Cloud-led Future · Intelligence-driven Hotline +86 21 6228 0217 中文 | EN

Domestic Supercomputer Wins Gordon Bell Prize at SC26

Overview

At the 2026 International Conference for High Performance Computing, Networking, Storage and Analysis (SC26) on November 19 local time, a quantum chemistry computation of the iron-molybdenum cofactor running on the domestically produced 100,000-card AI cluster “Sugon 8000 (Dengfeng)” won the ACM Gordon Bell Prize — often called the Nobel Prize of high-performance computing and one of the highest honors in computational science. A separate SD3 global kilometer-scale weather simulation system was a finalist for the climate modeling award.

The Gordon Bell Prize recognizes breakthroughs in computational science achieved on real, running supercomputers rather than peak-speed rankings. The win suggests China’s domestic computing power is not only at the global forefront in hardware scale, but is gaining international recognition for the software ecosystem and application capability that support frontier research.

Background & Interpretation

1. Technical context and evolution

Established in 1987 and awarded by the ACM, the Gordon Bell Prize honors outstanding achievement in parallel computing applications, with winning work required to run on real supercomputers and pass rigorous review. Past winners have focused on major scientific problems such as climate simulation, astrophysics, materials science and biomedicine. Chinese teams have repeatedly ranked high on the TOP500 list, but breakthroughs in this application-oriented top award — which represents “using computing well” — have been limited, a gap this win helps fill.

2. Core drivers and mechanisms

The winning iron-molybdenum cofactor project focuses on electronic-structure simulation of complex systems such as nitrogenase, which demand extreme precision and scale and underpin materials and catalysis science. Supporting it is the system capability of the 100,000-card domestic cluster: large-scale parallel efficiency, stable high-speed interconnects, the floating-point and mixed-precision performance of domestic chips, and supporting compilers, schedulers and algorithm libraries. The SD3 finalist validates domestic computing strength in climate modeling, a traditional supercomputing stronghold.

3. Market structure and value-chain impact

The breakthrough comes as the domestic computing ecosystem matures rapidly. Alibaba recently disclosed that Qwen is training its next-generation models on Pingtouge chips, with Bailian providing inference on domestic silicon; Huawei’s CloudRobo embodied world model was trained on Ascend computing and open-sourced; collaboration between domestic models and in-house chips continues to deepen. From “can we build it” to “can we use it well,” domestic computing is shifting its value center — software stacks, scheduling efficiency and scenario adaptation are becoming the next competitive frontier.

Implications & Outlook

In the near term, the award should raise the credibility of domestic supercomputing in the international scientific community and attract more research collaborations and application migrations. Over the medium term, AI for Science is becoming a new growth engine for computing demand, with needs in quantum chemistry, meteorology and life sciences driving continued iteration in domestic chips, interconnects and foundational software. A single award does not equal ecosystem-wide leadership, however; software richness, developer scale and cross-platform compatibility remain long-term challenges.

Takeaways for Industry Participants

Computing vendors should shift focus from peak metrics to real-application parallel efficiency and stability, accumulating reusable software stacks and algorithm libraries around key scientific scenarios. Research institutions can leverage the cost and controllability of domestic computing to migrate critical workloads faster, using the process to improve toolchains. Across the value chain, coordinated optimization of chips, interconnects, storage and foundational software matters more than performance gains in any single link.

This column compiles industry information and shares technical perspectives; it does not constitute investment advice.