AI: There is a lot of potential for investors in the supply chain  

When it comes to artificial intelligence, a lot of people say it feels like a new era was ushered in this year. Historically, similar breakthroughs happened roughly every ten years. If the pattern repeats itself, we are indeed witnessing the dawn of a new era.  

By Mike Glöckner, Analyst in the Research Team of DJE Kapital AG

Many stocks in the technology sector have already benefited greatly from developments in AI. These include cloud hyperscalers Amazon, Google, Microsoft and the well-known chip manufacturer NVIDIA. The companies mentioned are also among the so-called "magnificent seven" that have almost exclusively driven the positive overall performance of the S&P 500 this year. The valuations of these companies on the stock market are correspondingly high. But the enormous demand for computer chips that make artificial intelligence possible in the first place also benefits many companies that are active in the manufacturing process of these chips.   

Many of these companies have experienced a positive development since the beginning of the AI boom. This is no guarantee for future performance but the consensus is that the sector will continue to grow: Morgan Stanley Research estimates that the AI technology sector (chips, hardware, and network) is worth around $98 billion (USD) today and is expected to grow to $275 billion by 2027.   

Many of the companies in the supply chain are headquartered in the US or in Asia (primarily in Taiwan), but there are also some in Europe. These are for the most part highly specialised companies, some of which have market capitalisations of only a few billion US dollars and some of which have also achieved a performance like NVIDIA this year, although they are much less well known.   

NVIDIA's market share for the so-called artificial intelligence graphics processing units (AI GPUs) is around 80 percent. However, this market share is likely to decline somewhat in the future, as a competing company that makes high-performance and adaptive computing products plans to launch a new chip this quarter. The so-called AI-GPU MI300 chip is specifically designed for "large language models" such as Chat GPT. Other large IT corporations have also invested billions in the field of AI.   

Number of transistors per chip increases at high speed 

Technological advancements require ever more powerful chips, which must be designed and developed first. For almost 60 years, the number of transistors in a chip doubled about every two years. This "law" was an observation made by former Intel CEO Gordon Moore in 1965.   

However, this speed can no longer be maintained. Therefore, the chip industry has started to implement optimisations in other areas of the design or "architecture" of a chip, so that the overall performance of the chips continues to increase at a similar rate. This, however, has made the design process for chips very expensive. For example, the current architecture of NVIDIA's new H100 AI chip is home to 80 billion transistors, and the entire multi-chip HGX H100 board is made up of up to 45,000 different components that all need to communicate with each other and in parallel.   

The development of a chip today therefore costs several hundreds of millions of USD. In addition, the chip must be tested extensively before it can go into mass production.  

Companies have specialised in the development of these designs, which offer ready-made components consisting of several elements as buildings blocks. This in turn saves time and money for chip engineers in developing new chips, so that the wheel does not have to be reinvented every time.   

The global market leaders in chip design include two American and one English company. Although some of these companies initially specialised in one area, some are now expanding into other promising areas such as self-driving cars and high-performance servers for data centres as well as "smart home" devices and "wearables".   

Specialised chips 

In addition to the exploding demand for "general-purpose" chips for a wide variety of applications, there is also a strong trend towards chips that are developed specifically for certain purposes. Many of these chips are optimised in terms of cost/benefit. In this area there are highly specialised, smaller Taiwanese chip designers, whose main customers are companies like Amazon, Google, Microsoft and a well-known American electric car maker.  

In addition to chips for artificial intelligence applications, there are also manufacturers that specialise more in high-performance memory chips.   

These are called high-bandwidth memory (HBM) chips. Two South Korean companies dominate in this field. Even the precise installation of the chips and the connection to other components is so technologically demanding that there are also specialised companies in this field. In addition, there are manufacturers which specialise in the printed circuit boards that house the electronic components (PCBs). Suppliers from Taiwan are dominant in this area and known for their durable and reliable circuit boards of high quality. 

AI servers 

Customers like Amazon, Google, Meta or Microsoft often do not buy the AI circuit board directly (with all the components on it), but entire AI servers. These are computers in which all components, including network and power connections, AI circuit board, additional CPUs, memory, cooling, etc., set up for the best performance, so that all you have to do is to connect them in the customer's data centre. There are some interesting companies from Taiwan in this area that are worth considering for investors.   

Testing the components & the whole system  

The individual components must be tested extensively before delivery. An NVIDIA AI system (circuit board with components and chips) costs between about 20,000 USD and 75,000 USD, depending on the equipment. Testing is a very complex matter and can take several weeks. Companies from Japan and Taiwan are well positioned in this market segment.    

For investors it pays off to also look beyond the users of artificial intelligence and consider companies in the supply chain of the components as potential investments. 

Note: Marketing advertisement - All information published here is for your information only and does not constitute investment advice or any other recommendation. The statements contained in this document reflect the current assessment of DJE Kapital AG. These may change at any time without prior notice. All statements made have been made with due care in accordance with the state of knowledge at the time of preparation. However, no guarantee and no liability can be assumed for the correctness and completeness.