Thursday, 13 February 2025
Written by MYTHEO
‍

Key Takeaways
‍

The launch of DeepSeek triggered a steep sell-off in hardware-focused tech stocks, with established semiconductor players experiencing share price losses of up to 17.40% in a single day.
‍
Chinese AI models are not a new phenomenon. Since the rise of ChatGPT, several Chinese tech giants have introduced their own models, including Baidu’s ERNIE, Alibaba’s Qwen, Tencent’s Yuanbao, and SenseTime’s SenseNova
‍
DeepSeek’s arrival is set to intensify competition, pushing rivals to accelerate innovation. As companies strive to outdo one another, the escalating AI battle could fuel another investment boom, creating a fresh catalyst for long-term growth in the sector.

DeepSeek Shakes Up AI Rally: A Threat to Chip Giants or Just Another Hype Cycle?
‍

A new artificial intelligence (AI) model called DeepSeek launched in January, sparking panic in the market.
‍
DeepSeek, a China-based large language model (LLM) developer, has ignited debate over whether major semiconductor companies like Nvidia, TSMC, and Broadcom could face significant disruption. Nvidia and Broadcom are key components of several US-based ETFs, including those held by MYTHEO, while Taiwan Semiconductor Manufacturing Co. (TSMC) is a major holding in many non-US ETFs, particularly those focused on emerging markets and the Asia-Pacific region.

These semiconductor giants—alongside major tech firms like Alphabet, Meta, Tesla, Apple, and Microsoft—have been driving the equity market’s strong performance in recent years. However, following a market selloff on January 27, concerns arose over whether DeepSeek could derail the bull market anchored by these companies.
‍
What made the market’s reaction to DeepSeek particularly unusual was its disproportionate impact. Despite being an LLM competitor, DeepSeek’s announcement had little effect on existing AI developers such as Microsoft and Alphabet. Instead, the sharpest declines were seen in hardware and semiconductor stocks like Nvidia, TSMC, AMD, and Broadcom.

On that day, Nvidia, Broadcom, TSMC, and AMD saw their shares tumble by 16.97%, 17.40%, 13.33%, and 6.37%, respectively, as investors worried about potential disruptions to the demand for high-performance AI chips. Meanwhile, DeepSeek’s direct competitors, Microsoft and Alphabet, experienced relatively modest declines of 2.14% and 4.20%, respectively.

Performance of AI-related Stocks on 27^th January 2025
‍

Source: Google Finance, February 2025

The prevailing market narrative suggests that DeepSeek could challenge the long-term growth of high-end chipmakers like Nvidia and TSMC because DeepSeek Version 3 was developed and trained at a fraction of the budget spent by OpenAI, Alphabet, and Microsoft. Furthermore, DeepSeek V3 runs on an older, less powerful Nvidia H800 chip, raising speculation that AI models may no longer require cutting-edge hardware to compete.

‍As a result, investors assumed that DeepSeek could gain market share without relying on expensive or high-performance chips. However, we believe this narrative is overly speculative, and DeepSeek’s current advantage may be short-lived.

Top Large Language Models (LLMs) Available Today
‍

‍‍
The emergence of Chinese-developed LLMs is not new. Since the launch of ChatGPT in November 2022, several Chinese tech firms have introduced their own models, including Baidu’s ERNIE, Alibaba’s Tongyi Qianwen (Qwen), Tencent’s Yuanbao, and SenseTime’s SenseNova.

When DeepSeek was first launched in November 2023, it received little publicity. It only began gaining public attention after announcing its first free chatbot model, DeepSeek-R1. DeepSeek’s announcement drew comparisons to SenseTime’s launch of SenseNova 5.0 in April last year, which was claimed to outperform OpenAI’s GPT-4—the most advanced model at the time.However, such claims have repeatedly been challenged.

In fact, it didn’t take long for another Chinese competitors to respond. On the first day of the Chinese New Year, Alibaba unveiled an upgrade to its Qwen AI model, known as Qwen 2.5. Following the same playbook as SenseTime and DeepSeek, Alibaba also claimed that its AI model surpassed all existing ones in the market.

While DeepSeek is a significant development, it is far too early to declare any AI model the ultimate winner or to conclude that high-end chip demand will decline. Instead, we believe that DeepSeek’s emergence will further intensify competition among LLM developers, pushing them to invest more aggressively—including adopting cutting-edge chips—to gain an edge.

Rather than undermining semiconductor giants, DeepSeek’s rise has triggered a wave of reactivity among existing AI players, accelerating the evolution of LLM-based AI models. This rapid innovation could spark a new growth cycle in AI, ultimately benefiting companies like Nvidia, Broadcom, and TSMC.

Despite the hype surrounding DeepSeek, it is premature to crown any AI model as the dominant force in the industry. Instead, we expect it to push competitors to accelerate development and increase investments to reclaim leadership. This escalating competition will likely fuel another AI boom—providing a catalyst for continued growth in the semiconductor sector.  

This material is subject to MYTHEO’s Notice and Disclaimer.