Alibaba's aggressive ASI roadmap: Qwen models to 5-10T params, 20GW by 2032
bookwormengr · x · 2026-09-23
Alibaba reportedly laid out one of the most aggressive ASI roadmaps yet (unverified claim-by-claim):
- Qwen 4 already in training; Qwen 4.5 and Qwen 5 planned to scale to 5-10 trillion parameters
- Progress claimed on recursive self-improvement (RSI): models find their own weaknesses, design experiments, and synthesize data to improve themselves
- M890 supernodes already infer models above 2T parameters; V900 clusters scale to 500,000 cards
- Alibaba Cloud targets 20GW+ global datacenter capacity by 2032
- V900 moved up from Q3 2027 to mass production in Q1 2027
- Eddie Wu says machine "thinking" will eventually exceed all humans
Commenters argue private enterprises, if not governments, are clearly "AGI/RSI pilled," making dialogue and cooperation important.
More from Companies & People
- Databricks research lead is hiring — candidate inspired by the team's own papers — jefrankle · 2026-09-23
- NVIDIA's AI Day Singapore pushes Southeast Asia public-sector AI from pilots to production — nordicinst · 2026-09-23
- Aja Frost, HubSpot's AEO strategy lead, departs — who takes over? — gaganghotra_ · 2026-09-23
- Game dev looks like the programming field most resistant to AI — how much is it actually used? — marktenenholtz · 2026-09-23
- Claude Partner Network emails keep coming while our accounts sit banned — jdjohnson · 2026-09-23
- Microsoft plans ~30% per-seat Copilot discounts for 1,000+ seat buyers, per The Information — rohanpaul_ai · 2026-09-23