China’s LineShine Blasts Past Exaflop Barrier with Homegrown Tech

Get ready to witness a seismic shift in supercomputing power! China’s LineShine has shattered records and taken the crown as the world’s fastest supercomputer. This beast just crossed the critical two-exaflop threshold—something no CPU-only system has done before.

LineShine’s Unstoppable Performance Leap

On June 23, 2026, the TOP500 list revealed a stunning new leader. LineShine from China’s National Supercomputing Center in Shenzhen hit a blazing 2.198 exaflops on the High Performance Linpack (HPL) benchmark. That’s a massive leap over the previous champ, the US-based El Capitan, which scored 1.809 exaflops.

This is not just a new number—it’s a milestone. LineShine is the first supercomputer to break the two-exaflop barrier with sustained double-precision performance using CPUs only. No GPUs, no accelerators—just raw, massive CPU power.

How does it pull this off? With 13.79 million cores ticking at 1.55GHz. These cores link up through a proprietary interconnect built on LingQi technology, designed in China. The system runs on the Kylin operating system, another piece of homegrown tech powering this monster.

China’s Homegrown Hardware and Efficiency

LineShine’s heart beats with Chinese-designed chips. It uses LingKun processors, a powerful CPU line developed domestically. This is a major step for China’s semiconductor independence and technological prowess.

Power hungry? LineShine draws about 42.2 megawatts—enough to power a small town. But it packs a punch in efficiency, achieving 52.07 gigaflops per watt. That’s impressive muscle with a smart energy footprint.

Its theoretical peak performance is 2.736 exaflops. On HPL it hits about 80 percent of that peak, a solid real-world figure. But LineShine isn’t just about raw speed.

• On the HPCG benchmark, which simulates AI-like workloads, it ranks fourth with 22.00 HPCG-petaflops.
• On the mixed-precision HPL-MxP test, it reaches an eye-popping 7.92 exaflops, showing strength in AI-adjacent tasks.

These benchmarks highlight that LineShine is versatile, not just a one-trick speedster.

Global Supercomputing Landscape Shifts

The June 2026 TOP500 list shows a more diverse and interesting field than ever. Five systems now exceed one exaflop on HPL:

• LineShine at 2.198 exaflops
• El Capitan at 1.809 exaflops (Lawrence Livermore National Laboratory, USA)
• Frontier at 1.353 exaflops (Oak Ridge National Laboratory, USA)
• Aurora at 1.012 exaflops (Argonne National Laboratory, USA)
• JUPITER Booster at exactly 1.000 exaflop (Jülich Supercomputing Centre, Europe)

These top ten systems represent cutting-edge architectures from China, the US, Europe, and Japan. They run on a mix of Intel, AMD, NVIDIA, and custom processors. HPE/Cray supplied six of the top ten machines. Four use AMD processors, and three integrate NVIDIA technology.

Dr. Jack Dongarra, a leading expert in supercomputing, called LineShine “an impressive system.” That’s high praise from a pioneer who knows the field inside out.

But experts caution that TOP500 rankings don’t directly translate to leadership in AI-optimized computing. AI workloads demand different strengths than pure HPL benchmarks. Still, LineShine’s mixed-precision and HPCG results hint at serious AI potential alongside raw speed.

Looking Ahead: What This Means for Tech

LineShine is a game-changer. It proves China can design and build world-class supercomputers from the ground up. No reliance on foreign CPUs or operating systems—just pure homegrown tech muscle.

This breakthrough raises the stakes in the global race for high-performance computing. It pushes innovation in chips, interconnects, and software. It also expands the supercomputing landscape beyond traditional players.

The supercomputer battle is heating up. Will the US respond with new machines? Can Europe and Japan close the gap? The next TOP500 update will be thrilling.

One thing is clear: we’re witnessing the dawn of a new era in supercomputing power. And it’s only getting started.