Cover Story: DeepSeek Sets Up Race for Chinese Dominance in AI
Listen to the full version
In the weeks following the Lunar New Year, DeepSeek has shaken up the global tech industry, igniting fierce competition in artificial intelligence (AI). From computing power to applications, large language models and cloud services, companies are scrambling to position themselves in the next wave of dominance over machine learning.
DeepSeek’s emergence has ushered in a new era of AI in China, where the focus has shifted from consumer adoption to raw technological advancement. As tech giants race to refine their models, secure computing power and stake their claim in the next frontier for expert systems, one thing is clear: the competition is just getting started.
Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.
Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.
- RELATED
- DIGEST HUB
Digest Hub
Back
- DeepSeek's rise has intensified global AI competition, emphasizing technological advancement over consumer adoption and causing infrastructure challenges due to its rapid model adoption.
- Major companies like Tencent and Baidu have integrated DeepSeek's technology, while investments in AI infrastructure are surging, with Alibaba announcing a significant investment to bolster AI capabilities.
- DeepSeek's impact is reshaping the Chinese tech sector and market sentiment, with investors pivoting focus towards infrastructure and deployment firms as foundational AI models face reevaluation.
[para. 1] In the weeks following the Lunar New Year, DeepSeek has created a significant stir in the global tech industry by sparking fierce competition in artificial intelligence (AI). The company, based in Hangzhou, China, introduced the R1 inference model, which quickly gained popularity, although its success brought infrastructure challenges due to overwhelming demand. Consequently, companies realized the importance of integrating DeepSeek technology and securing computing power to manage the surge in demand for AI-powered applications. This has set off a race among tech giants to optimize their infrastructure and refine their models.
[para. 2] The response from the industry was swift. Infrastructure providers launched services tailored for DeepSeek, and model developers embraced open-source frameworks. The scramble to integrate DeepSeek has also spread internationally, with companies in the U.S. and China rolling out rival neural networks to keep pace with the fast adoption of DeepSeek-R1.
[para. 3] A major move came from Tencent, which quietly began an AI search function powered by DeepSeek-R1 within WeChat. Although not publicly announced, this integration attracted significant market attention due to WeChat's massive user base of 1.38 billion monthly active users. However, this led to substantial computing power consumption, necessitating a shift to Tencent's chatbot, Yuanbao, to manage demand.
[para. 4] DeepSeek's impact has been multifaceted, marking a technological shift by excelling in complex reasoning tasks. This has prompted Chinese tech giants such as Baidu, Alibaba, and ByteDance to enter the AI race, launching their offerings to compete in this evolving landscape.
[para. 5] The competitive environment has forced AI companies to reconsider their strategies, prioritizing technical advancements over mere user acquisition. The urgency to expand AI infrastructure has driven significant investments, like Alibaba's $52 billion plan over three years, to bolster AI and cloud infrastructure. The rise of DeepSeek has also caught the attention of global investors, boosting confidence in the Chinese tech sector significantly.
[para. 6] The rise of AI has intensified the demand for computing power, pushing companies to seek alternatives to Nvidia's GPUs. Some, like ByteDance, are aggressively expanding their AI capabilities, while others, like Tencent, face GPU shortages that strain their infrastructure. Globally, the race is on to develop advanced AI models, with U.S.-based companies like Elon Musk’s xAI and OpenAI releasing new models that challenge existing capabilities.
[para. 7] DeepSeek's rapid pace of development continues to catch the industry by surprise, with the next generation of its model, DeepSeek-R2, anticipated soon. This fast evolution puts pressure on other Chinese firms and tech giants to keep pace.
[para. 8] The demand for AI has transformed the AI infrastructure landscape, with Model-as-a-Service (MaaS) providers emerging as key players. These firms have rushed to launch DeepSeek-powered models, facilitating AI integration without hefty infrastructure investments. Adoption has spread beyond tech companies to include government and public services, illustrating the broadening impact of AI.
[para. 9] DeepSeek's rise has also shifted investment dynamics within the tech sector. Investors now pursue stakes in AI infrastructure and deployment technologies. Companies like SiliconFlow and Together AI have raised substantial funding, reflecting a pivot towards supporting AI inference and deployment solutions. This marks a significant shift in where potential growth and innovation are expected within the AI landscape.
- DeepSeek
- DeepSeek, a Chinese AI company based in Hangzhou, has rapidly gained prominence with its R1 inference model, setting off a competitive surge in AI technology. Its open-source framework and advanced capabilities in multi-step reasoning, dubbed "deep thinking," have shifted industry focus towards technological advancement and AI infrastructure. Main Chinese tech firms like Tencent, Baidu, Alibaba, and ByteDance quickly integrated DeepSeek's model, while the firm plans an R2 release, indicating robust iteration speed and influence on global tech markets.
- Tencent
- Tencent integrated the DeepSeek-R1 AI model into WeChat as a test, leading to substantial server load and subsequently moving the AI search function to the separate Yuanbao chatbot app. Tencent faced GPU shortages and strained resources due to increased demand. Despite having over 200,000 GPUs, efficient allocation is needed across Tencent's AI applications and services. The company's maneuvering reflects broader competition in AI adoption among Chinese and international tech giants.
- Baidu
- Baidu launched its Deep Search feature powered by DeepSeek-R1 and its Ernie program, attracting over 10 million users in its first hour. The Ernie 4.5 model is set for release on March 16, and Baidu plans to adopt an open-source model by June 30. However, it has struggled with computing power procurement and major updates since Ernie 4.0, facing challenges in balancing AI and autonomous vehicle goals amid declining profits.
- Alibaba
- Alibaba responded to the AI surge by announcing a 380 billion yuan investment to develop AI and cloud infrastructure. The company also introduced the Qwen2.5-Max model, claiming it surpasses DeepSeek-V3 and is expected to release the full version in March. Additionally, Alibaba's DingTalk has integrated DeepSeek, and the company is actively expanding its AI capabilities while competing in the rapidly evolving AI landscape.
- ByteDance
- ByteDance is expanding aggressively, acquiring 100,000 modified Nvidia chips and recruiting AI talent, including Wu Yonghui from Google's DeepMind. Although ByteDance denied the chip acquisition claims, it shows their intent to strengthen AI capabilities. They integrate DeepSeek, which influences their technological strategy. ByteDance also focuses on AI infrastructure expansion and talent acquisition to maintain competitiveness in the rapidly evolving AI market.
- Microsoft
- The article mentions that Microsoft, along with other cloud providers like Amazon Web Services, has integrated DeepSeek into their platforms. This indicates that Microsoft is actively participating in the AI arms race by adopting and supporting state-of-the-art AI models to enhance its cloud services.
- Amazon Web Services
- Amazon Web Services' Chinese partner, Ningxia Western Cloud Data, has collaborated with SiliconFlow and Huawei to expand the deployment of DeepSeek's R1 inference services. This partnership is part of a broader trend where cloud providers, including AWS, are integrating DeepSeek into their platforms, indicating a bet on the ongoing momentum and demand for advanced AI models like DeepSeek-R1.
- Nvidia
- Nvidia's high-performance GPUs remain essential for DeepSeek and other AI firms to expand computing capacity, despite a growing interest in domestic Chinese alternatives. ByteDance reportedly acquired 100,000 modified Nvidia chips, although the company denies this claim. Meanwhile, Tencent is rationing its computing power due to GPU shortages, highlighting Nvidia's pivotal role in the current AI landscape.
- SiliconFlow
- SiliconFlow is an AI infrastructure firm that rapidly rolled out DeepSeek-R1 inference services, experiencing explosive user growth. It partnered with Huawei and Amazon Cloud's Chinese partner, Ningxia Western Cloud Data, to expand DeepSeek deployment. SiliconFlow recently secured over 100 million yuan in pre-Series A funding led by China Growth Capital and Puhua Capital, positioning itself as a key player in AI model deployment and inference computing.
- WuWen Xinqiong
- WuWen Xinqiong is a Model-as-a-Service (MaaS) provider that has built a computing power pool integrating DeepSeek-R1 across seven Chinese AI chip platforms. By doing so, it eliminates compatibility concerns and simplifies large-model adoption for enterprises. The company emphasizes DeepSeek's role in accelerating AI accessibility, particularly for mid-sized enterprises that focus on core businesses while leaving AI infrastructure development to firms like WuWen Xinqiong.
- Huawei
- Huawei partnered with SiliconFlow to expand DeepSeek deployment using Huawei’s Ascend AI chips. This collaboration resulted in explosive user growth for SiliconFlow shortly after launching DeepSeek-R1 inference services. Huawei is part of the wave of hardware makers releasing DeepSeek-compatible AI devices, contributing to the surge in appliance adoption for local AI integration.
- Lenovo
- Lenovo is mentioned as one of the hardware makers rushing to release DeepSeek-compatible AI devices. They are part of the movement to provide AI appliances—integrated computing devices designed for local AI deployment—catering to the growing demand for AI infrastructure following DeepSeek's rise.
- Inspur
- Inspur is mentioned as one of the hardware makers rushing to release DeepSeek-compatible AI devices in response to the growing demand for AI appliances. These devices combine high-performance chips with large-capacity storage and pre-optimized software, allowing businesses to locally deploy AI models more cost-effectively.
- Volcano Engine
- Volcano Engine is mentioned as one of the hardware makers that have rushed to release DeepSeek-compatible AI devices. This indicates their involvement in the development and supply of AI appliances, which are integrated computing devices designed to facilitate local AI deployment, particularly in response to the growing demand for AI powered by DeepSeek's technology.
- China National Petroleum Corp
- China National Petroleum Corp. (CNPC) quickly pivoted to DeepSeek's AI model after observing its rapid adoption. During the Lunar New Year, CNPC partnered with China Mobile to deploy DeepSeek-R1 on domestic hardware, completing the full-stack adaptation and deployment within weeks. This shift indicates a broader move toward DeepSeek across China's industrial giants, following the trend of integrating advanced AI models into operations.
- China Mobile
- China Mobile, like other cloud providers, has integrated DeepSeek into its platforms, betting on its continued momentum. Over the Lunar New Year, China National Petroleum Corp. partnered with China Mobile to deploy DeepSeek-R1 on domestic hardware, showcasing DeepSeek's rapid full-stack adaptation and deployment across industrial giants in China.
- Sinopec
- Sinopec swiftly integrated DeepSeek after observing its rapid adoption. The company partnered with China Mobile to deploy DeepSeek-R1 on domestic hardware, completing the adaptation and deployment in a short period. This move signifies a broader shift towards DeepSeek across China's industrial giants.
- Sinochem Group
- Sinochem Group quickly pivoted to using DeepSeek after observing its rapid adoption. Over the Lunar New Year, it partnered with China Mobile to deploy DeepSeek-R1 on domestic hardware, completing full-stack adaptation and deployment within weeks. This shift signals a broader trend among China's industrial giants toward integrating DeepSeek's AI capabilities.
- February, 2025：
- ByteDance recruited Wu Yonghui, a former vice president at Google's DeepMind, to lead its foundational AI research.
- On or After Lunar New Year, 2025：
- Major MaaS providers launched full-scale DeepSeek-R-1 powered models following a surge in demand.
- Mid-February, 2025：
- SiliconFlow launched DeepSeek-R1 inference services on Huawei's Ascend AI chips, partnering with Huawei and Ningxia Western Cloud Data for deployment expansion.
- Feb. 16, 2025：
- Tencent quietly rolled out an AI search function within its social media app WeChat, powered by DeepSeek-R1.
- Feb. 16, 2025：
- Baidu launched Deep Search, powered by DeepSeek-R1 and its own Ernie program.
- Feb. 17, 2025：
- Tencent shifted AI search functions to its chatbot Yuanbao due to high computing power demand and server load issues.
- Feb. 18, 2025：
- Baidu's Deep Search feature appeared on its search homepage, drawing more than 10 million users in its first hour.
- Feb. 24, 2025：
- Alibaba announced a 380 billion yuan investment over three years to develop AI and cloud infrastructure.
- Late February, 2025：
- Governments in cities like Hohhot, Guangzhou, Wuxi, and Wenzhou integrated DeepSeek into public services.
- PODCAST
- MOST POPULAR
Sign in with Google
Sign in with Facebook
Sign in with 财新