The Decade Ahead What will generative AI mean for markets and economies?

Contribution to 15.1% S&P 500 performance
Magnificent 7 vs. remaining 493 S&P 500 stocks, year-to-date, in percentage points

Generative AI is the latest technology creating value across sectors
The innovation value chain: Breakthrough technology is powered by infrastructure and inputs, creating new hardware running on platforms of operators and enablers, benefiting the economy at large

Infrastructure and input providers. Demand for necessary infrastructure and inputs understandably often booms in the immediate aftermath of innovation—think steel and the railway, oil and the automobile, semiconductors and the smartphone. But history shows that while providers of key inputs can earn high profits initially, over time their products can become commoditized. Infrastructure and input providers therefore may eventually need to learn to run at a high scale and with lower profit margins.

Hardware manufacturers. Innovative technologies can often drive the adoption of new consumer hardware (e.g., cars, TVs), and hardware manufacturers can therefore benefit from a surge in demand. At the early stages of a technology boom, hardware manufacturers may benefit from the uniqueness, quality, and novelty of their hardware. But sustaining leadership is often harder. Hardware can also become commoditized over time, and successful hardware companies often must learn to differentiate their products, potentially by integrating hardware with a leading operating system.

Operators and enablers. The largest and most enduring value creation derived from new technologies has historically tended to accrue to their operators and enablers. Railroad companies, radio and TV networks, software operating system developers, and digital platform companies are all examples of such enablers. In many cases, they have become the largest companies in the world at some point in time.

Application beneficiaries. Innovative technologies often create ecosystems of beneficiaries that may not be directly involved in the development of a technology, but may be well positioned to use them to build new businesses or improve the profitability of existing ones. Companies that increased cross-country trade following the advent of the steam engine, retailers that boomed due to mass car adoption, or businesses that engage in e-commerce and social media as a result of the internet are all examples.

We can see this historical framework at play in the early stages of the AI revolution:

Infrastructure and input providers (cloud). Cloud computing is a crucial input for generative AI because it provides the computing power needed to both train and make generative AI applications work. Much like historical “inputs” into other key technologies, cloud computing is in some sense commoditized––it is a standardized service with a low unit cost. That means there are significant economies of scale, and the largest cloud platforms are already run by the largest technology companies. Cloud providers are also able to “lock in” customers through bundled services.

Hardware manufacturers (GPU manufacturers). Graphics processing units (GPUs) are chips that are crucial to the training of neural networks, a cornerstone of AI, and are therefore seeing a strong surge and significant share price appreciation for the leading manufacturers. In the short term, we expect this demand surge to continue. In the medium term, we may see a period of “digestion” in which buyers identify and refocus on only the most promising use cases. Over the longer term, it remains to be seen whether chipmakers can continue to develop, evolve, and innovate their products to maintain pricing power, or if GPUs will become commoditized, forcing manufacturers to run at a higher scale and with lower margins.

Operators and enablers (large language models). Large language models (LLMs) can be thought of as among the key “enablers” of the AI ecosystem. Developing an effective large language model requires significant scale in data, computing power, and talent, and while models may be better suited to a specific application (e.g., generating text, code, or images), they are generally not specific to an individual domain (e.g., finance, law, or marketing). Ultimately, there may only be a few LLMs on which most generative AI applications are built. So far, the most notable have mostly been built by, or in a joint venture with, the largest technology companies.

Application beneficiaries (text generation, programming, image/video generation). ChatGPT has been a clear example of how generative AI can exhibit human-like text-generation capabilities. It can also be used to generate computer code, images, or video. We see significant opportunities over the next few quarters in the integration of AI “copilots” in office productivity software, in the rising demand for AI analytics, and in AI integration in image, video, and other enterprise applications.

An unusual feature of generative AI is that, right from the onset of the new technology, many of the same companies are already operating in multiple stages of the value chain––from cloud, to the ownership of LLMs, to the development of end-user applications.

With that in mind, it is perhaps understandable why the Magnificent 7 in the S&P 500, mostly AI beneficiaries, have seen their market capitalization grow by 67% (or USD 4.6 trillion) so far in 2023. With the significant resources needed to build and benefit from complex AI models, we expect the large players to grow larger still.

We believe that investors looking for exposure to AI should seek broad exposure across the value chain, including in cloud, semiconductor, software, and internet names. Related semiconductor companies should continue to see robust demand in the near term; key generative AI product launches across many of the Magnificent 7 are likely to keep the momentum high; and internet stocks should benefit as AI becomes more integrated in consumer applications like gaming, entertainment, and advertising.

While the long-term growth potential is large, investors should be prepared for potential short-term volatility or drawdowns. As with other technology booms, an initial surge in demand can often be followed by a digestion period for consumers and businesses. For long-term investors, such periods could present attractive entry points to increase exposure.

We think AI should meaningfully improve efficiencies and worker productivity, but the implications for economic growth are less clear. Depending on how AI is ultimately applied, it may result in unchanged output but increased leisure time.

Some jobs will become obsolete because of AI—few offices today have “typing pools,” for example—but we should not necessarily expect a surge in unemployment. AI is likely to create at least some new jobs that were not previously thought of. Historically, roughly 10% of the labor roles that existed at the end of any given decade did not exist at its start. For example, employment in the entertainment industry has increased over the past decade as social media and streaming increased consumption of entertainment and reduced barriers to entry.

Over the longer term, AI will likely be disinflationary for prices in some sectors, but the extent of the economy-wide impact will depend heavily on the magnitude, location, and timing of AI’s use. It is already becoming clearer how AI text, image, and video generation could put downward pressure on pricing for sectors including customer services, computer programming, legal, and entertainment.

Finally, periods of economic upheaval tend to create social tensions as those who see their income and status decline seek someone or something to blame. These tensions can lead to populist and prejudice politics. AI could also further escalate geopolitical tensions as countries enter an AI “arms race” as a defense against the potential applications of AI by their rivals. This trend is already evident in the recent restrictions on AI technology introduced by the US, and other retaliatory measures by China.