For most of the 20th century, AI was mostly tasked with relatively “simple” computational problems like winning a game of chess (IBM’s supercomputer Deep Blue achieved a milestone victory over then world champion Gary Kasparov in 1997) or structuring “big data” (Amazon’s early success with recommending customers “what to buy next” was a commercial hit).
A breakthrough came in 2017 when Google released a paper on the concept of Transformer networks. This new concept, in combination with deep learning (an AI method that teaches computers to process data in a way that is inspired by the human brain, related to neural networks), led to a revolution in natural language processing which then led to the development of large language models (LLM) such as GPT-3 which can generate coherent output from natural language prompts. This new type of AI is called generative AI (GenAI) which can create new content such as text, images, code, audio, or video based on prompts or data.
What is the market for GenAI?
The simple answer is that we are creating so much unstructured data that no human being is able to deal with it. For example, in just a minute, internet users typically upload 500 hours of video on YouTube right through to sending an impressive 231 million emails.
It is therefore not difficult to see that this enormous amount of data needs to be processed, sent over a network, and stored, only to be analysed for a range of purposes after the fact.
It is guesstimated that in the last two years alone, 90 per cent of all the data in all of mankind’s history was created. And we are just getting started.
Intel has calculated that one fully autonomous car (we are not that far off) will generate 4 terabytes of data every day (equivalent to about 3,000 internet users). This number only goes up exponentially when more cars are on the road that need to communicate with each other. The internet of things will also see an explosion in connected devices and the giant data demand of virtual reality are two more examples of this exponential growth in data creation.
The need for help with this data is obvious and it is therefore no surprise that the first mainstream GenAI applications have taken off like nothing before. When OpenAI released ChatGPT in November 2022 to the wider public, demand exploded and quickly led to a queue to get access to the website.
The recently introduced smartphone app held the record of reaching 100 million users in two months – a record that has already been broken by Meta’s release of Threads that did the same in a handful of days. For perspective, it took the internet seven years and even Facebook 4.5 years to reach this milestone.
Estimating the market size of the various layers of GenAI is more art than science. What does seem to be clear is that the market for GenAI software and platforms is significantly larger than the market for GenAI semiconductors. IDC is forecasting the wider GenAI software market (which includes models and development) to grow 18 per cent a year to nearly US$800 billion in 2026. PitchBook has the GenAI semiconductor market also growing 18 per cent a year (to 2025) but to less than a 10th the size of the software market. The main semiconductor market will be in ASICs (applications specific chips) that are currently in use by all the hyperscalers.
Additionally, and even more impactful, GenAI will have a significant influence on global GDP. Here we are moving into the realm of speculation, given the early days of this new technology. Three of the larger studies on the impact of AI on GDP show a material uplift.
For example, PwC’s forecast of a 14 per cent global GDP uplift in 2030 translates into nearly US$16 trillion!
The PwC analysis (“Sizing the Price” 2018) notes three main drivers of economic impact:
- Productivity gains from businesses automating processes (including use of robots and autonomous vehicles).
- Productivity gains from businesses augmenting their existing labour force with AI technologies (assisted and augmented intelligence).
- Increased consumer demand resulting from the availability of personalised and/or higher-quality AI-enhanced products and services.
The negative impacts from GenAI, as has been the case for other revolutionary technology introductions, will be job displacements, as many repetitive tasks like coding, data input, and call centre interactions will be replaced by machines. The impact will resonate through most, if not all, conceivable sectors in our view and the examples of potential use cases are endless.
This list is far from exhaustive but it is worth noting that as has been the case with other eras of productivity transformation such as the industrial revolution, old jobs and industries are replaced with new jobs and industries so net-net, while disruptive in the short to medium term, the longer-term outcome can be positive.
Equity markets have taken a big leap of faith and have looked for the best possible “pure plays” to invest in this new thematic in 2023. The poster child of this phenomenon has been Nvidia, the market leader in GPUs, which is the ultimate pick and shovel of GenAI. The share price of Nvidia is up by 200 per cent this year to date alone, particularly driven by the increased revenue guidance in May, based on strong demand for their AI chips. The stock has now joined the exclusive US$1 trillion market capitalisation club and has become the largest semiconductor stock on this measure, ahead of other giants like TSMC, Samsung, and Intel.
The search by the market for GenAI beneficiaries has also resulted in an explosion of the share prices of the seven largest US companies by market cap, Microsoft, Apple, Alphabet, Amazon, Meta, Tesla, and Nvidia such that they are now close to one-third of the US market.
Conclusion
The rapid rise of GenAI has only just begun, creating exciting opportunities for investors. In our view, GenAI qualifies as a revolutionary development in technology, like the internet, iPhone, and cloud computing before it.
Theo Maas, portfolio manager, global equities, Northcape Capital
