DeepSeek's "shockwave" has a wide-reaching impact, with people regarding its launch as the AI field's "Sputnik moment," yet it seems more likely to be China's stock market's "Alibaba IPO moment."

In a report on February 12, the Bank of America pointed out that it could become a game changer in the AI landscape:

China's AI development "holy grail" may not necessarily be the "best in class," but rather "good enough" to support large-scale AI applications in China and drive productivity and economic growth, with DeepSeek potentially democratizing computing and altering the global AI ecosystem.

It is worth mentioning that Bank of America has compared the emergence of DeepSeek to Alibaba's IPO in 2014, suggesting it may mark a turning point for China's stock market:

Alibaba's IPO attracted long-term global capital inflow into the Chinese market, propelling the development of the XINJINGJI. The emergence of DeepSeek, likened to Alibaba's IPO in 2014, could have a structurally positive impact on China's stock market in the long term.

DeepSeek's technological breakthrough: China's AI "cost revolution."

DeepSeek has become the focus with the launch of its R1, prompting adjustments in American tech stocks and a rebound in Chinese stocks.

The R1 inference model has achieved performance comparable to leading models in the USA while reducing reliance on high-end Hardware, with training costs only a fraction of the latter. For example, it costs only $2.19 per million output tokens, while OpenAI's GPT-4 costs $60.

Bank of America believes that:

By leveraging a mixture of experts architecture (MoE), FP8 mixed-precision training, and reinforcement learning techniques, the R1 model has achieved breakthroughs in inference efficiency, with API pricing only 2%-4% of GPT-01. This innovation not only challenges the consensus that "AI hegemony relies on huge Hardware investments," but also proves that China can still promote AI inclusivity through software optimization despite limited computing power.

Bank of America also summarized the development of DeepSeek:

Technical highlights: The R1 model has only 37 billion active parameters, dynamically allocating computing resources to reduce dependency on high-end GPUs.

Market response: After its launch, R1 topped the free chart on the USA App Store, achieving 16 million downloads in the first month, surpassing ChatGPT's performance during the same period.

Ecosystem integration: Major global cloud service providers such as AWS, Microsoft Azure, Alibaba Cloud, and Tencent Cloud have integrated the R1 model.

Can DeepSeek replicate Alibaba's "structural turning point" from 2014?

In 2014, Alibaba's listing drove the rise of China's XINJINGJI Sector, attracting Global long-term capital inflow. Bank of America believes that DeepSeek could become the "Alibaba IPO moment" for the current Chinese stock market, but caution is needed regarding excessive reactions to short-term market sentiments.

Bank of America believes:

The emergence of DeepSeek can be compared to Alibaba's IPO in 2014, which could become a turning point for the Chinese stock market. Alibaba's IPO attracted Global long-term capital inflow into the Chinese market, promoting the development of the XINJINGJI. DeepSeek's technological breakthroughs may attract more Global capital attention to the Chinese AI field, fostering long-term growth of the Chinese stock market.

Bank of America emphasizes:

The "Holy Grail" of AI development in China doesn't necessarily mean "best-in-class," but rather "good enough" to support large-scale AI applications and drive productivity and economic growth in China. DeepSeek demonstrates that China's private sector still has innovation capabilities and competitiveness, and it also indicates the continued potential for China's progress in the AI field.

Bank of America also believes that DeepSeek will change the Global AI ecosystem:

The emergence of DeepSeek may alter the Global AI ecosystem, promoting the democratization of AI technology and lowering the barriers to AI applications. Through open-source models, DeepSeek provides Global developers with more efficient and cost-effective AI solutions, facilitating the widespread application of AI technology.

DeepSeek promotes the development of multiple Industries.

In addition, Bank of America pointed out that the emergence of DeepSeek will promote the development of multiple Industries such as Software, Datacenter, Semiconductors, Internet, and Electric Autos.

Specifically:

• Software Industry: AI applications may accelerate penetration, and DeepSeek's open-source model lowers the threshold for AI applications, allowing more software companies to quickly integrate AI functions, promoting the development of Software as a Service (SaaS) model.

• Datacenter: In the short term, the Datacenter Industry is limitedly affected, as revenue growth for 2025 has already been secured through previous Orders. However, in the long term, the increase in demand for AI model training and inference will drive growth in Datacenter demand. DeepSeek's technological advancements may further accelerate this trend.

• Semiconductors Industry: DeepSeek's technological advancements may accelerate the development of China's Semiconductors Industry, especially in AI accelerator chips and advanced packaging technologies. There are market concerns about Cloud Computing Service providers cutting capital expenditures, but Bank of America believes that demand has not been disproven, rather it may benefit from the demand growth brought by DeepSeek.

• Internet Industry: DeepSeek's technological advancements may accelerate the applications and developments of Chinese Internet giants in the AI field, enhancing their competitiveness in the Global market. Giants such as Alibaba, Tencent, and Baidu have already made significant investments in the AI field and are expected to further enhance their AI capabilities through DeepSeek's technology.

• Autos and Electric Autos Industry: DeepSeek's technology may accelerate the development of autonomous driving technology, enhancing the interactive experience of Smart Automobiles.

Editor/ping