DeepSeek’s strong new AI model – AI’s Sputnik moment?

Conviction Equities Boutique
Read 3 min

Shares in Nvidia and ASML have plummeted due to concerns that Chinese AI startup DeepSeek's new model could disrupt the current AI business model. DeepSeek's R1 model, released on January 20, exhibits performance similar to top US large language models (LLMs), but at a fraction of the cost. This has sparked concerns about the need for high-end chips and computing power. The development suggests Chinese AI engineers have found a way to work around US-led export bans on advanced semiconductor technologies.

Technological advancement shows much higher cost efficiency is possible, but we doubt this will cause slowdown in investments just yet

The astonishing performance of DeepSeek's model is mainly due to a combination of architectural optimizations such as Mixture-of-Experts design and custom communication schemes between chips. This technological advancement shows that much higher cost efficiency is possible in AI development. However, this is unlikely to cause a slowdown in overall AI investments just yet, as the race towards Artificial General Intelligence (AGI) is in full swing. Meta recently announced over 50% growth in capital expenditures on AI projects in 2025, and Project Stargate from OpenAI/Oracle/Softbank revealed during the World Economic Forum is expected to be worth $500 billion.

Sanctions look ineffective, what will likely provoke further tightening

Despite the US's efforts to stall China's progress in AI through export restrictions, these sanctions appear to be ineffective. DeepSeek states it used Nvidia H800 chips for training, a down-spec'ed version of the H100 for the Chinese market. But it's unclear if H100 chips bought prior to the restrictions were involved, or if the company leased more high-end chip capacity from data centers in South East Asia. The Biden administration's recently proposed tightening of export restrictions would address the latter, and further tightening of restrictions is likely.

Investment implications: AI arms race to continue with growing cost-effectiveness considerations

The investment implications of these developments are significant. The strong Q4 performance in tech hardware was mainly driven by more cost-effective custom AI chip demand, as demonstrated by Amazon. The DeepSeek revelations support the expectation for a continued drive for cost efficiency. However, with high expectations comes increased vulnerability, what lead us to a reduction in hardware exposure into 2025. So now with increased scrutiny on the sustainability of high AI infrastructure investments we expect some initial hit on the hardware stocks, but would perceived this as a buying opportunity as the improved cost-efficiency should ultimately boost the adoption of AI applications and benefit a wider tech supply chain.

Chinese internet companies' ability to succeed in AI development despite sanctions is supported by the news. Lower training and inference costs should generally support faster adoption and monetization of AI solutions. Investors may start to see value in diversifying their portfolios beyond the US-centric stocks and considering equities in markets like China, where companies are making significant strides in sectors like AI, despite regulatory challenges and while US leader’s barriers to entry are put more into question. This could mark a significant shift in investor sentiment, with emerging markets again becoming increasingly competitive and attractive for investment.

 

 

 

 

 

Important Information: This document was produced by a company of the Vontobel Group (collectively "Vontobel") for institutional clients. This document is for information purposes only and does not constitute a solicitation, or offer, research or recommendation, to buy or sell any investment instruments, to effect any transactions, or to conclude any legal act of any kind whatsoever. Information herein should not be considered investment advice or any other kind of advice on legal, tax, financial or other advice or a recommendation to purchase, hold or sell any investment. No representation is given that the securities, products, or services discussed are suitable for any particular investor. Past performance is not a reliable indicator of current or future performance. All investing involves risks including possible loss of principal. Any projections or forward-looking statements regarding future events or the financial performance of countries, markets and/or investments are based on a variety of estimates and assumptions. There can be no assurance that the assumptions made in connection with the projections will prove accurate, and actual results may differ materially. The inclusion of forecasts should not be regarded as an indication that Vontobel considers the projections to be a reliable prediction of future events and should not be relied upon as such. This information is not intended for use in any jurisdiction which would subject Vontobel and/or TwentyFour to any registration, licensing or other authorization requirement within such jurisdiction or country. It is the responsibility of the recipient to inform themselves and observe applicable regulations and restrictions for their respective jurisdiction(s). Information herein is not intended for retail investors, and/or distribution to the general public in any jurisdiction.

About the authors

Related insights