CIO Insights: What’s next for investing in AI?

10 minute read

Investor enthusiasm over AI has been a major force behind the rally in equities over the past couple of years. While enthusiasm had cooled in recent months, optimism over economic prospects under US president-elect Donald Trump has reignited interest in risk assets – including AI-focused tech companies like the Magnificent Seven (Mag-7).

But to understand what’s next for AI, it’s helpful to reflect on the past, to better understand the bigger picture and our place in the current cycle. In this month’s CIO Insights, we examine the evolution of other breakthrough technologies to gain perspective on where we stand today, and the investment implications going forward. 

Key takeaways:

• The rally in AI-related stocks has been driven by the fundamentals, but there are questions about how much further it can go. The gains in the tech sector since late-2022 – particularly for AI-related mega-cap firms – has been underpinned by robust earnings and ongoing capital investment. That said, there are still lingering investor concerns over an “AI bubble”. In particular, there are still questions over viable business models that can justify the billions of dollars invested into building the infrastructure to develop next-generation large language models (LLMs).
• AI excitement is nothing new: the current boom is the third cycle since the early days of AI in the 1950s, and is riding on breakthroughs in computing. From each of these cycles, breakthroughs have emerged that propelled the next development. The latest cycle started in the 2000s with the emergence of deep learning, and researchers believe that the current cycle will lead to Artificial General Intelligence (AGI) in the near future. Investments into AI infrastructure will likely continue, unless what’s known as the “scaling law” is proven wrong.
• Drawing parallels to the Internet era, the next beneficiaries of the AI era may have yet to appear. Looking forward, the next wave of AI beneficiaries may very well go beyond the current Mag-7 and semiconductor companies – any that are able to scale using AI technologies. These can include drug companies, autonomous vehicle networks, and a plethora of “personalised content producers”.
• Given the high level of uncertainty surrounding AI’s use cases, staying invested in a diversified index is one way to harness its investment potential. As the applications of AI evolve, we believe staying invested in a broad, passive market index is a more prudent strategy than trying to pick the next AI winner. This can help capture the economic upside as the technology is adopted in a wider range of companies and sectors, while mitigating the risks of a fast-changing and uncertain landscape.

The rally in AI-related stocks has been underpinned by fundamentals

The results of the US election have reignited the so-called “Trump Trade”, reflecting investor optimism about the impact of the Trump administration’s pro-business stance and preference for deregulation. That’s contributed to a bounce in technology stocks and risk assets more broadly. (Read more on that here: CIO Update: What another Trump presidency could mean for your investments.)

But let’s zoom out: the rapid rise of AI-related stocks – led by the Magnificent Seven – really started following the launch of ChatGPT in late 2022. At its forefront was Nvidia, a company that has transformed from a niche graphics processing player into a household name and the face of the AI revolution. The drivers behind its rally have been two-fold:

1. AI-driven earnings growth: Companies at the forefront of AI development and implementation have seen tangible improvements to their bottom lines. This isn't just speculative fervour: for example, Nvidia's quarterly earnings have grown more than 10x since 4Q 2022, with its data centre segment contributing almost 90% of its $35 billion top-line sales in the most recent quarter. Such strong earnings growth has been reflected in its share price – shown in the chart below. And in its outlook, Nvidia has signalled continued robust demand from cloud providers in the coming quarters as these firms continue to build out data centres.

2. TINA (There Is No Alternative): Companies outside of the Mag-7 have delivered less spectacular earnings growth over the past 18 months or so, amid sluggish manufacturing growth and a high interest-rate environment. The difference was stark: as of 2Q 2024, the earnings of these seven mega-cap companies grew by 35% while the rest of the S&P 500 posted growth of just 6%.

But concerns remain about a shifting macro environment and returns on AI investments 

Despite expectations of a positive macroeconomic backdrop going forward and continued strong earnings results from major AI frontrunners, there are still lingering investor concerns about the sustainability of the AI-driven rally:

1. The macro environment is shifting: As the Fed continues to cut rates, paired with expected fiscal stimulus from the Trump administration, the growth gap between AI forerunners and the rest of the market may start to narrow. The latest estimates from J.P. Morgan indicate that this gap will close in 2025, with the rest of the S&P 500 expected to deliver earnings per share (EPS) growth of 14% versus 18% for the Mag-7, as illustrated in the chart below.

2. Returns on AI investment are in question: While AI adoption has been rapid among employees and consumers, and hyperscalers have announced multi-billion dollar capital expenditure plans to build AI data centres, there's a growing concern about the lack of clear revenue models. This echoes the "productivity paradox" observed during the early days of computer adoption, where massive investments in IT initially failed to produce measurable productivity gains. ‘AI’s $600B Question’, a recent post published by Sequoia, has called into question whether AI investments will ultimately be justified by financial returns – with a potential gap amounting to $600 billion.

While the first concern is more cyclical in nature, the second concern is more structural. To understand whether this pause is merely a pit stop in a longer race or the beginning of a more significant correction, let’s take a deeper dive into the history and likely future developments of generative AI.

Before ChatGPT: History gives a better understanding of the AI cycle

While more recent developments like ChatGPT have become synonymous with AI for many, it's crucial to understand that AI has a rich and complex history dating back to the 1950s. The field has experienced several boom-bust cycles, each driven by breakthroughs and subsequent realisations of limitations. In short, we’ve seen the following cycles play out since the field’s inception in the 1950s:

1950s–1960s: The birth of AI and initial optimism

1970s: The first "AI winter" as early promises failed to materialise

1980s: Expert systems revive interest

1990s: Another "AI winter" sets in

2000s–present: The deep learning revolution

This historical context is crucial for understanding the current AI landscape. It reminds us that while the potential of AI is immense, progress is rarely linear, and setbacks are part of the journey.

The current boom in AI has been riding on the back of computing breakthroughs

The present AI boom, which began in the early 2000s, is fundamentally different from its predecessors. The key driver this time around has been the emergence of deep learning techniques (our Glossary below explains) which have dramatically improved AI capabilities across various domains.

Central to this boom is what is known as the “scaling law” – a theory suggesting that increasing model size, dataset size, and computational power leads to predictable improvements in AI performance. This principle has become the north star, guiding investments in the AI sector. 

In short, companies and governments are pouring billions of dollars into building larger language models and more powerful computing infrastructure, with the belief that the scaling law will hold true and eventually lead to Artificial General Intelligence (AGI). 

While some optimistic estimates point to AGI emerging as soon as 2027, history has shown that that road could be bumpy. Recent comments from leading AI researchers suggest that we're approaching the limits of scaling up large language models through pre-training – the initial phase where models learn from vast datasets. But there's promising news: researchers have identified a new "scaling law" in the post-training phase, allowing for further, continuous improvements.

Where we see the next big developments in AI 

How should investors think about the next phase of AI development? While the semiconductor sector and the Mag-7 have been clear winners so far, we think there are other areas to consider.

That’s because pivotal breakthroughs in LLMs have enabled a host of advancements and applications in other fields. Here, we can use the IT research firm Gartner’s “hype cycle” – a graphical representation of the maturity, adoption, and social application of different technologies – as a guide.

While some segments of AI are still at their early stage of development (like multiagent systems or causal AI), there are technologies that are closer to providing scalable productivity enhancements. As an example, autonomous vehicles have seen rapid development in the past few months thanks to the use of generative AI, with Tesla slated to announce its robotaxi in the near future and Alphabet’s Waymo planning to further expand its footprint.

The next wave of AI beneficiaries? Companies that can use the technology to scale 

While initial excitement and investments into AI have been mostly focused on its infrastructure, the ultimate winners of the technology may eventually emerge from applications of the technology, similar to the Internet boom in the 2000s.

As Chase Coleman, the legendary founder of Tiger Global Management observed: less than 1% of the current global Internet market cap was founded in the two years following the release of Netscape Navigator, the browser that helped popularise the World Wide Web.

Many of today's mega-cap Internet platforms – Alphabet, Meta, Amazon – were neither founded nor in their infancy during the early days of Internet adoption. As the Internet era progressed, the marginal cost of distributing content became essentially free, enabling these platforms to scale rapidly. 

Now, with AI, we're seeing a similar revolution in content creation itself: the marginal cost of creating content is approaching zero. This opportunity spans a wide range of industries; just to name a few:

1. Pharmaceutical companies creating personalised drug regimens.
2. Technology firms offering AI-powered personal assistants.
3. Education companies providing adaptive, personalised learning experiences.
4. Entertainment companies generating customised content based on individual preferences.
5. Financial services firms offering hyper-personalised investment advice.

In our view, companies that can effectively use generative AI to create and sell personalised content have the potential to create significant economic – and shareholder – value. This could fuel the next bull market in much the same way that Internet platform companies drove the tech boom of the 2010s.

Tap into AI’s evolution by staying invested in a diversified index 

As we move into the next phase of AI development, the landscape may become even more uncertain. While it's tempting to try to pick the next AI winner, history suggests that to be a challenging task. Instead, we believe a more prudent approach might be to:

1. Diversify across the AI ecosystem, including hardware, software, and AI-enabled services.
2. Consider exposure to other sectors that could benefit from AI-driven productivity gains.
3. Invest in companies with strong AI research and development capabilities.
4. Keep an eye on emerging players that demonstrate novel applications of AI technology.

For most investors, the simplest and most effective way to capture any upside that AI development brings is through ETFs exposed to the broad equity market. As emerging AI leaders grow and succeed, they naturally gain larger weightage in passive indexes. Staying invested in an index lets you participate in their success while mitigating the risk of betting too heavily on any single company or sector.

For a more targeted approach, our Thematic Portfolios give you diversified exposure to companies at the forefront of innovation. These portfolios are designed to capture multi-year, structural trends with strong, long-term growth potential. And there's more on the horizon – we're refreshing our Thematic Portfolios and expanding the ETF line-up in our Flexible Portfolios to give you even more ways to capture tomorrow's opportunities. Stay tuned for these updates in the coming weeks!

Glossary

Generative AI and Large Language Models (LLMs)

AI systems that create new content (text, images, code) based on patterns learned from training data. LLMs specifically focus on processing and generating human language at massive scale.

Artificial General Intelligence (AGI)

A theoretical form of AI that could match human-level intelligence. Unlike today's specialised AI, AGI would be able to understand and solve a wide range of problems independently.

Deep learning

A technique using layered networks to process data, recognise patterns, and make decisions, much like how our brain uses networks of neurons. This technology is behind many recent AI breakthroughs.

Data centre

Large facilities that house hardware for the storage, processing, and distribution of large amounts of data. Data centres are crucial for training and running large AI models.

Hyperscalers

Large companies that provide cloud computing, networking, and internet services at scale. Examples include Amazon (AWS), Microsoft (Azure), and Google Cloud.