Urgent.News

What's breaking now, across thousands of outlets.

AI

La brutal economía de la IA china encierra lecciones para EE UU

Los modelos del país asiático compiten mediante eficiencia, volumen y precios muy inferiores

La brutal economía de la IA china encierra lecciones para EE UU

China's rapid AI development contrasts sharply with the U.S.'s, fueled by massive capital funding and advanced tech from U.S. companies. Despite China's investments lagging behind U.S. firms' spending on chips, data centers, and related capital projects, Chinese AI companies are making strides in cost-efficiency and scale. A key factor is the prevalence of open-source models that can be downloaded, modified, and used by users, rather than subscription-based models favored by U.S. firms.

This strategy enables Chinese firms like Alibaba's DeepSeek and MiniMax to generate revenue through API usage and app subscriptions. Alibaba, for instance, has seen its API division's gross margin rise to 25% following cost reductions in inference services. While these Chinese companies still operate with slim profit margins compared to the U.S., they are leveraging their own technical expertise and computational capabilities to improve performance and meet growing demand.

Written by urgent.news from El Pais Economia's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at cincodias.elpais.com →

More in AI

They Put 7 Attention Mechanisms on a Latin Square. Then Removed Them One by One.

Since GPT, nearly every Transformer repeats the same attention mechanism at every layer. Forty-eight identical blocks, differing only in learned weights. Nobody tested that. It is a convention, not a conclusion. A paper out of VIDRAFT AI Research ( arXiv:2609.20269 , CC BY 4.0) tests it, and the interesting part is not the headline.

  • Seven attention mechanisms arranged in a Latin square
  • Removing mechanisms one by one had negligible effect on performance
  • At least one mechanism from a different family crucial for maintaining performance

More from Friday 18 September →