Chinese Startup Dongfang Suanxin Develops 3D Chip Architecture to Bypass U.S. Export Restrictions

In a significant move to enhance its semiconductor capabilities, Chinese startup Dongfang Suanxin has announced the development of a 3D chip architecture designed to circumvent U.S. export restrictions on advanced chip technology. This initiative comes in response to the tightening of export controls by Washington, which has increasingly targeted advanced semiconductor technologies, particularly those related to artificial intelligence (AI). Company representatives have emphasized that the chip is being developed entirely within China, utilizing domestic supply chains, and that its innovative 3D stacking method promises to boost performance without relying on cutting-edge lithography techniques.

The Context of U.S. Export Controls

The backdrop to this development is the U.S. Department of Commerce's imposition of restrictions in 2022 and 2023 on the export of advanced chips and chip-manufacturing equipment to China. These restrictions specifically aimed at high-performance chips, including Nvidia's A100 and H100 models, along with the essential tools for their production. Concerns have been raised by U.S. officials regarding potential breaches of these controls, particularly involving ASML's extreme ultraviolet lithography machines. Such developments have prompted Chinese chip manufacturers to explore alternative designs and manufacturing processes to reduce their reliance on foreign technology.

ChangXin Memory Technologies, for instance, has garnered attention from major companies like Apple Inc., yet its growth has also made it a target in the ongoing technology rivalry between the U.S. and China. As a result, the pattern of restrictions has inadvertently spurred innovation within the Chinese semiconductor sector, leading to new architectural approaches like that of Dongfang Suanxin.

Dongfang Suanxin's 3D Chip Design

Dongfang Suanxin's innovative approach involves stacking multiple layers of logic and memory vertically, a technique that the company claims enhances performance without the need for advanced lithography processes. This method, referred to as "3D heterogeneous integration," allows for the combination of various types of chips within a single package. According to a technical white paper released by the startup, this architecture can achieve performance levels comparable to Nvidia's latest chips while utilizing older, unrestricted manufacturing nodes. The concept of 3D chip architecture is not entirely new; previous research has demonstrated the potential benefits of stacking layers to improve both performance and spatial resolution.

By employing these methods, Dongfang Suanxin aims to bypass the need for extreme ultraviolet lithography tools, which are currently restricted for sale to China. The startup has asserted that its design can be fully realized with equipment available domestically, marking a significant step toward self-sufficiency in semiconductor manufacturing.

Commercial Viability and Future Plans

Looking ahead, Dongfang Suanxin plans to begin sampling its chip to select customers by the end of 2024. The company has secured funding from a state-backed venture capital fund, although the specifics of this financial backing remain undisclosed. Traditionally, semiconductor startups have operated as fabless design houses, outsourcing their manufacturing to contract chipmakers. This model has enabled new entrants to compete in niche markets effectively.

However, analysts have pointed out that while the 3D design may help bridge some performance gaps, compatibility with existing software ecosystems poses a significant challenge. Efficiently running AI models on the new chip will heavily depend on software optimization, which is critical for performance. To address this, Dongfang Suanxin is developing its own compiler and driver stack tailored to support common AI workloads.

Despite these advancements, experts caution that Nvidia's dominant CUDA software platform provides the company with a substantial advantage in AI applications, regardless of the hardware capabilities of competitors. Nvidia's market position has been bolstered by the surging demand for AI infrastructure, with reported quarterly revenues reaching $57.01 billion in the third quarter of 2025, reflecting a 62% year-over-year increase. Nvidia's CEO has also indicated that the potential for China to surpass the U.S. in the AI race could stem more from energy costs and regulatory factors than from hardware advancements alone.

The U.S. government has signaled that it may extend export controls to encompass 3D stacking equipment if deemed a national security risk. This evolving landscape underscores the high stakes involved in the technology competition between the two nations.

Why it matters

The emergence of Dongfang Suanxin's 3D chip architecture is emblematic of a broader trend among Chinese tech firms striving for self-sufficiency in semiconductor technology amid escalating geopolitical tensions. As China restricts overseas travel for top AI specialists to mitigate brain drain, the implications of these developments extend beyond individual companies. They reflect a concerted effort to innovate under the constraints of export restrictions, which could ultimately reshape the semiconductor landscape in the coming years.

The success of Dongfang Suanxin's chip will depend on overcoming both technical challenges and geopolitical hurdles. As the company forges ahead with its development, its progress will serve as a critical test case for China's ability to innovate independently of foreign technology. Other Chinese firms, such as DeepSeek, are similarly pursuing vertical integration strategies to lessen their reliance on established players like Nvidia and Huawei, indicating a collective push toward indigenization in the semiconductor sector.