Alibaba

Headlines · 4

1. 2h
   Alibaba's Joe Tsai: open source is Europe's only path to AI independence

   Alibaba chairman Joe Tsai said at the Wave by Vento event in Turin that open-source AI is the only way for Europe to achieve technology independence and "AI sovereignty," urging the region to build its own computing infrastructure to train and run models. He said Alibaba narrowed itself in 2023 to two businesses — e-commerce and "full-stack" AI — and uses about $25bn a year of e-commerce free cash flow to fund the AI push, doubling computing-infrastructure capex each of the past three years. Tsai also estimated US hyperscalers will invest roughly $1tn this year, a scale China cannot match, though China accounts for about 30% of global industrial production, giving it valuable factory data and EV, battery and robotics supply chains.

   Bloomberg Technology · 🔥 19
2. 6h
   Researchers link AI agent fleet to Tencent and Amap scraping

   Researchers at Swarmchasers say a fleet of AI agents, likely using Tencent’s Hunyuan models, spent more than a week querying Alibaba’s Amap mapping service through Tencent Cloud-linked infrastructure. The preliminary report describes attempts to bypass anti-bot protections and says 211 scans were labeled “claude,” although the researchers’ tests found the behavior more consistent with Tencent models than Anthropic’s Claude.

   Tom’s Hardware · 🔥 8
3. 2d
   China’s 1990s-born AI leaders take the helm as the real test begins

   A 36Kr feature examines a group of Chinese AI leaders born in the 1990s who now oversee major model, multimodal, agent and AI infrastructure efforts at Alibaba, Xiaomi, ByteDance and Tencent. It argues that their rapid promotions reflect the shorter cycle of the foundation-model industry, while emphasizing that technical performance, commercialization, cost control and organizational execution remain unproven.

   36氪 人工智能 · 🔥 3
4. 2d
   Strata reportedly runs 125B Qwen model on 12GB GPUs

   Developer Niko1221 has open-sourced the Strata engine, which reportedly runs a quantized Qwen3.8-Flash-Next model with 125 billion parameters on consumer GPUs with at least 12GB of VRAM. The engine keeps the MoE model in system RAM, loads frequently used experts into VRAM, and uses a lightweight model for speculative decoding; reported tests reached 94 tokens per second on an RTX 5070 with a 2-bit quantization.

   IT之家 AI · 🔥 3

Experience and discussion from the community

Share my Alibaba experienceAsk about Alibaba