On July 28, 2025, leading Chinese artificial intelligence (AI) firms announced the formation of two new industry alliances aimed at fostering a domestic AI ecosystem and reducing reliance on foreign technologies. These initiatives come in direct response to escalating U.S. export restrictions on advanced semiconductor technologies, particularly those involving Nvidia’s high-performance chipsets. The announcements were strategically made during the three-day World Artificial Intelligence Conference (WAIC) in Shanghai, which concluded on Monday.

The first initiative, titled the Model-Chip Ecosystem Innovation Alliance, seeks to create a comprehensive technological framework that connects large language model (LLM) developers with AI chip manufacturers. This collaboration aims to integrate the full value chain—from semiconductors to AI models and infrastructure—within China. Zhao Lidong, Chief Executive Officer of Enflame, one of the participating chip manufacturers, described the alliance as an “innovative ecosystem” designed to foster synergy across the AI supply chain.

Notable participants in this alliance include major Chinese GPU developers such as Huawei, Biren, and Moore Threads, all of which have been subjected to U.S. sanctions restricting access to cutting-edge technologies developed using American expertise. The initiative was spearheaded by StepFun, a prominent LLM developer, emphasizing a shared commitment among AI firms to innovate domestically and circumvent geopolitical restrictions.

The second initiative, known as the Shanghai General Chamber of Commerce AI Committee, has been established to advance the integration of AI technologies with broader industrial transformation efforts. This alliance also includes StepFun, as well as other key players such as MiniMax, another LLM developer, and chipmakers Metax and Iluvatar CoreX. SenseTime, a major AI firm previously focused on facial recognition technology and currently sanctioned by the United States, is also a significant participant. The company has since pivoted towards LLM development as part of its strategic shift.

A major highlight of WAIC 2025 was the unveiling of Huawei’s CloudMatrix 384, an AI computing system composed of 384 of its latest 910C chips. According to analysis from U.S.-based research firm SemiAnalysis, the system outperforms Nvidia’s flagship GB200 NVL72 chipset in certain performance metrics. The research highlighted Huawei’s strength in system-level engineering, enabling it to utilize higher chip volumes and advanced clustering designs to offset individual chip limitations.

At least six other Chinese companies showcased similar technological solutions, particularly in clustering and supercomputing applications. Metax, for instance, presented an AI supernode system that integrates 128 C550 chips and is engineered to support advanced liquid-cooled data center operations.

Other notable product launches at the conference included Tencent’s Hunyuan3D World Model 1.0, an open-source tool that enables users to generate interactive 3D environments through text or image inputs. Baidu introduced its next-generation "digital human" technology, capable of creating virtual livestream hosts using advanced cloning technologies that replicate voice, tone, and body language from only ten minutes of footage.

Additionally, Alibaba revealed its Quark AI Glasses, powered by the company's Qwen AI model. Scheduled for release in China by the end of 2025, the glasses will allow users to navigate using Alibaba’s mapping services and conduct mobile payments via voice-activated QR code scanning through Alipay.