News
Europe Asks Whether It Can Train a Frontier AI Model on the Compute It Actually Owns
A widely discussed GitHub project and analysis ('euromesh', 137 points / 273 comments on HN) examines whether Europe's domestically controlled compute is sufficient to train a frontier-scale model without reliance on US hyperscalers. The debate sharpens the EU's sovereignty problem as capex and chip access concentrate among a few American and Chinese players. It crystallizes a recurring policy anxiety into a concrete capacity question.
↳ Follow the thread