A great leap forward

GenAI is a type of machine learning, whereby users feed data into a computer which can not only solve many problems and learn as it goes along, but also provide output as if it were a human thanks to the use of “large language models.”

Due to its broad potential usage, GenAI applications such as ChatGPT, launched in November 2022, have captured the world’s imagination.

Among other things, GenAI applications can:

• Generate new content such as writing essays and composing emails;
• Edit and summarize text;
• Quickly classify or digest large amounts of data and draw conclusions from it;
• Answer complex queries; and
• Create digital art.

GenAI has the potential to radically change the way tasks are performed, much like the computer did in the late 1980s.

Impact on productivity and economic growth

Inevitably this transformational technology has sparked a debate on the impact it may have on the economy.

One optimistic estimate from the McKinsey Global Institute (2023) suggests that GenAI stands to add up to $4.4 trillion to the global economy on an annual basis compared to a world GDP of just over $100 trillion in 2023.

However, this extreme optimism is not universally shared. In a May 2024 paper, Daron Acemoglu, institute professor of economics at the Massachusetts Institute of Technology, calculates that productivity would increase by a mere 0.06 percent annually.

In the sections that follow, we review key aspects of GenAI technology.

GenAI ecosystem basics

Tech enablers

Hardware

High-powered advanced semiconductors facilitate GenAI as the technology needs to be fed huge amounts of data from which to train machine-learning models.

Such workloads require significant amounts of computing power provided mainly by semiconductors called graphic processing units (GPUs), and by custom accelerator chips. Both can handle large amounts of data and perform an enormous number of calculations simultaneously. NVIDIA is the main designer of the advanced chips used to train and run AI models such as OpenAI ’s GPT-4, the brain power behind ChatGPT. Design companies outsource to Taiwan Semiconductor Manufacturing Company (TSMC) to produce the chips for them.

NVIDIA’s dominance has not gone unnoticed. Established semiconductor firms such as Intel and Advanced Micro Devices are launching rival products and the established cloud companies (including Google’s corporate parent Alphabet, Amazon, and Microsoft) are now designing their own chips to reduce reliance on NVIDIA. Many smaller firms, including startups, are also in the race.

Other than AI semiconductor manufacturers, hardware also encompasses semiconductor equipment manufacturers (such as ASML) plus server and networking equipment makers (such as Dell and Amphenol).

Cloud computing platforms

“Cloud computing” is the term coined to refer to the online availability of computer system resources such as data storage and computing capacity. It offers access to these resources without having to manage them directly.

Most businesses find it more cost effective to build, tweak, and run large AI models in the cloud, rather than have this key hardware platform on premises given that infrastructure requirements are expensive.

Model makers

AI models are pretrained to create content and can be adapted to support a wide array of tasks. Microsoft-backed OpenAI had an early lead with its GPT series, but much competition has emerged, such as Alphabet- and Amazon-backed Anthropic, a privately-held U.S.-based rival, or French AI model maker and private company Mistral. Meta, Facebook’s parent, also issued its own model, Llama, in 2023.

Non-Tech enablers: Data centers, energy providers, industrials

Data centers, already a vital infrastructure required to support mass digitization, are indispensable for AI services. The size of multiple U.S. football fields, data centers host many thousands of servers which process and store online information.

As GenAI takes off, more and larger data centers will be needed as training GenAI models requires placing thousands of advanced chips into data centers and operating them at full capacity for extended periods.

Data centers use vast amounts of energy. Natural gas remains a key source of energy due to current grid constraints, though demand for renewables and nuclear in particular, a more reliable source, should grow over time, as Big Tech has committed to drastically reducing carbon usage over the next decade.

AI data centers also require power management and distribution systems to ensure the efficient use of power and industrial-strength cooling solutions so that the heat from servers remains controlled as they process data.

Software engineering

Close to 50 percent of new code is now generated by AI. The technology helps developers code more than 50 percent faster.

Software engineering is embedded in many corporate operations, as well as in goods and services.

Customer services

Scores of companies already use AI chatbots for customer support, but GenAI brings it to a new level. For instance, by rapidly processing customer data and browsing histories, GenAI can identify product suggestions tailored to customer preferences.

Marketing and sales

GenAI can significantly improve the efficiency and efficacy of marketing material. It can reduce the time required for content drafting and ensure consistency for a uniform brand voice and writing style.

Research and development (R&D)

In manufacturing, GenAI can optimize designs, reducing costs in logistics as well as production and testing time.

In Biotech and Biopharma, GenAI can shorten trial phases by drafting scenarios and profile testing candidates, expediting drug development for complex disease.

How to invest

The adoption of GenAI is still in its infancy. We think the technology seems very promising, but a full realization of its potential will take time and require intensive management as well as regulation to address the challenges its adoption will present. As has almost always been the case in the past, investors may be overestimating what GenAI can deliver in the short term, but underestimating what it can do in the long term given the technology’s great promise.

Enablers selling AI equipment and software have been the clear beneficiaries of the new technology, their valuations having expanded markedly. We think portfolios would likely benefit from exposure to the infrastructure beneficiaries of GenAI, where the eventual spending may take a decade or more to arrive.

As for the GenAI adopters, investors should assess how the new technology is being implemented—to increase sales, reduce costs, or improve productivity—and keep an eye on the competition. If competitors are also using GenAI effectively, any competitive advantage may erode quickly.

For more details about the GenAI ecosystem, see the full report.

In an upcoming Special Report, we will look more closely at the specific impact we believe the technology will have on a wide array of industries and how it will likely create new ones, emphasizing potential opportunities for investors.