European firms reduce reliance on American AI providers

Traders Union
All News
Financial News
European firms spread AI risk

Highlights

U.S. limits on some Anthropic models exposed access risks for European firms.
Siemens, Renault, Orange and ChapsVision use multiple AI providers.
European companies want resilience through choice, not isolation.

Access risk becomes a business issue

The latest warning came after the U.S. government ordered Anthropic to suspend foreign access to its Fable 5 and Mythos 5 models, citing national security concerns. The move alarmed European technology users because proprietary AI services are typically accessed remotely and remain under the control of the developer, Reuters reports. If a provider is forced to restrict access, customers cannot simply move the same system onto their own servers.

Executives from Siemens, Renault, Orange and ChapsVision said at the VivaTech conference in Paris that they are already spreading AI workloads across multiple providers. Siemens uses models including China’s DeepSeek and Alibaba’s Qwen, Nvidia’s Nemotron and other U.S. and European systems. Renault works with Google, Microsoft, Mistral, DeepSeek and Dataiku, using a mix of open-weight and proprietary models.

For many companies, the goal is not to cut off U.S. technology, but to avoid being locked into it. Cedrik Neike, chief executive of Siemens Digital Industries, said sovereignty should not be confused with autarky, or complete self-sufficiency. The message from European industry is narrower: companies want the ability to switch providers if access, price or regulation changes.

Europe searches for control

The Anthropic restrictions have strengthened the argument for European AI capacity, particularly models that companies can run on their own infrastructure. France’s Mistral remains the region’s most visible general-purpose AI provider, while firms such as DeepL are stronger in specialized areas. But European executives acknowledge that the local market still lacks the depth of U.S. and Chinese alternatives.

Orange said open-source or open-weight models can reduce data-transfer risks when operated on European infrastructure. ChapsVision, which has won government contracts in France and Germany as an alternative to Palantir, uses models from Mistral, Anthropic, OpenAI, and Qwen. Its approach reflects a broader view: sovereignty means having credible substitutes ready, not relying on one national champion.

Cost adds a second pressure point

Access is not the only concern. Token costs — the fees charged for processing information through AI systems — are rising as companies deploy software agents that can perform tasks automatically. Orange said executives could become increasingly focused on cost per token by the end of the year.

Renault said rising AI token costs are already forcing it to adapt. Celonis, whose clients include BMW and Siemens, said companies need better internal context models before deploying AI agents. Without that, AI systems must repeatedly extract basic facts from company data, which can quickly inflate usage bills.

The result is a more pragmatic European AI strategy. Companies are not abandoning U.S. models, but they are building portfolios of providers, infrastructure and cost controls. The Anthropic episode has made one point clearer: for large companies, AI dependence is now a supply-chain risk, not just a software choice.

We have previously highlighted that EU prepares new cloud rules for Amazon, Microsoft, and Google.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.