Alibaba reportedly plans to train an artificial intelligence model with 5000000000000 to 10000000000000 parameters.
Article
Alibaba plans to train an artificial intelligence model with 5 to 10000000000000 parameters
Beating AI Flash reported, citing market sources, that Alibaba plans to train an artificial intelligence model with 5000000000000 to 10000000000000 parameters. Alibaba CEO Wu Yongming said at the Yunqi Conference that current customer demand for AI is very strong. The company will devote its full efforts to developing AI infrastructure, with the goal of expanding the global data centers operated by Alibaba Cloud to more than 20GW by 2032. On the same day, T-Head released the Zhenwu V900, its most powerful domestically produced AI chip, whose computing power is 3 times that of the Zhenwu M890. Reports indicate that T-Head’s annual AI chip shipments will increase substantially.
Original link: https://m.theblockbeats.info/flash/368343
Key points
01
Market sources say Alibaba plans to train an artificial intelligence model with 5000000000000 to 10000000000000 parameters.
02
Alibaba CEO Wu Yongming said at the Yunqi Conference that current customer demand for AI is very strong.
03
Alibaba will devote its full efforts to developing AI infrastructure, with the goal of expanding the global data centers operated by Alibaba Cloud to more than 20GW by 2032.
04
T-Head released the Zhenwu V900, whose computing power is 3 times that of the Zhenwu M890.
05
Reports indicate that T-Head’s annual AI chip shipments will increase substantially.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
Parameter count refers to the scale of the parameters contained in a model; the article says this model’s scale is 5000000000000 to 10000000000000. GW is a unit of power used in the article to describe data center scale.
Why it matters to readers
If implemented, the model training and data center development described in the article would involve substantial AI infrastructure investment.
This report covers AI model scale, data center development, and AI chip performance, helping explain why training large models requires supporting infrastructure.
Risks and unknowns
The model plan is attributed to “market sources”; the article does not state whether it has been officially confirmed.
The article does not specify the model’s name, training timeline, training results, or technical approach.
The article provides no update on progress toward the goal of exceeding 20GW in data center capacity.
The article provides no specific figure for annual AI chip shipments.
Related Developments
Loading event timeline…
Related concepts
Parameter count
This term is not in the glossary yet. Browse related concepts in the glossary.
