The competitiveness of AI in China has shifted from 'catching up in computing power' to 'defining applications.'
In the past two weeks, rumors about the "DeepSeek new model R2 is about to be released" have continued to gain traction.
At the end of April, the post by Clément Delangue, the CEO of the world's largest AI open-source community Hugging Face, on social media, which consists solely of three eye emojis and is accompanied by an entry link to the #DeepSeek team's official repository on the Hugging Face platform, caused quite a stir online. Everyone is guessing: "#DeepSeek is up to something big!"
As time moves into mid-May, discussions about DeepSeek R2 have intensified, and posts about R2 keep coming in.
Market news reveals that DeepSeek will launch the upgraded model DeepSeek R2 in May, expected to adopt a more advanced Mixture of Experts model (MoE), with a total parameter count about double that of the previous generation R1, further pushing AI from "content generation" to "logical validation," seen as an important advancement towards general artificial intelligence (AGI).
Market opinions suggest that the DeepSeek team has maintained a historical rhythm of significant quarterly updates, and it is believed that the official release is likely to be very soon, which will undoubtedly stir up the DeepSeek topic once again.
In fact, earlier this year, DeepSeek emerged and led global capital to reassess the value of Chinese assets. Morgan Stanley recently published a 97-page global technology blue paper titled "China – AI: The Sleeping Giant Awakens," indicating that China is focusing on how AI can drive industrial transformation on a large scale, turning constraints into opportunities.
This report conveys a clear message: the competitiveness of AI in China has shifted from "catching up on computing power" to "defining applications."
It is worth highlighting that Morgan Stanley has also released the "China AI 60" list for the first time, covering three major levels: Hardware, Platforms, and Applications.
Infrastructure:
Chip stocks $SMIC (00981.HK)$ 、 $LENOVO GROUP (00992.HK)$ 、 $ACM Research (ACMR.US)$；
Datacenter $GDS Holdings (GDS.US)$ 、 $21Vianet (VNET.US)$ ；
In the Electrical Utilities Industry $WEICHAI POWER (02338.HK)$ Etc.;
Platform Layer:
$TENCENT (00700.HK)$ 、 $Alibaba (BABA.US)$ 、 $Baidu (BIDU.US)$ 、 $Iflytek Co.,ltd. (002230.SZ)$ ；
Application Layer:
Software Company $MEITUAN-W (03690.HK)$ 、 $MEITU (01357.HK)$ 、 $KUAISHOU-W (01024.HK)$ 、 $KINGDEE INT'L (00268.HK)$ 、 $Trip.com (TCOM.US)$ 、 $NetEase (NTES.US)$ 、 $KE Holdings (BEKE.US)$ 、 $PDD Holdings (PDD.US)$ 、 $JD.com (JD.US)$ 、 $Kanzhun (BZ.US)$ Etc.;
Autonomous Driving Industry $HORIZONROBOT-W (09660.HK)$ 、 $XPENG-W (09868.HK)$ 、 $BYD COMPANY (01211.HK)$ 、 $NIO-SW (09866.HK)$ 、 $LI AUTO-W (02015.HK)$ 、 $ZEEKR (ZK.US)$ ；
In the Medical Industry $ALI HEALTH (00241.HK)$ 、 $GUSHENGTANG (02273.HK)$ Etc.;
Consumer Industry $HAIER SMARTHOME (06690.HK)$ 、 $POP MART (09992.HK)$ 、 $GIANT BIOGENE (02367.HK)$ 、 $MIDEA GROUP (00300.HK)$ 、 $YUM CHINA (09987.HK)$ ；
In electronic products $XIAOMI-W (01810.HK)$ 、 $Tuya Inc (TUYA.US)$ ；
Logistics stocks $SF HOLDING (06936.HK)$ ; Education stocks $TAL Education (TAL.US)$ ; Financial Software $BAIRONG-W (06608.HK)$ 。
Notably, in this research report, Morgan Stanley categorizes China's AI applications into two main lines, covering the Consumer field (2C) and the Enterprise field (2B).
In the 2C field, the integration of AI-native applications and super applications will drive rapid adoption. For example, ByteDance's Doubao and Tencent's Yuanbao, as AI-native applications, attract a large number of users by providing personalized content recommendations and smart interaction features.
In the 2B field, the application of AI in enterprises will become more diverse, covering multiple industries from manufacturing to services. For example, AI technology can be used for quality inspection in manufacturing, supply chain optimization, as well as risk assessment and customer service in the financial industry.
The firm also stated that, despite the demand for cost reduction in the 2B field, factors such as lower enterprise IT spending, a preference for private cloud deployment, and an immature Software industry may cause the large-scale application of AI in the 2B field to proceed slower than in the 2C field. However, B2B Software companies with proprietary data are still expected to create valuable enterprise-level AI applications.
Monetization path: The 2C end relies on advertising (Tencent's advertising ROI increased by 20%) and transaction commissions, while the 2B end shifts to 'pay per token' (DeepSeek's inference cost as low as $0.4 per million tokens, only 4% of GPT-4's).
Overall, this research report from Morgan Stanley serves not only as an investment guide for investors but also as a panoramic development 'white paper' decoding the evolution of China's AI industry. From self-sufficiency in computing power to blossoming applications, China's 'rise script' of AI is being staged.
Want to know more interpretations?Come and use Futubull AI!Providing precise answers, comprehensive insights, and grasping key opportunities!
Editor/Somer
