A Milestone Week for Independent Computing Power

The week of June 23, 2026, will be remembered as a turning point in China's pursuit of independent computing power. At the ISC 2026 conference in Hamburg, Germany, two historic achievements were announced on back-to-back days:

First, the domestically-developed "Lingcheng" supercomputer topped the global TOP500 list with 2.19 EFlops of sustained double-precision floating-point performance, returning China's supercomputing to the world's number one position for the first time in nine years [citation:2][citation:4].

Then, on June 24, Sugon's ParaStor F9000 all-flash storage system won double first place in the IO500 production rankings — both in the full-node and 10-node categories. This marks the first time a Chinese vendor has achieved this "double crown" in the industry's most authoritative storage benchmark [citation:5][citation:6].

Computing and storage, the two pillars of digital infrastructure, both reached world-leading milestones in the same week — a powerful signal that China's independent computing power ecosystem is not just catching up, but setting new global standards.

"Lingcheng" Tops the Global Supercomputer Rankings

On June 23, 2026, the National Supercomputing Center in Shenzhen released the "Lingcheng" supercomputer at ISC 2026 with a measured sustained performance of 2.19 EFlops — the world's first supercomputing system to exceed 2 exaflops of sustained performance [citation:3][citation:4].

"Lingcheng" displaced the U.S. Department of Energy's El Capitan system (1.809 EFlops) from the top spot, with Oak Ridge National Laboratory's Frontier (1.353 EFlops) now in third place [citation:2]. It is also the fifth exascale-class supercomputer in the world, and the first built entirely with domestic technology [citation:2].

The achievement has drawn international recognition. Jack Dongarra, Turing Award winner and professor at the University of Tennessee, praised "Lingcheng" as showing "a glimmer of hope for AI4Science system architecture" [citation:10][citation:11]. Forbes noted that "Lingcheng" is not GPU-driven but runs entirely on standard CPUs — a design choice that may offer a better path for integrating AI with traditional scientific computing [citation:1].

The Architecture Behind "Lingcheng": Why It's Different

What makes "Lingcheng" truly remarkable is not just its performance, but its architecture. Unlike most of today's leading supercomputers, which rely on GPU accelerators from NVIDIA or AMD, "Lingcheng" takes a different path.

"Lingcheng's" chief architect, Professor Lu Yutong of Sun Yat-sen University, explained that supercomputing has entered a new era of "super-intelligence convergence." The system's Online Acceleration architecture breaks down the traditional CPU-GPU barrier, embedding AI acceleration directly into the CPU to enable efficient collaboration between supercomputing and AI workloads [citation:2][citation:9].

This architectural choice is significant for two reasons. First, it demonstrates that a non-GPU path to exascale computing is viable — particularly relevant given ongoing export restrictions on advanced GPUs. Second, by integrating AI acceleration at the chip level, "Lingcheng" positions itself as a platform for "science AI" (AI4Science), supporting applications ranging from drug discovery and materials science to climate modeling and brain research [citation:2][citation:9].

Storage Double Crown: ParaStor F9000 at the Top of IO500

On the following day, June 24, Chinese storage technology took its own place in the spotlight. Sugon's ParaStor F9000 all-flash storage system took first place in both the full-node and 10-node production categories of the IO500 benchmark [citation:5][citation:7].

The IO500 is the storage performance counterpart to TOP500, evaluating the I/O capabilities of high-performance storage systems under real workloads. What makes Sugon's achievement particularly significant is the category: Production List rather than Research List [citation:5][citation:6].

The Production List only includes storage systems that have been running in actual production environments for extended periods, handling real business workloads with redundancy and continuous operation [citation:5][citation:6]. In other words, Sugon didn't build a system just to win a benchmark — it proved the system's capabilities through real deployments, and then the benchmark confirmed what the production environment had already shown.

ParaStor F9000 has been running in a tens-of-thousands-card cluster for over a year, supporting more than 100 AI and high-performance computing production applications, including large model training, scientific computing, and data analytics [citation:6][citation:7].

Production Validation: From Benchmarks to Real Business

Sugon's storage achievement is backed by real-world deployments across multiple frontier industries:

These aren't demonstration projects — they are critical production workloads that run 24/7 in environments where failure means millions in lost productivity. This kind of real-world validation carries more weight than any benchmark alone.

Key Takeaways