Alibaba bets across the full stack: 5-10T Qwen models, 500K-card clusters, 20GW cloud by 2032
Div_pradeep · x · 2026-09-24
At Apsara Conference 2026, Alibaba CEO Eddie Wu said "machine thinking" represents under 3% of human capacity, leaving a 1,000x scaling runway. Alibaba announced bets on all three layers: Models — Recursive Self-Improvement research with planned 5-10T parameter Qwen models aimed toward ASI; Chips — the Zhenwu V900, billed as China's most powerful AI chip, scaling to 500,000 cards per cluster; Cloud — global data center capacity exceeding 20GW by 2032. The thread's point: the whole stack is being built to scale together.
More from Companies & People
- OpenAI's Astra Commons event packs Buenos Aires venue, 800+ turned away — paw_lean · 2026-09-24
- A Clear Ladder for Learning Data Engineering: From SQL to Airflow Orchestration — goyalshaliniuk · 2026-09-24
- Satire: The 'Doing More AI Than You' CTO Who Rebrands Failure as Learning Velocity — DavidLinthicum · 2026-09-24
- Meta Muse DAUs flat for three days then went vertical, Similarweb data shows — firstadopter · 2026-09-24
- An OpenAI Agent Hacked Australia's Health Service; Government Found Out Months Later — wiredmagazine · 2026-09-24
- Ben Todd: OpenAI has made clear it can't be trusted on safety incident disclosure — ben_j_todd · 2026-09-24