Sharp corrections

On Monday, January 27, 2025, the technology sector literally "bled". Semiconductor market leaders such as Broadcom (AVGO) and Nvidia (NVDA) turned out to be some of the worst performers, writing off around 17 % of their value. Broadcom shares fell to $202.13, while Nvidia closed at $118.42. The slump also led to a reduction in Nvidia's market capitalisation, with the giant losing around $600 billion of its value, dropping it to the third most valuable company in the world, according to Companiesmarketcap.com. The NASDAQ Composite Index (IXIC) lost more than 3 percent on the day to close at 19,341 points. Yet, like the shares of both companies, it continued to hover near its all-time highs. Other companies in the sector, such as Alphabet and AMD, also faced slightly lower declines.*

Broadcom's share price development over the last 5 years. (Source: Investing.com)*

Nvidia's share price development over the last 5 years. (Source: Investing.com)*

Price development of the NASDAQ Composite Index over the last 5 years. (Source: Investing.com)*

The sell-offs were global

European companies such as ASML (ASML) and Siemens (SIEGn), which provides electrical hardware, were also affected by the declines. On the Tokyo Stock Exchange, for example, semiconductor test equipment supplier Advantest (6,857), Softbank (9,984), owner of Britain's ARM Holdings, and cable maker Fujikura (5,803) all suffered significant losses.* Exchanges in other Asian countries were closed for holidays on the day, so potential changes won't become apparent until later.

It also outperformed the leaders

The negative sentiment among investors and the associated wave of declines was sparked by China's artificial intelligence model DeepSeek. The eponymous startup, founded in 2023, released a new R1 version of its open-source language model earlier this week, taking advantage of Nvidia's reduced chip capacity. This followed a model released last year. As CNBC reports, both products were able to outperform the US tech giants' models in independent third-party tests with their performance. These were the Llama model from Meta, ChatGPT from OpenAI and Claude Sonnet from startup Anthropic. It also convinced consumers of its qualities when it became the most downloaded free app in the US on the Apple Store on 27 January 2025, overtaking ChatGPT and Google Gemini. DeepSeek took a few months and about $6 million to achieve such a remarkable feat, according to their words. In addition, the use of Nvidia chips is also interesting, as the US has been trying to restrict exports of advanced semiconductors to China for several years. As reported by CNBC and CNN, not only did the Chinese startup manage to develop a more powerful model with lower-end chips in such a case, but it even raises the question of whether it was able to circumvent the US rules or if they don't work in practice as the US government thought.

The model has also been praised by competitors

Representatives of technology companies have also commented on DeepSeek, acknowledging its great performance. For example, Microsoft CEO, Satya Nadella, praised its computational efficiency and warned that Chinese development should not be taken lightly. The founder of OpenAI, Sam Altman, wrote on the social network X (formerly Twitter) that it is "encouraging to have a new competitor". According to information published by Reuters, the US President, Donald Trump, has called DeepSeek a "wake-up call" for the US tech sector and Japan's Minister of Digitalization, Masaaki Taira, in turn, has acknowledged that AI from China is not years behind the times as thought.

Big Tech's AI spending

But it was the performance-to-price ratio that generated the most interest, which made investors uncertain whether the big tech companies' spending in AI was really justified. OpenAI will spend $500 billion to expand US infrastructure, the largest project in history, according to Trump. Around $65 billion is planned to be invested by Facebook's owner, Meta. Microsoft intends to spend $200 billion on data centres, while Amazon will use "only" $10 billion. On the other hand, there are those in the market who, according to NBC News, don't believe that the DeepSeek model only cost that little.

* Past performance is no guarantee of future results.