AI: disruption, opportunity
and portfolio implications

History offers few examples of truly transformational technologies. Electricity reshaped the physical world, and the internet redefined how we communicate and conduct business. Many now argue that AI is the next such general-purpose technology – ‘thinking and doing on demand’ – just as electricity offered energy on demand.

AI is no longer confined to niche applications. It is becoming a key infrastructure layer supporting productivity across industries, offering solutions from self-optimising supply chains to customer service automation, code writing and drug discovery. In this context, AI resembles a new industrial revolution. Cloud data centres and inference engines could well be the electricity grids and steam engines of the 21st century.

Real-world impact

Major corporations are already weaving AI into the fabric of their operations. Examples include:

• Amazon integrates AI across personalised shopping, warehouse logistics and drone delivery. We particularly like how Amazon monetises both AI enablement (via selling data centre capacity to others through AWS) and adoption (via enhanced robotics in its warehouses and stock planning in its retail business).
• Microsoft is using Copilot to accelerate software development and automate internal functions such as IT and HR – 35% of its code is now AI-generated.
• Walmart uses AI to enhance customer support while optimising delivery routes and store layouts with predictive algorithms.

These cases highlight that AI isn’t simply automating isolated tasks – it is rearranging entire workflows, improving employee efficiency, and enabling faster, data-driven decision-making. AI could well change the entire corporate cost structure by reducing the number of people required to run complex organisations.

Complex economics

AI’s core economic promise is productivity – growth without inflation – by enabling employees to accomplish more in less time. (In my own case, it has boosted weekly output by roughly 15% to date.) The unresolved question is how those gains will ultimately be divided between companies, consumers and employees. We suspect that employees may find themselves last in line.

The economics are complex. Building and training large language models (LLMs) requires billions in upfront capital, and many market leaders are still burning through venture capital with no guarantee of profitability. Meanwhile, model performance is beginning to converge, increasing the risk of commoditisation. Despite this, valuations keep rising, raising concerns about a potential speculative bubble.

Businesses at risk

While AI’s great economic lure is its capacity to drive efficiency, it also poses a direct threat to business models built on low-value cognitive labour, repetitive processes, or scale alone. Companies that fail to adapt risk accelerated competition and early disruption. Even global leaders with clear AI advantages, such as Google, face challenges as LLMs threaten to bypass traditional search behaviour.

The highest risk probably lies with organisations that fail to integrate AI into their operating model, or worse, choose to ignore it altogether. In the age of rapid adoption, denial is not a defensive strategy – it’s an exit strategy.

Capital and competition

AI is in the midst of an investment surge rarely seen in any sector. Just four private model developers – OpenAI, Anthropic, xAI and Perplexity – have collectively raised roughly US$100 billion, implying average price-to-sales multiples near 40 times, compared with around five times for today’s megacap tech giants.

Meanwhile, hyperscalers like Microsoft, Amazon and Google are pouring hundreds of billions into AI infrastructure annually. The open question: will that capital deliver the promised returns? Although management teams are enthusiastic, we are watching closely for any early signs of demand slowing.

One potential ‘canary in the coalmine’ could be declining profitability among infrastructure providers like CoreWeave, which may indicate that downstream AI businesses aren’t yet generating the returns needed to sustain the ecosystem.

If growth expectations aren’t met, a cascade of negative sentiment could hit both valuations and reinvestment. For now, however, capital expenditure guidance continues to rise.

Lessons from history

History counsels caution. Technologies with long feedback loops and heavy capital requirements often spark waves of overinvestment, followed by harsh corrections – such as the railroads of the 1840s and the dot-coms of the early 2000s.

But disruption also creates outsized winners. Amazon took 27 quarters to post a profit, burning billions along the way, before becoming a US$2 trillion company. Apple, Tesla and Google all endured near-death moments before reaching trillion-dollar status.

For every Amazon, there are dozens of Pets.coms. That’s why we favour quality businesses that are early adopters of AI yet possess the balance-sheet strength to survive missteps – in our view, this offers a more durable path for investors.

Positioning portfolios

Bearing all the above in mind, how should investment portfolios be positioned? The largest portfolio risk today may be underexposure to AI. In recent years, avoiding AI-related stocks would have hurt performance more than holding a few poor performers.

That said, valuation discipline remains critical. Over time, equity returns are driven by returns on capital and the price paid for future cash flows. Even a great business will only deliver outsized returns if its purchase price embeds expectations lower than the eventual reality.

Given that many of today’s AI beneficiaries are already well-run, highly profitable enterprises, investors need not chase loss-making businesses still burning through capital. We believe the most promising opportunities lie in businesses where:

• AI can materially enhance operations – whether by unlocking new revenue streams or delivering cost efficiencies – particularly where the scale of these opportunities is expanding and remains underappreciated by the market
• Pricing power enables efficiency gains without these being fully passed on to customers
• Competitive moats remain defensible in an AI-driven landscape.

In our view, some of the most compelling opportunities lie in the following categories of firms, illustrated with examples from our global portfolios:

• Core AI beneficiaries (e.g., Microsoft, Prosus, TSMC)
• Industry leaders integrating AI (e.g., Visa, Yum! Brands, Thermo Fisher)
• Disruptive AI-native firms (e.g., Harvey, though almost all remain privately held for now).

We believe that companies with scale advantages and proprietary data are best positioned to capitalise on AI. Incumbents with entrenched customer bases and the capital to invest are extending their leads – for now. We anticipate that genuine disruption of mainstream leaders is still several years away. Beyond ChatGPT, AI’s definitive ‘killer app’ has yet to emerge – much as Facebook only entered the mainstream in 2008, nearly a decade after the internet first reached 100 million users.

However, the democratising nature of AI means that today’s leaders are not assured of tomorrow’s dominance. New entrants will emerge, and AI-native firms will ultimately reinvent what incumbents can only retrofit – making vigilance essential.

A last word

AI is not mere hype – it is reshaping industries, business models and the nature of work itself. For investors, the biggest risks are complacency and misallocation: either ignoring the change or overpaying for speculative plays. With AI set to benefit a wide range of businesses, there is no need to take excessive risk to participate.

In the end, as with electricity and the internet, AI will eventually become invisible – ubiquitous and indispensable. The moment to prepare is not when that happens, but now.