Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

Dominic Loyster argues Europe should not focus on training leading frontier AI models, which it lacks the compute infrastructure to do independently anyway. Instead, near-frontier models like Mistral adequately satisfy the business demands of European corporations. Prioritizing technological sovereignty and avoiding complete dependence on US and Chinese tech stacks serves Europe better than trying to win an expensive frontier race.

Related Insights

Writer's CEO claims enterprises are tired of the high costs and lack of control associated with "frontier" models from big labs. The market is shifting towards purpose-built, sovereign AI solutions that deliver reliable performance at a lower cost, creating an opening for specialized providers.

Europe's backing of Mistral is a strategic geopolitical play to create a sovereign AI alternative, much like it did with Airbus to compete with Boeing. The goal is to avoid technological dependence on the US, whose government can restrict access to its AI models, rather than to create the world's best model.

Mistral's $1.4B investment in Swedish AI infrastructure is more than an expansion; it's a political move. By building a "fully European AI stack," Mistral is positioning itself as the regional alternative to US tech giants, capitalizing on growing desires for data sovereignty amid fraying political ties.

Most nations' sovereign AI strategies will not involve creating frontier models from scratch. Instead, they will adopt the best open-source models, customize them with local data and values, and run them on-premise for national security.

Lagarde concedes the U.S. leads in pioneering AI due to advantages in chips, data, and capital. She argues Europe's competitive strategy should be to excel at the rapid and widespread diffusion of AI technologies, especially within its small and medium-sized enterprises.

Lacking the hardware, frontier models, and data centers, Europe's pursuit of AI sovereignty is futile. The optimal strategy is to be a fast, aggressive adopter of American-developed AI. By becoming a critical customer, Europe can capture the vast welfare gains from US R&D and gain bargaining power with vendors.

A global trend is emerging where nations refuse to be dependent on closed-source American AI. They are actively building their own "sovereign AI" stacks, often using open-source models, preferring to control their own destiny even if the technology is only 95% as good.

Europe is in a strategic trap: it wants to regulate AI for safety but lacks a domestic frontier AI industry to give it leverage. Over-regulation could cause US AI labs to either abandon the European market, using the freed-up compute to accelerate R&D, or serve Europe with weaker, compliant models.

The likely path for most countries' sovereign AI strategies is not to compete with the US and China in building frontier models from scratch. Instead, they will license the best available open-source models and then use reinforcement learning and supervised fine-tuning to align them with their specific language, culture, and values.

Europe can secure its place in the AI future through a 'compute for access' deal. By providing favorable conditions for American hyperscalers to build data centers, Europe can demand guaranteed access to the frontier models running on them. This leverage is backed by Europe's control over key semiconductor manufacturing equipment from companies like ASML.

Adopting Near-Frontier AI Models Fits European Corporate Needs Better Than Frontier Chasing | RiffOn