Moffett AI Showcases Specialized Sparse Inference Chips at 2026 Inclusion·Bund Conference
Moffett AI appeared as an Ant Group ecosystem partner at the 2026 Inclusion·Bund Conference and presented AI compute cards, outlining a path to improve compute efficiency with specialized sparse inference chips as inference demand rises.
The company's appearance came as AI technology achievements are accelerating into products and services for digital and physical worlds, driving the formation of the AI new economy and sustained growth in compute demand, especially on the inference side. Improving compute efficiency allows more inference tasks to be completed with the same computing resources, providing support for the scaled development of AI products and services and further releasing the development potential of the AI new economy.
Facing rising inference demand, Moffett AI says increasing compute supply must be accompanied by innovation in computing methods to improve compute efficiency. As a leader in sparse computing, Moffett AI uses coordinated optimization across algorithms, software and hardware to reduce redundant computation and improve computational efficiency through specialized sparse inference chips, enabling limited compute resources to complete more effective computation and offering a more efficient computing method for the AI new economy.
Moffett AI's inference clusters, built on its self-developed chips, have been deployed in data centers across four major regions: Northwest, Southwest, East China and North China. Combined with local industrial and resource characteristics, the clusters provide efficient inference support for scenarios such as intelligent manufacturing, life sciences and urban governance, promoting faster integration of compute and industrial demand and jointly expanding the application space of the AI new economy.
As AI applications deepen, inference workloads are becoming more diverse and dynamic, placing new demands on chips' data processing and scheduling capabilities. To address this, Moffett AI is about to launch its next-generation SparsePrime® compute card. Through architecture upgrades and optimization of scheduling mechanisms, the card is intended to further meet evolving AI inference needs.
Moving from technological innovation to ecosystem building, Moffett AI says that making more efficient computing methods truly serve the AI new economy requires not only continued technology and product innovation but also lowering the barriers to technology adaptation and industrial collaboration. While advancing the scaled deployment of specialized sparse inference chips, Moffett AI is further participating in industry standard development and promoting industry-university-research cooperation to build a prosperous AI compute ecosystem with all parties.
In July 2026, Moffett AI became a member unit of the Artificial Intelligence Standardization Technical Committee of the Ministry of Industry and Information Technology. Relying on its technical accumulation and industrialization practice in sparse computing, it will participate in standard development for AI basic hardware, software platforms and large models, promoting more efficient coordination among chips, software, models and applications and lowering adaptation costs across the industrial chain and barriers to scaled AI application.
In August 2026, Moffett AI initiated the establishment of a sparse computing industry-university-research alliance, carrying out collaborative innovation around frontier research, key technology breakthroughs and achievement transformation, and promoting the conversion of scientific research innovation into development momentum for the AI new economy.
As AI products and services accelerate toward scaled application, they continue to release the development potential of the AI new economy. Moffett AI says it will continue to advance innovation in computing methods, using specialized sparse inference chips to improve compute efficiency, allowing limited compute to carry more intelligent applications and accelerating the implementation of AI productivity across industries.
Editor's Summary Moffett AI showcased AI compute cards at the 2026 Inclusion·Bund Conference as an Ant Group ecosystem partner, pitching specialized sparse inference chips as a way to improve compute efficiency amid rising inference demand. The company said its chips-based inference clusters are deployed in four regions and that it will launch its next-generation SparsePrime® compute card. It also joined a MIIT standardization committee and initiated a sparse computing alliance to support ecosystem and standards work.