A German search engine tagged with tree planting and independence has replaced the flagbearer of European AI. Based in Berlin, Ecosia recently announced a shift in its artificial intelligence strategy, ending its partnership with French Mistral and turning to various open-source solutions, including Chinese open-source models. This move has caused quite a stir in the European tech circle — because just a few months ago, Ecosia had seen Mistral as a lifeline to reduce dependence on American giants.
Ecosia has long embedded environmentalism and technological independence into its brand core. Its products have already entered public institutions such as the German Federal Environment Ministry and the UK's National Health Service (NHS). In May this year, it just replaced OpenAI with Mistral as its main AI provider, aiming to build a more sovereign AI ecosystem using local European technology. But less than half a year later, this "European marriage" turned red.
Founder and CEO Christian Kroll gave a straightforward reason for the breakup: the company was not satisfied with Mistral's overall model performance, believing that its current model capabilities are about a year behind major competitors and cannot meet Ecosia's development needs. What's more, technical issues occurred frequently during the cooperation — Mistral's servers were overloaded multiple times, and service stability was compromised; Kroll even said directly that Mistral is "too big" for Ecosia, and its existing infrastructure could not handle their traffic.
More subtly, there was a mismatch in values. Ecosia questioned whether Mistral truly aligned with its long-term strategy: France's energy structure has a high proportion of nuclear power, which somewhat conflicts with the company's environmental narrative; it further doubted whether Mistral could represent true European technological sovereignty, as the latter also relies on international capital, which, in Kroll's view, does not count as technological independence.
Facing the public criticism from its former client, Mistral co-founder and chief scientist Guillaume Lample did not remain silent. He publicly called on Ecosia to test Mistral's latest released model and promised to immediately open up early access permissions; he emphasized that building advanced AI requires time, computing power, and research accumulation, and Mistral is in a period of rapid growth, and it also has the ability to customize models according to customer needs. This response shows reluctance but also indirectly proves the mismatch between the two parties' paces — one is eager for a stable and usable ready-made solution, while the other is still on the way to growth.
