Alibaba's Zhenwu V900: China's Most Powerful AI Chip

Key Takeaways

• Alibaba's Zhenwu V900 chip offers three times the performance of the previous Zhenwu M890
• Alibaba plans Qwen models up to 10 trillion parameters, compared to 2.4 trillion in current flagship
• Alibaba aims for over 20 gigawatts of computing capacity by 2032
• US export controls on Nvidia chips are pushing Chinese firms toward domestic alternatives
• ChatGPT's share of AI chatbot prompt volume dropped from 70% to 50% in the first half of 2026
• Gemini grew from 17% to 30% of prompt volume while Claude rose from 2% to 11%
• Maryland Governor Wes Moore proposed AI rules covering worker protections, discrimination, and child safety
• An ILO study found AI boosts productivity in Chinese firms but reduces job numbers
• Researchers warn isolated users are more likely to replace human connection with AI companions
• OpenAI's Navier-Stokes proof uses an external force, which mathematicians say does not solve the actual problem

Alibaba reveals powerful new AI chip and larger model plans

Alibaba announced a new AI chip called Zhenwu V900 at its Apsara Conference in Hangzhou. The chip delivers three times the performance of its predecessor. The company is also developing a Qwen AI model up to four times larger than its current flagship. These moves come as Chinese firms seek alternatives to Nvidia chips due to US export limits.

Alibaba launches Zhenwu V900 chip and plans 10 trillion parameter models

Alibaba unveiled its Zhenwu V900 AI chip, built by its semiconductor arm T-Head. The chip offers three times the performance of the earlier Zhenwu M890. Alibaba also plans to train future Qwen AI models with up to 10 trillion parameters. This compares to its current Qwen3.8-Max model at 2.4 trillion parameters. The announcements aim to help Chinese companies compete amid stricter US export controls on Nvidia chips.

Alibaba reveals bold AI chip and model expansion plans

At its annual conference in Hangzhou, Alibaba CEO Eddie Wu called the new Zhenwu V900 chip the most powerful AI chip in China. It delivers three times the performance of the previous generation. Alibaba plans to train AI models with five to 10 trillion parameters to match advanced US models. The company also expects to reach over 20 gigawatts of computing capacity by 2032. Wu compared AI growth to the industrial revolution.

Haryana upgrades 60 ITIs with focus on AI and EV training

Haryana is overhauling 60 government ITIs across 12 clusters under the Rs 60,000-crore PM-SETU programme. The state already uses an industry-linked training model through Maruti Suzuki's Japan-India Institutes for Manufacturing. Annual apprentice engagement grew from 5,225 in 2016-17 to 86,565 in 2025-26. Cumulative engagement since 2016-17 crossed 4.73 lakh by September 19. The ITIs will now include AI, EVs and cybersecurity training.

AI chatbot usage spreads beyond ChatGPT as rivals grow fast

Comscore's Q2 2026 report shows AI chatbot usage is fragmenting across platforms. ChatGPT's share of prompt volume fell from 70% to 50% between January and June 2026. Gemini grew from 17% to 30% and Claude rose from 2% to 11%. Google's AI Overviews in search also grew from 25.8% to 39.4%. ChatGPT reached 168 million desktop conversations in June 2026 while Gemini hit 89 million.

Squirrel photographer and Taylor Swift insects top weird AI-era news

A columnist shared funny animal and nature stories to take a break from AI fears. A University of California doctoral student named a new genus of insects after Taylor Swift. The insects include four species with names inspired by Taylor Swift and her albums. The author also wrote about lovebugs in South Carolina that mate in pairs and stick to cars. He joked that killer AI bots can wait since nature is already wild enough.

Maryland Governor Wes Moore proposes new AI rules for state

Maryland Governor Wes Moore announced proposals to regulate artificial intelligence in the state. The rules would regulate frontier AI companies with worker protections and incident reporting. They also aim to stop algorithms that discriminate in tenant screening or target problem gamblers. Moore wants AI protections for children in schools and for private-sector workers. Some proposals can start through executive action while others need approval from state lawmakers.

New book pairs human and Claude AI to guide everyday users

Steve Schonberg worked with Claude, an AI assistant from Anthropic, to write a book called With All My Circuits Showing. The book helps regular workers learn to use AI tools for jobs involving words and ideas. Claude admits where it falls short, like agreeing too easily with users and failing some safety tests. The guide covers platforms including Claude and gives practical advice for teachers, writers and sales professionals. The paperback is being distributed to bookstores and libraries worldwide.

Trump orders U.S. government to call AI super intelligence

President Trump said U.S. government documents will use the term super intelligence or SI instead of artificial intelligence. He argued the new name is more accurate because artificial can sound fake. Trump said the U.S. leads China in this technology and should keep that lead. He also announced an AI Force and an AI czar. Experts note that superintelligence already has a technical meaning for AI that exceeds human abilities, which could cause confusion.

Study warns AI companions may harm isolated users

Researchers from Singapore Management University and Duke-NUS Medical School studied AI companions and inequality. They found lonely or socially isolated people are more likely to use AI as a replacement for human connection. People with strong social ties use AI to supplement their relationships instead. The study created a framework using the Swiss cheese model to show how failures in multiple layers create risk. Governance was identified as the biggest gap since AI companions are mostly unregulated.

OpenAI may have solved the wrong Navier-Stokes problem

OpenAI created a proof for the Navier-Stokes problem that could win a $1-million prize. The proof uses an external force like gravity to make the math break down. Most mathematicians want to solve the problem without any outside force. They say OpenAI's answer does not solve the real problem they care about. The result has created debate about whether the achievement truly matters.

AI adoption in Chinese firms boosts productivity but cuts jobs

A study by the International Labour Organization looked at AI use in Chinese enterprises. It found that AI makes workers more productive but reduces the number of jobs available. Workers may need to learn new skills to stay employed. The ILO says governments and businesses must invest in training. It also calls for careful planning so AI benefits are shared fairly.

Sources

• Alibaba deepens AI push with new chip and bigger model
• Alibaba unveils new AI chip to challenge NVIDIA, plans Qwen models with up to 10 trillion parameters
• China's Alibaba unveils new powerful chip and ambitious AI model plans
• AI, EVs, cybersecurity on course list as Haryana upgrades govt ITIs
• Comscore data shows how AI discovery is splintering beyond ChatGPT
• DeWitt Humor: I trust rabid squirrels more than I trust artificial intelligence
• Moore proposes AI rules to protect Marylanders, wading into national debate
• Human and Artificial Intelligence Collaborate on Tell-All Guide to Getting the Most Out of AI Platforms and
• Trump Renames AI “Super Intelligence” For U.S. Government Documents
• AI companions may erode social skills among isolated users, framework suggests
• Did OpenAI solve the wrong Navier-Stokes problem?
• AI adoption in Chinese enterprises boosts productivity but raises concerns about jobs and skills