Artificial intelligence (AI) is displaying textbook characteristics of General Purpose Technologies (GPTs) seen in previous industrial revolutions over the past 300 years. Pervasive, dynamic, and capable of spawning entirely new industries, GPTs have significantly changed the landscape over time. However, while factors such as connectivity, early-stage investment and faster deployment are creating a shrinking time scale and expanding opportunity, the bulk of capital currently flowing into the sector is being captured at the infrastructure layer of the value chain. 

While this is typical of the early stage of adoption, it does not reflect where value is likely to migrate as adoption accelerates and the industry matures.

This is according to presenters at Old Mutual Investment Group's latest quarterly market update, who outlined why today's AI infrastructure boom, while a rational and even necessary phase of the technology's development, is unlikely to be where investors see the strongest long-term returns.

The comments come against a backdrop of markets broadening out well beyond the US mega-cap technology names that have dominated returns in recent years. “Over the past 12 to 18 months, Magnificent 7 and broader US leadership have given ground to a wider set of AI players, with Korea and Taiwan posting exceptional gains off the back of their semiconductor industries, alongside other emerging markets such as Brazil, Japan and South Africa,” explained Portfolio Manager John Orford. “Technology still dominates returns, but increasingly in different guises, across more markets and business models than the handful of US hyperscalers that led the prior cycle.”

That broadening is itself a function of the AI infrastructure build-out, with Korea and Taiwan's gains substantially driven by exports feeding US demand for AI-related hardware. 

“It's this build-out – now one of the largest in history – that we’re watching most closely, not because of the scale of the spending on the build-out, but because of what markets are now assuming about the earnings that are needed to justify the spend. Earnings could be where the bubble lies, with lofty earnings expectations built into current valuations," Orford says. 

This raises the question of how value is created in the AI sector and where is the opportunity set to be found as we move rapidly forward. 

While AI and its associated innovation spawn isn’t slowing down anytime soon, this could be a double-edged sword. Unpacking research on the current AI value chain and how this is likely to evolve in the future, Portfolio Manager Zain Wilson pointed out that rapid capability gains and falling costs are widening AI's addressable market and pulling in more competitors and more capital – a pattern typical of GPTs at this stage of their cycle. 

“However, the same over-investment and intensifying competition also mean money committed to infrastructure today may not generate the best returns for investors.”

Wilson explains that he expects value to migrate up the chain over time:

• Away from hardware and data centres, which face cyclical rather than structural bottlenecks.
• Closer to enterprise applications and products that solve specific customer problems.
• Toward businesses able to close the gap between AI's raw capability and demonstrated customer value, rather than those simply reselling computation by the token.

"While this migration of the value chain attracts significant capital and competition, it erodes returns and prices over time, and the market ultimately consolidates into a few large winners as opportunity shifts to secondary innovations and products," he explained. Prior competitive advantages can become vulnerable in this phase – which is precisely why disciplined positioning matters more than momentum-chasing at this point in the cycle, he points out.

Turning to the broader macro and portfolio implications, Orford outlined that the AI-driven investment cycle is unfolding against a US economy that is late-cycle but still resilient, supported by easy financial conditions. "There are some concerns around inflation, and that is pushing up bond yields. That poses a risk should central banks have to raise rates more than currently expected," he said.

Domestically, he pointed to improving fundamentals – continued structural reform progress, fiscal consolidation and a firming growth path – against upcoming local government elections as a near-term watchpoint. "South African asset valuations in equities remain attractive on a relative basis, and we could see a re-rating if the cyclical story around oil prices improves," he added.

This view is reflected directly in portfolio positioning. Old Mutual's Balanced Fund is running slightly underweight global growth assets, primarily global equity, reflecting stretched valuations and a late-stage global cycle that warrants some risk being taken off the table. The fund is modestly overweight South African fixed income, property and bonds, while diversifying its defensive holdings away from global bonds – where sovereign, fiscal and inflation risks are elevated – toward a broader basket of cash, African fixed income and gold.

Within South African equities, the Fund's overweight sits in resources excluding gold rather than domestically-linked cyclical shares – a position Wilson linked directly back to the AI theme. "That overweight isn't being driven by the South African economy. It's driven by more than a decade of under-investment in resources, and by demand vectors including energy security, the copper intensity of AI-driven electrification, and rising capital spend on defence and the energy transition," he said.

"AI's long-term disruptive potential isn't in question. What is in question is who captures the value, and when," Wilson concluded. "Investors who chase today's infrastructure spend may be early to the theme and late to the returns."