China has unveiled its first domestically developed near-memory computing 3D AI chip, marking another step in the country’s push to strengthen its semiconductor independence and expand the computing infrastructure needed for artificial intelligence.
The chip, known as DF1000, was launched on July 13 by Shanghai-based AI chip startup Dongfang Suanxin. Its architecture uses near-memory computing and 3D integration, an approach designed to address some of the performance bottlenecks created when large amounts of data must constantly move between processors and memory.
The launch comes as access to advanced AI chips and semiconductor manufacturing technologies has become increasingly important to global technological competition.
DF1000 takes a different approach to AI computing
Modern AI workloads require enormous amounts of data to move rapidly between computing and memory components. This movement can create bottlenecks that affect performance and energy efficiency, particularly as AI models become larger and more computationally demanding.
Near-memory computing attempts to address this problem by bringing computation closer to where data is stored.
According to Dongfang Suanxin, the DF1000 combines this architecture with a software-defined 3D chip design, potentially allowing greater computing performance without depending entirely on the most advanced semiconductor manufacturing nodes.
Company chairman and CEO Wei Shaojun described the launch as an important step toward reducing one of the major constraints facing high-end computing chips: their dependence on increasingly sophisticated manufacturing processes.
The company says its underlying technology is based on more than two decades of research and is supported by a broader effort to develop an independent hardware supply chain and software ecosystem.
China pushes for greater semiconductor self-reliance
The DF1000 arrives at a strategically important moment for China’s technology sector.
Artificial intelligence development increasingly depends on access to high-performance computing infrastructure, particularly GPUs and specialized AI accelerators. At the same time, restrictions surrounding access to advanced semiconductor technologies have increased pressure on Chinese companies to develop domestic alternatives.
Rather than competing exclusively through smaller manufacturing nodes, alternative chip architectures such as 3D integration, chiplets, and near-memory computing could provide another path toward improving AI performance.
If successfully commercialized at scale, these approaches could help Chinese semiconductor companies reduce some of their dependence on the world’s most advanced fabrication technologies.
Shanghai emerges as a major chip innovation hub
The launch also reflects the rapid expansion of Shanghai’s semiconductor ecosystem, particularly in the Pudong New Area.
According to local officials, Pudong’s integrated circuit industry grew approximately 23% year over year in 2025, with industrial output exceeding 360 billion yuan, or roughly $53 billion.
The district reportedly represents around one-fifth of China’s integrated circuit industry and a significant majority of Shanghai’s semiconductor activity.
Shanghai more broadly has developed a substantial semiconductor ecosystem spanning chip design, manufacturing, equipment, materials, research, and talent. The city is home to more than 1,200 integrated circuit companies and has become one of China’s most important centers for semiconductor innovation.
Dongfang Suanxin plans faster AI chip development
Founded in 2024, Dongfang Suanxin is still a relatively young semiconductor company, but it has outlined an aggressive development strategy.
The company says it plans to simultaneously mass-produce its current generation of chips, develop its next generation, and conduct research into future architectures.
Its current products are primarily focused on AI training workloads, while next-generation chips are expected to target performance comparable with mainstream international competitors.
Dongfang Suanxin also plans to showcase a broader portfolio spanning chips and computing systems, signaling ambitions beyond a single processor architecture.
AI chip competition is expanding beyond manufacturing nodes
The significance of the DF1000 extends beyond the launch of another domestic Chinese AI chip.
For years, semiconductor competition has largely centered on who can manufacture smaller, more advanced process nodes. But as those processes become increasingly expensive and geopolitically constrained, chipmakers are exploring architectural innovations to extract more performance from available manufacturing capabilities.
Near-memory computing and 3D integration represent one such approach.
China’s challenge will now be turning these architectural breakthroughs into commercially scalable products capable of competing on performance, energy efficiency, software compatibility, manufacturing yield, and cost.
The DF1000 therefore represents more than a new chip launch. It reflects a broader strategy to find alternative routes to high-performance AI computing as China works to build a more self-reliant semiconductor ecosystem.
