Alibaba Group has unveiled a new AI processor, the Zhenwu V900, which the company says is China's top-performing chip and positions it in direct competition with Nvidia.

Chief Executive Eddie Wu said the new hardware delivers three times the performance of its predecessor, the Zhenwu M890, and can scale to 500,000 units for advanced AI training. Wu also outlined a long-term goal of reaching 20 gigawatts of global data center capacity for Alibaba Cloud by 2032 to keep pace with surging AI demand. The figures point to the physical scale frontier AI development now demands, where training trillion-parameter models rests on chip fleets and power capacity of the kind Alibaba outlined.

The announcement carries weight as China works to build a stronger domestic AI computing ecosystem while access to Nvidia's most advanced processors becomes increasingly restricted. Alibaba has developed its own semiconductor capabilities through its T-Head unit while expanding its cloud infrastructure to support AI workloads.

The Zhenwu V900 therefore represents more than a single new processor. It forms part of Alibaba's broader effort to control more of the computing stack required for large-scale AI development, from chips and cloud infrastructure to foundation models and AI agents. How effectively the company can manufacture and deploy these processors at scale will be an important factor in supporting increasingly demanding AI workloads.

New products unveiled at Apsara Conference

Speaking at Alibaba's Apsara Conference, Wu cited strong customer demand for AI and the company's renewed focus on expanding infrastructure. The firm has committed more than $53 billion to a three-year buildout, backed in part by a $10.2 billion stock offering in Hong Kong in August. The funding provides the financial backbone for targets that stretch years out, spanning chip production, data center construction, and the model roadmap Alibaba laid out alongside the hardware.

Beyond the 20GW plan and the new processor, Alibaba introduced a series of AI products at the conference, including Qwen foundation models, multimodal models, a purpose-built agentic cloud, and a mobile AI agent platform.

Joe Tsai, chairman of Alibaba Group, reinforced the company's AI vision at the event. "AI possesses vast potential for development–it can be deployed and scaled in real-world scenarios, boosting productivity across thousands of industries," he said.

Wu added: "Last year, we asserted that the more capable AI becomes, the more powerful humanity will be. Today, I remain steadfast in that conviction."

The announcements arrived as AI stocks pushed higher globally. Shares of Alibaba climbed more than 5% in Hong Kong trading, while Tencent, the parent company of WeChat, rallied over 7%.

According to a Reuters report, Alibaba is looking to train a large AI model with between 5 trillion and 10 trillion parameters. Annual shipments of its T-Head AI chips are also projected to scale up significantly—a scale of model ambition that aligns with the chip and power targets announced at the conference.

Washington alleges improper distillation

Alibaba has been at odds with the U.S. government for some time. U.S. officials recently leveled new allegations against Alibaba and DeepSeek, claiming the companies improperly obtained knowledge from leading U.S. AI models to advance their own technology. China has dismissed the claims and warned of countermeasures if Washington weaponizes them against Chinese firms.

In a joint statement, the FBI, NSA, and CISA said Chinese tech giants have methodically mined top-tier U.S. AI models for more than two years. The agencies claim proprietary data from ChatGPT, Gemini, Claude, and Grok was used as a blueprint for China's own systems, as detailed in a CISA cybersecurity advisory. The accusations land in a technology relationship already defined by restricted access to Nvidia's most advanced processors and China's parallel push to build domestic alternatives.

The U.S. has a long history of accusing China of stealing corporate secrets. With the latest accusations, the U.S. government noted that Chinese officials were likely aware of the widespread AI scraping but did not explicitly accuse China's intelligence agencies of running the operation.

The White House frequently portrays the U.S. and China as competing for the lead in cutting-edge AI. President Trump is scheduled to meet Xi Jinping on September 24, with AI set to dominate the talks—a point underscored by the inclusion of tech executives such as Nvidia's Jensen Huang and Microsoft's Satya Nadella at an upcoming White House state dinner. The V900 launch, the 20-gigawatt target, and the distillation allegations now form the backdrop to that meeting.

China, for its part, has advocated for greater collaboration between the two countries. "We always believe that all parties should strengthen cooperation to foster open, inclusive, beneficial, and ethical development of AI for the well-being of all," said Mao Ning, spokesperson at the Ministry of Foreign Affairs.