai's argument about Europe's AI future. But I think the practical meaning of “sovereignty” deserves more attention than the slogan.

Tsai's proposal: use open models and keep control locally

Speaking at Wave by Vento in Turin on October 7, Tsai argued that open-source AI offered Europe its route to technological independence. In the event coverage, he describes European companies adapting models using their own industrial data and running them on infrastructure they control.

I understand the attraction. A business may have valuable internal documents or manufacturing knowledge that it does not want to send to an external model service. The ability to choose where processing happens creates a useful option.

Tsai also has a commercial interest in this argument: Alibaba develops the Qwen family and sells cloud infrastructure. I take his “only way” position as his assessment, not a settled finding that Europe has exactly one viable strategy.

Downloadable weights are a starting point

A concrete example is the official Qwen3.8-27B model card, which lists Apache 2.0 licensing. That gives a prospective user something specific to examine: the model files, licence and supported software.

Even here, the terminology matters. The Open Source Initiative's definition goes beyond access to trained weights, covering the information and code needed to study and modify the system. I do not assume a downloadable model automatically provides full transparency into how it was trained.

Nor do I assume the licence of one Qwen release applies to every model with the same family name. The exact release is what needs checking.

Who runs it after the download?

This is where the idea becomes more interesting to me. Once an organisation runs a model itself, someone has to maintain the software, control access, monitor problems and keep backups. Those jobs do not disappear because the model files are free to download.

There is also the hardware. The number of parameters is not a complete memory specification: the numerical format, context length, runtime and workload affect the machine required. A model that can be downloaded is not necessarily a model that runs comfortably on the office laptop.

That is why I find small local models for existing devices interesting alongside larger releases. For many tasks, the useful question is how little infrastructure gets the job done well.

My practical test for independence

I would start with four questions before accepting a claim that a system is sovereign:

  • Can it keep working? What happens if the original supplier changes its service, pricing or availability?
  • Where does information travel? A locally running model may still call external search, send logs, use cloud backups or connect to other services.
  • Can another team maintain it? Documentation, reproducible deployment and usable exports matter as much as possession of the files.
  • What is the full operating cost? Include computing, electricity, support and staff time, then compare that with the value of the additional control.

Running a model inside Europe can help meet a location requirement. On its own, it does not tell me who holds administrative access, which subcontractors are involved or what dependencies remain.

More choice is the part I want to keep

My preference is for useful options: a local model for suitable private work, external services where they make sense and a practical route between them. The balance will differ for a researcher, a small business and a national institution.

I find Tsai's argument persuasive when it is about expanding those choices. The harder work is building the infrastructure and expertise to exercise them. Independence becomes meaningful when someone can explain how the system keeps working, who can access it and what it costs to maintain.

Comments

Popular posts from this blog

Obsidian’s New Search: Less Friction for Notes in Two Languages

OPPO Find X9 Ultra vs Find X10 Pro Max: Is the Older Ultra Still Worth Buying?

OPPO Find X10 Technical Preview: Cameras, Design and Missing Specs