US chip curbs ускорate china’s AI hardware push

US semiconductor export controls have paradoxically accelerated China's domestic chip manufacturing ambitions, with government-backed industrial policies playing a pivotal role. Companies like Huawei, rather than being held back, have doubled down on investments in domestic alternatives, spurred by Beijing's strategic financing and industrial models inspired by broader Asian policy frameworks. This concerted push has not only fostered rapid technological advancement but also strengthened China's resolve to achieve technological sovereignty amid external pressures.

Despite a 72.3 percent drop in direct semiconductor imports following the 2018 Section 301 tariffs, China’s technology competitiveness has surged, with the AI performance gap against the US shrinking dramatically from an estimated 6.3 years to near parity by 2026. This growth is mirrored in semiconductor-dependent sectors such as industrial robotics and new-energy vehicles, which have expanded over fivefold and twelvefold respectively, underscoring how China’s vast domestic market and robust industrial base have absorbed and adapted to US technological coercion.

China’s semiconductor industry is rapidly advancing in specialized segments like security cameras and satellite imaging, where some CMOS image sensor products now rank among the world’s best, and semiconductor exports have hit record highs. However, the primary bottleneck remains access to EUV lithography equipment, unavailable domestically, which limits progress at leading-edge nodes such as 5nm and 3nm. In response, Chinese firms like Huawei are innovating through alternative architectures—such as 3D integration and heterogeneous integration—to boost chip performance without relying solely on transistor scaling or EUV technology.

Looking ahead, China is expected to develop its own EUV lithography solution or a functional alternative within five years, a timeline accelerated by current export restrictions that have galvanized domestic innovation. This anticipated breakthrough signals a strategic pivot from dependency to self-reliance, highlighting how US sanctions have inadvertently catalyzed a new era of technological ingenuity within China’s semiconductor sector.

In response to escalating US semiconductor export controls, Chinese startups have sharply pivoted towards resolving critical supply chain chokepoints, particularly within the semiconductor sector. By 2026, approximately 20% of companies launching IPOs on Shanghai’s Star Market were focused on these technological bottlenecks, a significant rise from just 8.1% in 2022, underscoring a strategic realignment towards domestic innovation in core manufacturing tools and raw materials.

This surge in chokepoint-focused startups reflects a broader market transformation aimed at achieving supply chain self-sufficiency, with about 60% of firms contributing to this goal in 2026 compared to 41% four years earlier. The concentration of 19 out of 21 identified chokepoint companies within the semiconductor supply chain highlights a deliberate narrowing of focus, targeting upstream vulnerabilities such as specialized parts and complex machinery to reduce reliance on foreign technology.

The maturation of these domestic startups has reached a pivotal stage where many are now ready to tap capital markets to scale their operations, signaling growing investor confidence in China’s homegrown solutions to supply chain challenges. As Morgan Stanley analysts observed, the expansion of chokepoint technologies over the past several years has cultivated a cohort of companies poised for growth, effectively turning US export restrictions into a catalyst for China’s indigenous semiconductor and hardware development.

Chinese firms have ingeniously exploited legal and structural loopholes in US semiconductor export controls to maintain access to advanced AI hardware. Inspur Group, after being blacklisted, rebranded and leveraged its US-based affiliate Aivres Systems—now a sister company with only a 33% stake held by Inspur—to legally sell Nvidia’s cutting-edge GB200 and Blackwell chips to Southeast Asian telecoms like Indosat, enabling Chinese startups such as INF Tech to rent compute power abroad without importing GPUs directly into China. This strategic corporate restructuring, combined with the US Commerce Department’s focus on physical hardware movement rather than remote compute access, has allowed China to circumvent restrictions while technically complying with export rules, as long as the compute is not used for prohibited military applications.

While US export controls have slowed China’s AI hardware progress, Beijing is doubling down on technological self-reliance and alternative business models to sustain its semiconductor ecosystem. China’s latest five-year plan emphasizes accelerated scientific breakthroughs and AI integration across industries, with state-directed investments fostering domestic manufacturing clusters such as the MLCC production hubs along the Guangdong-Fujian coast. Companies like Jialichung are innovating with small-batch PCB production, defying traditional manufacturing economics, while Tencent is ambitiously developing an indigenous agentic AI hardware ecosystem, reflecting a strategic pivot from dependence on US technology toward homegrown solutions.

The evolving US-China AI rivalry has prompted the US to introduce nuanced export policies, such as case-by-case licensing for advanced AI chips like Nvidia’s H200 and AMD’s MI325X to approved Chinese customers, signaling recognition of the complex enforcement landscape. Meanwhile, Chinese firms continue to exploit cloud compute rentals and smuggling networks—evidenced by US prosecutions of Super Micro’s co-founder for diverting $2.5 billion worth of Nvidia servers through Southeast Asian fronts—highlighting the challenges in fully containing China’s AI hardware access. This dynamic has led to a hybrid AI hardware ecosystem in China that balances reliance on imported US technology with a gradual shift toward domestic chip alternatives as performance gaps narrow.

Facing limited compute capacity due to export restrictions, Chinese AI companies are adopting innovative business models such as releasing open weights of their AI models to circumvent deployment constraints, as seen with Moonshot AI suspending new subscriptions after demand overwhelmed its infrastructure. Concurrently, the surge in IPOs among semiconductor and AI accelerator companies signals robust domestic market confidence despite external pressures, illustrating how China’s AI hardware ecosystem is adapting not only technologically but also financially to sustain growth and competitiveness amid ongoing US restrictions.