by Raphael Dudler | Jul 27, 2026 | europe & ai
On August 2, 2026 — six days from now — the EU AI Act stops being a law with obligations but no teeth. The European Commission gains the power to fine general-purpose AI providers up to 3% of global turnover, the Article 50 transparency rules on chatbots and deepfakes activate, and national market-surveillance authorities can investigate and sanction. But roughly 24 companies signed the voluntary compliance code while Meta declined outright and xAI signed only part — and all of this arrives as Washington treats the American labs as strategic assets. gafam.ai reads the real test: not what the law says, but whether Europe dares to enforce it.
by Raphael Dudler | Jul 20, 2026 | europe & ai
SAP has completed its acquisition of Prior Labs, the German pioneer of tabular foundation models, committing over €1 billion over four years to build it into a European frontier AI lab. The target is not language models but the structured data — tables, transactions, records — that businesses actually run on, and that LLMs handle poorly. Founded just 18 months ago, Prior Labs went from research project to billion-euro European AI lab faster than almost any startup on record. gafam.ai reads why this is the clearest example yet of the strategy by which Europe can genuinely win.
by Raphael Dudler | Jul 14, 2026 | gafam watch
TSMC reported record June and Q2 2026 revenue — June sales up 67.9% year-on-year to the largest monthly figure in its 40-year history, driven by AI chip demand, with 3nm and advanced packaging sold out through year-end. The company makes nearly every leading-edge AI chip on earth and holds 73% of the foundry market, concentrating the entire AI economy on one island in a geopolitically tense strait. But beneath that concentration lies a fact rarely stated: none of those chips can be made without one machine, built only by Europe’s ASML. gafam.ai reads the chokepoint Europe actually holds.
by Raphael Dudler | Jul 9, 2026 | europe & ai
On July 8, France’s Mistral released Robostral Navigate, a physical-AI model that lets robots navigate complex environments using a single camera and basic language prompts — hardware-agnostic, trained entirely in simulation, and already tied to major European industrial customers. Alongside an open-weight frontier model entering July early access and a €4 billion data-center buildout, it reveals a deliberate strategy: Europe’s clearest AI champion is not trying to out-scale OpenAI at the chatbot frontier, but competing on sovereignty, openness and physical AI. After a month of documenting European dependency, gafam.ai reads what Europe is actually building.
by Raphael Dudler | Jul 6, 2026 | gafam watch
On June 30, Chinese food-delivery giant Meituan open-sourced LongCat-2.0 — a 1.6-trillion-parameter model released under a permissive MIT license, which the company says is the first model of its scale trained and run entirely on domestic Chinese chips, with no Nvidia GPUs. The benchmark claims are self-reported and the full weights aren’t yet public. But the hardware milestone is the story — and it lands as Washington gates access to American models. gafam.ai reads what it means for Europe’s impossible AI choice.
