On July 27, 2025, the "Intelligent Computing Technology Innovation and Industry Practice" forum hosted by the Intelligent Computing Industry Development Committee of the Global Computing Consortium (GCC) successfully concluded. Several groundbreaking achievement were released at the forum, including the "White Paper on Heterogeneous Computing Power Collaboration" and the "White Paper on Vector Databases", injecting strong momentum into the development of the intelligent computing industry.

The "White Paper on Heterogeneous Computing Power Collaboration" was led by the Heterogeneous Computing Power Collaboration Working Group of the GCC Intelligent Computing Industry Development Committee, with China Telecom as the leading unit, and China Academy of Information and Communications Technology and Zhiyuan Research Institute as the deputy leading units. It was jointly compiled by 8 partners in the industrial chain.

The "White Paper on Heterogeneous Computing Power Collaboration" was led by the Heterogeneous Computing Power Collaboration Working Group of the GCC Intelligent Computing Industry Development Committee, with China Telecom as the leading unit, and China Academy of Information and Communications Technology and Zhiyuan Research Institute as the deputy leading units. It was jointly compiled by 8 partners in the industrial chain.

This white paper focuses on intelligent computing heterogeneous computing power, aiming to identify the bottlenecks of heterogeneous computing power collaboration, build an ecological technology system, realize interconnection and efficient collaboration between heterogeneous computing powers, continuously drive the collaborative development of the intelligent computing industry, and work with industry partners to promote technological innovation and enabling applications of heterogeneous computing power.

The "White Paper on Heterogeneous Computing Power Collaboration" addresses three core challenges in heterogeneous computing power collaboration: "computing power resource wall", "software stack ecological fragmentation" and "low collaborative scheduling efficiency". It details the core content of heterogeneous computing power in four dimensions: collaborative system construction, key technological breakthroughs, application practice solutions and future development trends. In terms of collaborative system construction, it focuses on the heterogeneous computing power collaborative system of unified computing, unified communication, unified scheduling and unified evaluation, realizing seamless computing between heterogeneous computing devices, unimpeded communication collaboration, efficient resource scheduling and automated evaluation; in the field of key technologies, it tackles core technologies such as cross-architecture compilation, unified operator acceleration library, unified collective communication library, intelligent computing network interconnection optimization, heterogeneous interconnection bus, distributed training framework, distributed reasoning system and cross-domain heterogeneous computing power scheduling; at the solution practice level, it forms "one model, multiple chips" heterogeneous mixed pool training and low-cost heterogeneous mixed reasoning solutions and benchmark cases; in terms of future development trends, it gradually realizes the reconstruction of computing paradigm from chip level to cluster level, and builds a "four-computing integration" collaborative system for the next generation of artificial intelligence.

With the accelerated evolution of new-generation intelligent applications such as large models, heterogeneous computing power has become an important foundation driving innovation in the intelligent era. The release of the "White Paper on Heterogeneous Computing Power Collaboration" provides strong support for accelerating the construction of a new type of infrastructure for heterogeneous computing power with multiple integration and efficient collaboration. It will help consolidate industry consensus, promote key technological breakthroughs and collaborative ecological construction, and inject new momentum into the sustainable development of China's digital economy and artificial intelligence industry.

Welcome to join the GCC Intelligent Computing Industry Development Committee!