🔴 4 Views on the Future of AI

AI is gradually becoming the main subject of this newsletter, from a perspective broadened to 2025-2026.

I am preparing my new talk, The Sovereign Way: New Rules for a New Internet, for the fall. It brings in the notions of cognitive autonomy and post-sovereignty, and I hope it will break with the theoretical, disconnected discourse carried by a new generation of consultants.

For this July newsletter, I offer you an analysis of four texts that help anchor us in reality:

• An analysis of AI sovereignty as seen from Switzerland

• A view of China's AI strategy

• An update on the new profiteers of AI

• An analysis by an OpenAI alumnus of the reasons behind the company's success

For your information, the last two texts are reserved for paying subscribers.

We are switching to vacation mode, but I am preparing a list of the best texts from this newsletter for our new members.

The weekly publishing rhythm will continue in the fall, but the organization will be improved.

The Digital Resilience Council (Conseil de la Résilience Numérique), informal until now, is finally going to take flight, and I am preparing a preview presentation of my new keynote as well as meetups.

Consider becoming a paying member (the price will go up a little in the fall).

Naturally, my monitoring for paying subscribers does not stop during the vacation, and you will find the link at the end of the newsletter.

Enjoy your vacation, disconnect a little, and see you sometime in August!

AI sovereignty: the urgency of betting on talent

Sovereignty in AI has become a strategic goal openly pursued by many countries. But according to Marcel Salathé, a researcher at EPFL, the concept is an illusion for nearly every state, a conviction forged in Switzerland, where the subject is everywhere yet disconnected from reality.

Sovereignty implies decision-making independence and control of the technological means: computing power, semiconductor infrastructure, massive access to data, energy independence, an autonomous regulatory framework and a pool of talent. Yet only the last of these conditions is met in a country like Switzerland. The others involve a technological, industrial or geopolitical gap that is almost impossible to bridge.

Political rhetoric on the subject often masks a simple reality: for most countries, aiming for total sovereignty is a costly and counterproductive delusion. Even the European Union, the only entity outside the two superpowers with the potential for critical mass, struggles to adopt an ambitious strategy equal to what is at stake.

The realistic alternative is called talent sovereignty. Training, attracting and retaining the technical skills that make it possible to appropriate and adapt existing technologies is now the most effective lever. It allows rapid adoption, local adaptation, diffusion across sectors and value creation. Provided, however, that a high level of investment in education, research and innovation is maintained.

What is missing is not ambition but the alignment of means. Academic rankings starkly illustrate the gap between the United States and Europe, particularly in the production of talent in computer science and artificial intelligence.

For decision-makers as for investors, the challenge is not to dream of complete technological autonomy, but to build the conditions for a capacity to act that is grounded in skills. Because in the AI economy, it is uses, not infrastructure, that will create tomorrow's margins.

Giving up the illusion of sovereignty means making the strategic choice of realism.

LINK

AI in China: deploy everywhere

According to Axel Badalian, China is not trying to outpace the West in AI, but to anchor it in every link of its economy.

While the United States bets on technological breakthroughs to maintain its lead, China's strategy rests on a logic of endurance: industrialize AI at scale, quickly integrate “good enough” models and deploy them at every level. This approach favors infrastructure over disruptive innovation, with tight coordination between talent, data and public policy.

Initiatives such as “AI+” reflect this intent: predictive maintenance in factories, automated medical triage, optimized logistics. The goal is to turn AI into a macroeconomic lever to offset demographic constraints, deflation and the end of debt-fueled expansion. But this efficient centralization carries a risk: premature technological lock-in.

For startups, this means fertile ground for industrial applications, but under heavy pressure for profitability and control. For corporate customers, it is the promise of greater efficiency, at the price of growing dependence on standardized systems. For investors, what matters is no longer only innovation, but the ability to industrialize quickly.

This model is accelerating as China applies to AI the recipes that allowed it to dominate electronics, electric vehicles and e-commerce: mass production, vertical integration, local adaptation.

The global question is no longer who will reach AGI first, but who will manage to embed intelligence across the whole economy, quickly, broadly and sustainably.

LINK

For our subscribers, two more texts to read this summer (with our analysis):