Alibaba plans AI model with 5 trillion to 10 trillion parameters, unveils new chip
Alibaba is training Qwen 4, a next-generation large language model targeting more than 10 trillion parameters. To support this scale, the company introduced a new AI processor it describes as the most powerful in China. The company announced these plans during its annual Apsara Conference. As of November 1, Alibaba has not released a model ID, pricing, or weights for Qwen 4. These efforts aim to expand domestic computing capabilities amid ongoing hardware restrictions.
Listen to Live Briefing
Real-time synthesized voice briefing · Live Feeds Desk
- ✓ Alibaba is training a next-generation model called Qwen 4.
- ✓ The company aims for Qwen 4 to exceed 10 trillion parameters.
- ✓ Alibaba introduced a new AI processor described as China's most powerful.
What changed
Alibaba identified the forthcoming model as Qwen 4 and confirmed its training status at the Apsara Conference.
Live updates
-
Alibaba targets 10 trillion parameters for Qwen 4 model
Alibaba is training Qwen 4, a next-generation large language model targeting more than 10 trillion parameters. To support this scale, the company introduced a new AI processor it describes as the most powerful in China. The company announced these plans during its annual Apsara Conference. As of November 1, Alibaba has not released a model ID, pricing, or weights for Qwen 4. These efforts aim to expand domestic computing capabilities amid ongoing hardware restrictions.
Why it matters
The development follows the rollout of the Zhenwu V900 processor, which features 216GB of memory. This hardware strategy seeks to bypass global high-bandwidth memory shortages. Alibaba is positioning these tools as domestic alternatives for advanced machine learning workloads.
What is confirmed
- Alibaba is training a next-generation model called Qwen 4.
- The company aims for Qwen 4 to exceed 10 trillion parameters.
- Alibaba introduced a new AI processor described as China's most powerful.
What to watch next
- Release of Qwen 4 model weights and pricing
- Official specifications for the new AI processor
confidence 100%Sources used for this update (4)
- economictimes.indiatimes.com — alibaba launch: Latest News & Videos, Photos about alibaba ...
- beijingpost.com — Alibaba Unveils New AI Chip and Plans Model With Up to 10 ...
- newslantern.com — Alibaba to Train Qwen 4 on Parameters Exceeding 10 Trillion ...
- omniakey.com — Qwen 4 Release Date: What Alibaba Has Confirmed So Far
-
Alibaba Unveils Zhenwu V900 Chip and Tests Qwen 4
Alibaba is advancing its artificial intelligence ecosystem with the rollout of the Zhenwu V900 processor and preparations for its next-generation language model. The new hardware features 216GB of memory to support models scaling up to ten trillion parameters, arriving during a worldwide shortage of high-bandwidth memory. Simultaneously, testing is underway for a forthcoming model developed under the codename Kalmia. These developments reinforce the company strategy to build domestic alternatives for advanced machine learning workloads against a backdrop of strict hardware limitations.
Why it matters
The hardware and model rollout follows prior updates detailing a full-stack artificial intelligence strategy that combines cloud infrastructure, proprietary silicon, and open-source models. The domestic market for large language model services reached a significant financial scale, providing a strong commercial foundation for these domestic infrastructure investments. Alibaba continues to position its hardware and software stack to secure local market share as global supply chains face constraints.
What is confirmed
- Alibaba's Zhenwu V900 chip packs 216GB of memory to support AI models scaling to ten trillion parameters.
Still unconfirmed
- The next-generation Qwen 4 model is being tested under the codename Kalmia, with tokenizer tests and kernel code pointing to an imminent launch.
What to watch next
- Official commercial release dates for the Qwen 4 model under the codename Kalmia
- Detailed benchmark performance metrics for the Zhenwu V900 processor under production workloads
confidence 90%Sources used for this update (2)
- shattered.io — Alibaba Zhenwu V900: 216GB AI Chip Targets 10T Params
- ia4pymes.tech — Qwen 4 Is Being Tested Under a Codename: Facts vs Rumors
-
Alibaba Unveils Full-Stack AI Strategy, New Chips, and Qwen Roadmap
Alibaba released comprehensive updates outlining a full-stack artificial intelligence strategy that integrates cloud infrastructure, models, and specialized silicon. The initiative includes a multi-year hardware roadmap alongside open-source developments tracked through software repositories like llama.cpp, vLLM, and Transformers. Meanwhile, market data from an International Data Corporation report indicates that China's large language model-as-a-service market reached CNY 10.63 billion, highlighting the commercial backdrop for Alibaba's infrastructure push as it works to establish domestic alternatives to foreign hardware.
Why it matters
The technology giant aims to reduce reliance on external hardware providers by scaling its custom silicon and advanced cloud services. This hardware and software integration supports the company's Qwen artificial intelligence models, which target massive parameter counts between 5 trillion and 10 trillion. As Chinese enterprise demand shifts toward domestic cloud solutions, hardware rollouts like these dictate market positioning.
What is confirmed
- Alibaba announced comprehensive updates and a roadmap for its full-stack artificial intelligence strategy.
- An International Data Corporation report shows that China's large language model-as-a-service market reached CNY 10.63 billion.
Still unconfirmed
- Alibaba has unveiled the Zhenwu V900 artificial intelligence chip for training and inference, according to aivancity.ai.
- GitHub pull requests in llama.cpp, vLLM, and Transformers point to Alibaba's next open Qwen model nearing release.
- Alibaba scheduled a new processor release for the first quarter of 2027.
What to watch next
- Official commercial release date and performance metrics for the Zhenwu V900 chip
- Formal deployment and public release details for the next open Qwen model tracked on GitHub
- Customer adoption rates for Alibaba's newly detailed full-stack artificial intelligence cloud infrastructure
confidence 80%Sources used for this update (5)
- aivancity.ai — Zhenwu V900: Alibaba's AI Chip, Performance and Limitations
- www.philstar.com — Alibaba unveils roadmap on full-stack AI strategy from chips ...
- finance.biggo.com — China's LLM Services Market Reshuffles as Alibaba Cloud ...
- startupfortune.com — GitHub pull requests point to Alibaba's next open Qwen model ...
- imperyomanila.com — Alibaba announces AI upgrades, schedules new processor ...
-
Alibaba Unveils New AI Chip and Prepares 10T Parameter Qwen4
Alibaba Cloud unveiled a new artificial intelligence chip called the Zhenwu M890 and revealed plans to scale its upcoming Qwen4 model to between 5 trillion and 10 trillion parameters. Developed by the company semiconductor unit T-Head, the Zhenwu chip series has shipped over 560,000 cumulative units to more than 400 customers across over 20 industries. Clients utilizing the technology include China Telecom, FAW Group, and SPD Bank. The company also announced a multi-year semiconductor roadmap aimed at creating domestic alternatives to Nvidia infrastructure.
Why it matters
Alibaba Group Holding Limited operates a massive multinational technology business encompassing e-commerce, cloud computing, and digital media. The recent hardware deployment and large-scale model training reflect a broader national push toward domestic semiconductor independence in China. By pairing its proprietary silicon with advancing language models, the technology giant aims to secure a dominant position in enterprise artificial intelligence.
What is confirmed
- Alibaba Cloud launched the Zhenwu M890 training-and-inference integrated AI chip developed by its semiconductor arm T-Head.
- The Zhenwu series has shipped more than 560,000 units cumulatively and serves over 400 customers across more than 20 industries.
- Clients of the Zhenwu series include China Telecom, FAW Group, and SPD Bank.
Still unconfirmed
- Alibaba is training Qwen4 on a new architecture with subsequent versions expected to reach 5 trillion to 10 trillion parameters.
- A next-generation video generation model is scheduled for release in November.
- Alibaba launched an AI smartphone full-stack solution called Qwen Intelligence.
- Alibaba has deployed 10,000 Zhenwu chips in a new China data center.
What to watch next
- Official commercial release and performance benchmarks for the 5 trillion to 10 trillion parameter Qwen4 model.
- Deployment updates on the next-generation video generation model slated for November.
confidence 90%Sources used for this update (3)
- www.computeforecast.com — Alibaba Deploys 10,000 Zhenwu Chips in AI Push
- global.chinadaily.com.cn — Alibaba unveils new AI chip - Chinadaily.com.cn
- internetquadrant.com — 阿里大模型新进展：Qwen4在训参数冲刺10T，视频生成与AI手机方案齐发
-
Alibaba Advances AI and E-Commerce Operations
Alibaba Group Holding Limited operates as a Chinese multinational technology company specializing in e-commerce, retail, Internet, and technology services. Founded in Hangzhou on June 28, 1999, the enterprise provides consumer-to-consumer, business-to-consumer, and business-to-business sales services via global and domestic marketplaces, alongside logistics, digital media, entertainment, and cloud computing operations. The company's international trade site connects manufacturers, suppliers, exporters, and importers worldwide, supported by trade assurance protections for transactions.
Why it matters
Alibaba launched its first business venture in 1999 before achieving a historic initial public offering on the New York Stock Exchange in September 2014, raising US$25 billion. The corporation expanded into one of the top ten most valuable companies globally, eventually becoming the second Asian firm to exceed a US$500 billion valuation in January 2018. It continues to provide technology infrastructure and marketing reach for merchants and brands globally.
What is confirmed
- Alibaba Group Holding Limited is a Chinese multinational technology company specializing in e-commerce, retail, Internet, and technology.
- The company was founded on June 28, 1999, in Hangzhou, Zhejiang.
- Alibaba's American initial public offering on the New York Stock Exchange raised US$25 billion on September 19, 2014.
- Alibaba became the second Asian company to break the US$500 billion valuation mark in January 2018.
What to watch next
- Further updates regarding Alibaba's cloud computing and artificial intelligence initiatives.
- Developments concerning domestic hardware and export controls.
confidence 100%Sources used for this update (8)
- www.alibaba.com — Alibaba.com: Manufacturers, Suppliers, Exporters & Importers from...
- www.alibabagroup.com — Alibaba.com-Alibaba Group
- en.wikipedia.org — Alibaba Group - Wikipedia
- www.1688.com — 阿里1688首页
- login.alibaba.com — Alibaba Manufacturer Directory - Suppliers, Manufacturers,...
- www.alibabagroup.com — Alibaba Group Official Website
- french.alibaba.com — Alibaba, la plus grande plateforme de commerce B2B en ligne ...
- www.alibaba.com — Alibaba.com
-
China Proposes AI Export Controls as Alibaba Advances Model Roadmap
Beijing has proposed draft export controls to limit the overseas transfer of advanced artificial intelligence models, model weights, and chip design technology, while exempting open-source projects under interim filing requirements. This move coincides with ongoing technological developments by major Chinese artificial intelligence laboratories in 2026, including Alibaba's Qwen, DeepSeek, and Moonshot's Kimi. Alibaba previously announced plans to scale its model lineup to between 5 trillion and 10 trillion parameters alongside a new processor that triples the performance of its predecessor, marking a push for domestic alternatives as firms seek local hardware.
Why it matters
The proposed export restrictions highlight Beijing's tightening grip on domestic artificial intelligence capabilities as local labs scale up their infrastructure. Alibaba continues to position its hardware as the leading domestic option for data center workloads. Software support and manufacturing capacity remain critical factors for widespread hardware adoption.
What is confirmed
- China proposed draft export controls limiting the overseas transfer of advanced artificial intelligence models, model weights, and chip design technology.
- Open-source projects are largely exempt from the proposed Beijing draft under interim filing requirements.
- Alibaba Group previously unveiled a new artificial intelligence processor that triples the performance of its predecessor alongside plans for models with 5 trillion to 10 trillion parameters.
What to watch next
- Finalization and official implementation of China's draft export controls on artificial intelligence and chip design technology.
- Adoption rates of Alibaba's new artificial intelligence processor in data centers.
- Further model size and context length releases from Chinese laboratories including Qwen, DeepSeek, and Kimi.
confidence 100%Sources used for this update (2)
- www.briefs.co — China proposes export controls on AI and chip design
- safeguard.sh — Chinese AI labs in 2026: DeepSeek, Qwen and Kimi releases
-
Alibaba Unveils New AI Chip and Targets 10 Trillion Parameters
Alibaba Group has unveiled a new artificial intelligence processor that triples the performance of its predecessor, alongside plans to scale its AI model lineup to between 5 trillion and 10 trillion parameters. The company announced the new hardware and model roadmap at the Apsara Conference, positioning the processor as China's most powerful chip for artificial intelligence work in data centers. Analysts note that software support and manufacturing capacity will dictate how rapidly the chip sees adoption at scale, as Chinese firms increasingly seek domestic alternatives to imported accelerators.
Why it matters
The hardware rollout aligns with Alibaba's broader strategy to build out custom silicon and expand data center capacity. This approach combines AI model development at Tongyi Lab, cloud services, and proprietary chips from its Pingtouge unit to establish an integrated infrastructure stack. Meanwhile, other nations are aggressively scaling their own capabilities, evidenced by South Korea's separate plans for a 4.7 trillion won frontier AI model programme starting in March 2027.
What is confirmed
- Alibaba Group unveiled a new AI chip that it said triples the performance of its predecessor.
- Alibaba announced plans for an AI model with up to 10 trillion parameters.
- Alibaba describes its new processor as China's most powerful chip for artificial intelligence work.
- Alibaba expects mass production of its AI chips in the coming years.
Still unconfirmed
- Alibaba has not yet given detailed benchmarks or public timelines for the scale and technical details of its data center buildout.
What to watch next
- Detailed benchmarks and technical specifications for Alibaba's new AI chip
- Release dates and licensing details for the Qwen 4 model lineup
- Adoption rates and manufacturing capacity for domestic Chinese AI accelerators
confidence 90%Sources used for this update (13)
- www.daily8news.com — Alibaba unveils what it calls China's most powerful AI chip
- zglg.work — Alibaba Unveils AI Chip and Data Center Plans - zglg.work
- finance.yahoo.com — Chinese AI Chips Could Flood the Market. Alibaba’s New Chip ...
- www.yahoo.com — South Korea plans to develop $3.5 billion frontier AI model starting next year
- edgeconsultancykw.com — Alibaba's New AI Chip Triples Performance as Shares Jump 5.1 ...
- klse.i3investor.com — Alibaba Unveils New Chip Design to Meet Surging Demand for AI
- www.yottalabs.ai — Qwen 4 27B: Release Date, What’s Confirmed, and the Hardware ...
- kie.ai — Qwen4-Max: Deep Dive Into Alibaba's Flagship - kie.ai
- global.chinadaily.com.cn — New chip completes Alibaba's AI 'golden triangle'
- www.straitstimes.com — China plans AI export curbs to exert control over key tech
- kga-it.com — Qwen 4: what's official and what's rumor - kga-it.com
- economictimes.indiatimes.com — qwen: Latest News & Videos, Photos about qwen | The Economic ...
-
Alibaba begins training Qwen4 and targets 10 trillion parameter models
Alibaba has started training its Qwen4 model and plans to scale subsequent versions, Qwen4.5 and Qwen5, to between 5 trillion and 10 trillion parameters. Announced at the 2026 Apsara Conference, this roadmap integrates proprietary accelerators and a target of over 20 gigawatts of cloud capacity by 2032. The strategy shifts focus from raw benchmark competition toward a full-stack infrastructure approach, embedding AI into daily consumer routines like payments and shopping to create a habit-based competitive advantage.
Why it matters
The push follows the reveal of the Zhenwu V900 chip, which offers three times the performance of its predecessor. This infrastructure allows Alibaba to cluster up to 500,000 units to support massive model scaling. The company is positioning its AI as a seamless part of existing user habits rather than a standalone destination.
What is confirmed
- Alibaba confirmed Qwen4 has entered training at the 2026 Apsara Conference.
- The company plans to extend parameters for Qwen4.5 and Qwen5 into the 5 to 10 trillion range.
- Alibaba aims for more than 20 gigawatts of cloud capacity by 2032.
- The company announced its full-stack AI push on September 22.
Still unconfirmed
- Alibaba is pursuing a habit moat strategy by embedding AI into everyday routines to outlast competitors focusing on raw capability.
What to watch next
- Release date and benchmarks for Qwen4
- Deployment of the Zhenwu V900 in large-scale clusters
- Progress updates on the 20 gigawatt cloud capacity goal
confidence 90%Sources used for this update (6)
- www.thejakartapost.com — Indonesia Business, Economy, and Investment News
- finance.yahoo.com — Alibaba Stocks Rise as Ten-Trillion-Parameter Plan Expands AI Bet
- www.aimodeling.com — Qwen4 Enters Training as Alibaba Pushes LLMs to 5–10…
- www.ainchina.com — Alibaba Qwen 4 Apsara 2026: The Real Strategy | AI in China
- economictimes.indiatimes.com — alibaba qwen News and Updates from The Economic Times - Page 1
- www.gurufocus.com — Alibaba Stocks Rise as Ten-Trillion-Parameter Plan Expands AI Bet
-
Alibaba unveils China's most powerful AI chip, plans 10-trillion-parameter model
Alibaba has unveiled the Zhenwu V900 AI chip, which it claims is China's most powerful AI chip, and announced plans to build an advanced AI model with 5 trillion to 10 trillion parameters. The Zhenwu V900 delivers three times the performance of its predecessor and can be clustered with up to 500,000 units. The company also detailed a broad full-stack artificial intelligence strategy that includes scaling cloud infrastructure.
Why it matters
The push forms part of a wider roadmap encompassing domestic processors, cloud services, and autonomous agents. Alibaba's AI ambitions are driven by the growing demand for AI computing power and the need for homegrown silicon amid export controls. The company's shares surged following the announcements, reflecting positive market reaction.
What is confirmed
- Alibaba unveiled the Zhenwu V900 AI chip, which delivers three times the performance of its predecessor.
- The Zhenwu V900 can be clustered with up to 500,000 units.
- Alibaba plans to build an advanced AI model with 5 trillion to 10 trillion parameters.
- The company targets more than 20 gigawatts of global data centre capacity by 2032.
- Alibaba's shares rose 5.1% following the announcements.
What to watch next
- Mass production of the Zhenwu V900 AI chip planned for the first quarter of 2027
- Validation of the economics of scaling chips, models and data centers
- Progress on the Qwen models with 5-10 trillion parameters
confidence 100%Sources used for this update (11)
- www.wionews.com — Alibaba unveiled 'China's most powerful AI chip' and a plan for a 10-trillion-parameter model
- note.com — 【米国株】9/26 AIニュース（AI要約）
- thearabianpost.com — Alibaba debuts Zhenwu V900 with threefold performance gain
- note.com — [Morning Edition] September 23 (Wed) 20 Selected Business & Market News | Nasdaq Hits Consecutive Record Highs, Crude Oil Falls for 5th Day, Takaichi-Tru…
- menastartupdigest.com — Alibaba Plans AI Models of Up to 10 Trillion Parameters ...
- www.aimodeling.com — Alibaba Qwen next gen to scale to 5-10 trillion parameters
- finance.yahoo.com — Alibaba is Betting Bigger on AI as the Race for Scale Intensifies
- W.Media — Alibaba unveils new GPU as it expands AI chip roadmap
- currenttribune.com — Alibaba Wants a 10-Trillion-Parameter AI Model, and It Just ...
- finance.yahoo.com — Alibaba Holds Flat as 10-Trillion-Parameter Ambition Tests Power
- daily8news.com — Alibaba unveils Zhenwu V900 AI chip and plans Qwen models up ...
-
Alibaba Unveils New AI Chip and 10 Trillion Parameter Model Plan
Alibaba unveiled a new artificial intelligence chip and announced plans to build an advanced model featuring 5 trillion to 10 trillion parameters. The company also detailed a broad full-stack artificial intelligence strategy that includes scaling cloud infrastructure. Market reaction was positive, with company shares surging following the announcements. The push forms part of a wider roadmap encompassing domestic processors, cloud services, and autonomous agents.
Why it matters
The sweeping hardware and model announcements arrive ahead of a meeting between Chinese and United States leaders where technological competition is expected to be a central topic. The strategy aims to bolster domestic capabilities across chips, cloud infrastructure, and data centers. However, analysts note the rapid expansion faces significant questions regarding energy demand, regulatory pressures, and overall commercial viability.
What is confirmed
- Alibaba plans to build an AI model with 5 trillion to 10 trillion parameters.
- Alibaba unveiled a new artificial intelligence chip described by the company as China's most powerful.
- Alibaba shares jumped following the unveiling of the new chip and data center buildout plans.
- Alibaba's full-stack strategy spans chips, cloud infrastructure, models, and agents.
Still unconfirmed
- Alibaba is reportedly planning to charge large users for its next open-source artificial intelligence model.
What to watch next
- Implementation progress on data center capacity targets toward 2032
- Discussions on artificial intelligence competition during upcoming meetings between Chinese and United States leaders
- Further commercial rollout and pricing details for the next open-source model
confidence 100%Sources used for this update (14)
- Bloomberg.com — Alibaba Unveils AI Chip to Drive 20GW of Data Centers by 2032
- Reuters — Alibaba plans AI model with 5 trillion to 10 trillion parameters, unveils new chip
- Macau Business — Alibaba Unveils Roadmap on Full-Stack AI Strategy from Chips, Cloud Infrastructure, Models to Agents
- Investing.com India — Why is Alibaba stock surging today? By Investing.com
- careplusvn.com — Alibaba Reportedly Planning to Charge Large Users for Next Open-Source AI Model - Earnings Call Highlights
- CNBC — Alibaba shares jump as new AI chip, data center buildout plans unveiled
- Investing.com — Alibaba unveils new AI chip, calls it China’s most powerful
- digitimes — Alibaba expands full-stack AI push as cloud data centers target 20 GW by 2032
- pulse2.com — Alibaba Unveils New AI Chip And Plans Model With Up To 10 Trillion Parameters
- ua.news — Alibaba unveils Zhenwu V900 AI chip and announces new model — Taipei Times
- japantoday.com — China's Alibaba unveils new powerful chip and ambitious AI model plans
- www.devdiscourse.com — Alibaba’s Trillion-Parameter AI Bet Reshapes China’s Race for Chips, Cloud, Power and Computing
Community Sentiment: How do you assess this situation?
Voice your perspective · Real-time aggregated sentiment from the Live Feeds community