Alibaba bets full-stack on AI: 5-10T param Qwen, Zhenwu V900 chip, 20GW cloud by 2032
Scobleizer · x · 2026-09-22
Speaking at Apsara Conference 2026, Alibaba CEO Eddie Wu said machine thinking is under 3% of human capacity, with a 1000x scaling runway. Alibaba is betting across all three layers: recursive self-improvement toward 5–10T parameter Qwen models targeting ASI, the Zhenwu V900 chip (China's most powerful, scaling to 500K cards per cluster), and global data center capacity exceeding 20GW by 2032—a full-stack bet on intelligence as infrastructure.
More from Companies & People
- Anthropic Is Setting Up a Biology Lab Where Claude Guides Robots Through Drug Experiments — The Decoder · 2026-09-22
- Ben Thompson: Meta's Muse Is a Bearish Signal for Frontier Labs — vaibhavbetter · 2026-09-22
- Report: OpenAI and Anthropic allegedly exaggerated AI security threats to protect market position — evilsocket · 2026-09-22
- Similarweb data shows AI bot app DAUs nearing 150K and still climbing — ns123abc · 2026-09-22
- OpenAI's biggest Codex Meetup draws engineers from Apple, Google, Meta — paw_lean · 2026-09-22
- UChicago president outlines integrating AI into revised core curriculum — lpachter · 2026-09-22