무슨 일이 일어났나요?
According to the South China Morning Post, several Chinese pure-play AI chipmakers, including Moore Threads and MetaX, reported their first profitable quarters in the third quarter. Moore Threads posted an adjusted net profit of 49.26 million yuan, driven by a 1,774% year-on-year revenue surge, while MetaX recorded a net profit of 26.5 million yuan. These gains occurred as domestic firms raced to replace Nvidia hardware restricted by US export controls. However, the report highlights severe headwinds, including a critical shortage of high-bandwidth memory (HBM) that has driven grey-market prices for 36GB HBM3e modules from approximately $500 to over $4,000. This cost inflation has compressed gross margins for Moore Threads and poses significant risks for MetaX, which faces limited scale among top-tier cloud service providers. Additionally, Cambricon Technologies reported a 502 million yuan inventory impairment loss due to supply chain pressures, despite securing testing contracts with major cloud providers. Analysts from Morgan Stanley, Goldman Sachs, and Nomura provided mixed outlooks, noting that while AI demand is strong, profitability depends on resolving local fabrication and memory supply issues.
The South China Morning Post reports that Chinese AI chipmakers Moore Threads and MetaX achieved their first profitable quarters in Q3, driven by a surge in domestic demand for alternatives to Nvidia hardware. Moore Threads recorded an adjusted net profit of 49.26 million yuan, supported by a 1,774% year-on-year revenue increase, while MetaX posted a net profit of 26.5 million yuan, its second profitable quarter of the year.
Despite these earnings, the sector faces significant supply chain challenges, particularly an acute shortage of high-bandwidth memory (HBM). Morgan Stanley notes that factory prices for 36GB HBM3e modules have risen from around $500 to over $4,000 on grey markets due to US export controls reducing supply. This cost surge has compressed Moore Threads' gross margin by more than 8 percentage points year-on-year as it stockpiled expensive components.
MetaX faces additional hurdles due to limited scale among tier-one cloud service providers, making it difficult to absorb cost spikes despite raising prices for its flagship C600 processor. Morgan Stanley has lowered its 2027 and 2028 gross margin forecasts for MetaX, citing insufficient evidence of broader hyperscaler adoption beyond a 20,000-unit order from JD.com. The company's next-generation C700 processor may also face delays until late 2027.
Cambricon Technologies, another major player, reported a 502 million yuan inventory impairment loss over the first nine months due to a 3.42 billion yuan quarter-on-quarter surge in inventory. However, the company has made progress in mainstream cloud infrastructure, with its MLU692 chip delivered to a major CSP for testing in Q3. Morgan Stanley projects Cambricon will ship nearly 350,000 units to CSP clients next year alongside its next-generation MLU580.
Analysts from Goldman Sachs and Nomura offer differing perspectives on the sector's future. Goldman Sachs emphasizes that long-term survival depends on sustaining profitability while improving local fabrication capacity and yields. Nomura argues that financial markets underestimate the breadth of AI demand and the earnings potential across the value chain, though sustained cuts to computing commitments could challenge this outlook.
왜 중요한가요?
This development marks a pivotal shift in the global AI hardware landscape, demonstrating that Chinese domestic chipmakers are achieving commercial viability in the absence of Nvidia's high-end processors. The transition from loss-making to profitability for key players like Moore Threads and MetaX suggests that the Chinese market is successfully absorbing local alternatives, reducing reliance on US technology. However, the severe cost inflation in HBM memory and the resulting margin compression indicate that this profitability is fragile and heavily dependent on supply chain stability. The situation underscores the economic impact of US export controls, which have inadvertently accelerated the maturation of China's domestic AI hardware ecosystem while simultaneously driving up costs for all participants. For the broader AI industry, this signals a bifurcated hardware market where performance and cost-efficiency are increasingly determined by regional supply chain resilience rather than just architectural innovation. The ability of these firms to sustain profits while navigating memory shortages will determine whether they can compete globally or remain confined to the domestic Chinese market.
The profitability of Chinese AI chipmakers signals a maturing domestic ecosystem that is successfully filling the market void left by US export controls on Nvidia hardware. This shift reduces China's reliance on foreign AI chips and strengthens the local supply chain, which is a strategic priority for Beijing.
The severe cost inflation in HBM memory highlights the fragility of these early profits. The jump in HBM prices from $500 to over $4,000 demonstrates how supply chain bottlenecks can erode margins, posing a significant risk to the long-term viability of these companies if not resolved through local production or alternative sourcing.
The limited adoption by top-tier cloud service providers, as noted by Morgan Stanley, suggests that while government and regional projects are driving initial demand, broader enterprise adoption remains uncertain. This could limit the scale and profitability of these chipmakers in the near term, keeping them dependent on state-backed initiatives.
The competition among domestic players like Moore Threads, MetaX, and Cambricon is intensifying, with each firm vying for market share in a constrained environment. This rivalry may drive innovation and cost reductions but also risks overcapacity and price wars if demand does not grow as expected.
The situation underscores the global impact of US export controls, which have accelerated the development of China's domestic AI hardware sector while simultaneously driving up costs for all participants. This bifurcation of the AI hardware market could lead to divergent technological trajectories and competitive dynamics in the years ahead.
대화형 메커니즘: 실제로 작동하는 방식
이 개발의 이면에 있는 기본 기술을 대화식으로 살펴보세요.
Which component of an AI application is the machine-learning model itself?
다음에 무엇을 볼 것인가
Investors and industry observers should monitor the resolution of the HBM memory shortage and its impact on the gross margins of Moore Threads, MetaX, and Cambricon in the coming quarters. The timeline for MetaX's next-generation C700 processor, potentially delayed to late 2027, is critical for assessing its ability to capture enterprise accounts. Additionally, the extent of adoption by major Chinese cloud service providers (CSPs) beyond government projects will indicate whether these chipmakers can achieve the scale necessary for long-term profitability. Finally, the evolution of local fabrication capacity and production yields in China will be key to determining if domestic chipmakers can sustain their competitive edge against potential future easing of export controls or new technological breakthroughs.
The resolution of the HBM memory shortage and its impact on the gross margins of Moore Threads, MetaX, and Cambricon in upcoming quarters will be a key indicator of their long-term profitability and sustainability.
The timeline for MetaX's next-generation C700 processor, potentially delayed to late 2027, is critical for assessing its ability to capture enterprise accounts and maintain its competitive position against other domestic chipmakers.
The extent of adoption by major Chinese cloud service providers (CSPs) beyond government projects will indicate whether these chipmakers can achieve the scale necessary for long-term profitability and market dominance.
The evolution of local fabrication capacity and production yields in China will be key to determining if domestic chipmakers can sustain their competitive edge against potential future easing of export controls or new technological breakthroughs from US and other international competitors.