According to reports by McKinsey and PwC, the AI revolution could have an annual impact of $13-15 trillion on global GDP by 2030. Now some of these figures are not the latest estimates, but they give us a good sense of where things are headed. Incorporating recent developments in AI could increase these estimates significantly. For those who wish to invest internationally, in the US technology industry and global chip makers, here’s an interesting investment idea.
We believe that the state-of-the-art in current AI is so advanced that companies have been trying to hide, or underplay, the capabilities so as to avoid panic-mongering by the media, hysterical responses from the general population, and calls for stringent and highly restrictive regulations by politicians on the use of data and AI techniques.
In any case, the AI revolution is here as emphatically demonstrated by ChatGPT. In an earlier article we discussed the threats and opportunities from an investment perspective that ChatGPT is a harbinger of.
If AI is akin to the human mind, which thinks and generates insights, then AI chips are akin to the human brain, which stores and processes the sensory inputs. The processing of the new sensory inputs and stored inputs as well as thoughts by the human brain results in new insights by the human mind. AI chips, too, store sensory inputs in the memory chips and process those inputs to produce insights.
In this piece, we will focus on some of the key firms in the AI chip design and manufacturing ecosystem and potential investment opportunities. Of course, any mention of names of companies shouldn’t be assumed to be recommendations to buy their stocks. While a company might play an important role in the ecosystem, an investment decision has to be far more sophisticated than buying the best company. It should incorporate the strength of its resources, including monetary, human and intangible assets, its competitive advantages, and market price versus intrinsic value. We have discussed the Scientific Investing Framework in several other articles earlier.
Why artificial intelligence will shake up tech
AI requires large amounts of storage, and fast storage and retrieval, and this brings the need for specialised memory chips designed for AI applications. It typically requires around five times more bandwidth and could be three times more expensive. Micron, SK Hynix and Samsung are some of the companies focused on these memory chips.
ALSO READ: Coming soon: Mental health, disabilities, HIV/AIDS to be covered under a standard insurance policy
However, the more interesting battle is going on in the AI-focused processor chips. While conventional CPUs can be used for AI computing, these are much slower and inefficient. The chips for AI include GPUs (Graphics Processing Units), FPGAs (Field-Programmable Gate Arrays) and ASICs (Application-Specific Integrated Circuits).
The typical AI process requires training as the first step where large amounts of data are fed to the AI engine and it creates a model out of it. Now this model can be given new data and it will make inferences from that and share these with the user. The system requirements for training are much more intensive and this is where the GPU chips have an advantage. FPGA chips are better suited for inferencing. ASICs can be used for both.
Mapping the companies with their strategies
One of the top names in AI chips is NVIDIA. However, there is competition from Intel, AMD, Qualcomm and a host of others. Interestingly, nearly all the Big Tech names have started designing their own chips. For example, the TPUs (Tensor Processing Units) from Google are ASICs designed to solve matrix and vector operations for neural network models for deep learning.
ALSO READ: Investors start warming up to debt funds, equity flows continue to be strong: AMFI
Amazon has been designing chips for nearly a decade, making the crucial acquisition of Annapurna Labs in 2015. Currently, its two chips—Trainium and AWS Inferentia—designed, quite obviously, for training and inferencing, respectively. Graviton series is another AWS chip which is powerful.
Then there is the AIU (Artificial Intelligence Unit), another ASIC, from one of the pioneering companies in AI, IBM. This is the latest evolution of their AI chip Telum. IBM’s AI Hardware Center was launched in 2019 with the aim of training and running AI models 1,000 times faster by 2029.
Microsoft’s Project Brainwave is also focused on specialised AI chips based on FPGA. Microsoft recently acquired a chip design company called Fungible.
Apple has been working on its own chips for quite some time, termed Apple Silicon. These chips are being used in all their own hardware ranging from the iPhone to iPad, Mac, etc. Apple has a specialised AI chip called ANE (Apple Neural Engine). ANE is an NPU (Neural Processing Unit) which is designed to speed up matrix operations and convolutions which are required for neural network operations.
While Meta Platforms (Facebook) has also dabbled in designing its own chips, it has partnered with Qualcomm and Broadcom for the near future to develop customised chips for its requirements for AI, VR and Metaverse.
ALSO READ: High cash generating companies | Mutual funds love to hold these midcap and smallcap stocks
While we have not elaborated on NVIDIA’s GPUs, they are currently the largest-selling chips for AI. Intel, AMD and Qualcomm are the other important mainstream players. These are the traditional, well-known semiconductor chip players and so we wanted to rather highlight the other efforts going on in the AI battle besides these.
What should investors do?
Besides these there are numerous specialised AI companies but many of them might not be listed or might be too small or might not be generating profits yet. But they offer a peak of what’s to come in the future.
ALSO READ: Last-minute tax savings: Why an ELSS works well for patient investors
Investors keen to invest in the US technology sector—and even some of the other global chip makers—must use this as a starting point to dig deep into this highly interesting AI chips battle. This is just the beginning. In future, this space will discuss many other exciting aspects of the AI revolution growth vector.
Disclaimer: Please note that any mention of company names is not a recommendation to buy, sell or hold. Equity investments are subject to market risks. Past performance is no guarantee of future performance. Global Investing has additional risks. One should invest based on the advice of their financial advisor based on their investment objectives, financial situation and risk profile. OmniScience Capital, its management and employees and its clients might be buying, selling or holding the mentioned companies.