Alibaba maps Qwen models toward 5–10 trillion parameters
Alongside the chip, Alibaba said it plans Qwen-scale models with 5–10 trillion parameters, up from about 2.4 trillion in Qwen3.8-Max, and projected more than 20GW of cloud compute capacity by 2032 as U.S.–China AI rivalry frames a Trump–Xi meeting.
AI-assisted · US Brief desk · Sources listed below
What happened
Alongside the chip, Alibaba said it plans Qwen-scale models with 5–10 trillion parameters, up from about 2.4 trillion in Qwen3.8-Max, and projected more than 20GW of cloud compute capacity by 2032 as U.S.–China AI rivalry frames a Trump–Xi meeting.
More context
Alongside the Zhenwu V900, Alibaba said it plans Qwen-scale models with 5–10 trillion parameters, up from about 2.4 trillion in Qwen3.8-Max, and projected more than 20GW of cloud compute capacity by 2032. Those targets remain roadmap claims pending silicon and power build-out.
2 listed sources
- AP
- Gadgets Now, Sep 23, 2026
References listed by US Brief; a source count is not a verification score.
Editorial sourcing notes
Explainer-style expansion of the Sep 23, 2026 briefing bullet. Stays within verified claims from named outlets; not an exclusive investigation.
Like US Brief? You can support it with a tip.