Welcome back to The AI China Report. This week we’re covering China’s homegrown semiconductor industry and Chinese companies getting caught stealing US technology and spreading misinformation campaigns.
QUOTE OF THE WEEK
“Every new technology has its pros and cons. There’s huge upside with AI, and then there are some disadvantages…But by far—orders of magnitude—the higher risk to worry about is China, not sentient AI killing us”
- Vinod Khosla, billionaire investor of Khosla Ventures at the Fortune Brainstorm AI forum last week.
ONE BIG STORY: HUAWEI’S RISE IN THE SEMICONDUCTOR INDUSTRY
The Financial Times reported on the moves Huawei has made in recent years, backed by the Chinese government, to grow its semiconductor development capabilities to compete with the US in order to reduce reliance on other countries to produce powerful AI chips.
In late 2020, Huawei, a major player in the mobile phone industry, faced significant challenges due to sanctions imposed by the Trump administration. These sanctions severed Huawei's access to global semiconductor supply chains, hindering its ability to procure chips for advanced handsets. To navigate these challenges, Huawei formed a strategic partnership with the Semiconductor Manufacturing International Corporation (SMIC), a state-backed Chinese foundry. This collaboration aimed to produce advanced chips using older equipment, a move fraught with uncertainty and high costs.
By August 2023, despite the odds, Huawei launched the Mate 60 series phone, powered by the Kirin 9000S chip. This chip, although not as advanced as its contemporaries from Qualcomm, was well-received in China, marking Huawei's comeback in the chip market after years of sanctions. The US expressed surprise and concern over Huawei's ability to circumvent sanctions, sparking discussions about the implications of this technological breakthrough.
The Kirin 9000S's development relied on less efficient deep ultraviolet (DUV) machines instead of the more advanced extreme ultraviolet (EUV) lithography. This choice was initially driven by budget constraints and later by necessity, as EUV equipment became inaccessible due to export controls. The use of DUV machines increased production steps, costs, and led to a lower yield rate (percentage of non-defective chips), but it was a critical factor in Huawei's ability to continue chip production.
Significant state support and investment in Huawei and SMIC were evident. The Chinese government's financial backing played a crucial role in offsetting the high costs of chip production. This state support is part of China's broader strategy to achieve technological self-sufficiency and reduce reliance on foreign technology, especially given the ongoing geopolitical tensions and trade restrictions.
Looking ahead, Huawei and SMIC are expanding their focus to include chips for artificial intelligence (AI) systems. This shift is in response to U.S. restrictions on high-performance chip sales to China. Huawei's Ascend series of AI chips have been identified as potential alternatives to Nvidia's products, despite lagging in overall performance. Huawei has already started trials of these chips with major Chinese internet companies including Tencent and Baidu.
The success of the Kirin 9000S and the ongoing development of AI chips reflect China's determination to sustain and advance its semiconductor industry. However, challenges remain, particularly in the manufacturing process of data center chips, which are prone to a higher defect rate. Moreover, supply chain restrictions and the potential exhaustion of critical machine parts and materials pose additional hurdles for SMIC's production capabilities.
The journey of Huawei and SMIC, supported by the Chinese state, underscores China's ambition to become a semiconductor superpower. It also highlights the complexities and interdependencies of the global tech industry, where geopolitical dynamics can significantly impact technological progress and market competition.
NEWS ROUNDUP
OpenAI suspended TikTok parent company ByteDance’s account after finding that the company was using data generated by OpenAI to train its own competing model which violated developer licenses of both Microsoft and OpenAI. (The Verge)
A prominent computer scientist and founder of Chinese AI company SenseTime, Tang Xiao’ou, suddenly passed away on Saturday “due to illness”. The specific cause of death has not been revealed. This comes just two weeks after SenseTime experienced stock market turmoil due to short-selling from US-based activist firm Grizzly Research, which claimed the company was inflating revenues. (WIO News)
China's Cyberspace Administration is intensifying its crackdown on short videos that spread misinformation, targeting content that promotes pessimism, extremism, or incorrect values. The campaign, part of the annual "Qing Lang" initiative since 2020, aims to improve mental health and foster a healthy competitive environment in the short video industry, while also cracking down on AI-generated fake videos and content that manipulates public perception. (SCMP)
The Australian Strategic Policy Institute reported a network of over 4,500 YouTube channels spreading pro-China and anti-U.S. narratives, using AI-generated voices and avatars. This campaign, seeking to influence global opinion, promotes Chinese technology and geopolitical stance while criticizing the United States, and has garnered over 120M views and 730K subscribers across 30 different channels on platforms like YouTube. (NYT)
Microsoft is launching a service to combat misinformation and enhance election security, featuring a digital watermark to identify AI-generated content. This service, part of a broader initiative to protect electoral integrity, will provide tools for verifying content provenance and safeguarding against deepfakes, especially in the context of upcoming worldwide elections. (The Verge)
China's autonomous vehicle companies, led by WeRide, are expanding into international markets after domestic growth, with recent licensing achievements in Singapore for testing self-driving buses and robotaxis in the United Arab Emirates. (TechCrunch)
Alibaba's Damo Academy has introduced SeaLLM, an AI large language model tailored for Southeast Asian languages, marking the company's focus on expanding in this region. This model, superior in linguistic tasks and safety for languages like Vietnamese, Indonesian, Thai, and others, reflects Alibaba's ambition to strengthen its presence in Southeast Asia, as exemplified by its e-commerce platform Lazada's goal of reaching a $100 billion turnover by 2030. (SCMP)
DATA TO USE
China is projected to spend $26B annually in AI technology investments by 2026, accounting for 8.9% of global investment, trailing only behind the US.
Know someone who should be in the know about AI and China? Share The AI China Brief 🇨🇳 with your friends and colleagues!
Have feedback on how we can improve? Want to discuss this issue’s content? We’d love to hear from you in comments below!