Winner takes all in the AI economy, and SA wasn’t invited

As markets charge ever higher, driven inexorably by artificial intelligence (AI) mania, an unusual dynamic is playing out.

The economic impact of past market booms, especially ones that change the course of history, has typically broadened as the new technology is adopted. Early winners and first adopters share the first spoils, after which reality sets in and a financial bubble turns to bust.

Then, laggards catch up, the process of structural transformation takes hold, and the economic process of diversification reasserts itself as the productivity benefit of the new technology spreads across the economy.

With AI, however, the opposite seems to be happening. The longer it runs, the more ruthlessly it is concentrating gains among an increasingly narrow set of beneficiaries, punishing everyone else in a zero-sum dynamic. The question is, how long will this go on for?

The relative performance of the world’s equity markets over the past year is an almost perfect illustration; they have been a function of where that country’s economy fits into the so-called AI technology “stack”.

Top dogs

At the top is the US – which dominates foundational LLM development – and China, whose homegrown AI ecosystem is catching up much faster than sceptics had expected.

Below them come the hardware-related economies; the poster children of Taiwan and South Korea, who are at the crest of the semiconductor craze. Japan and Israel too have exposure to a broad array of semiconductor and AI-related capabilities.

Secondary suppliers – Malaysia, Singapore, Vietnam and Thailand – have benefited from the insatiable demand for components such as servers, circuits and electronic components in the vast datacentre buildout.

Then there is everyone else. Much of Europe is in this bracket, except for the Netherlands, home to ASML, the indispensable supplier of chip lithography equipment.

India, for example, which has long been celebrated as an IT services juggernaut, finds itself on the wrong side of the AI disruption trade. As a beneficiary of the old outsourcing model, it is now seen as highly vulnerable to obsolescence, with companies like Infosys and Tata Consultancy Services in the firing line.

The Indian benchmark Nifty 50 is down almost 10% in the past year, while the South Korean Kospi is up 174%.

SA languishes

South Africa, sadly, sits squarely in this last group. The JSE has been pedestrian in comparison with the technology-heavy Asian emerging markets. The exception, albeit a partial one, is copper miners; this is seen as a secondary play on wiring-intensive data centre construction. But beyond that, South Africa’s listed companies offer little in the way that speaks to the AI buildout. It is a worrying reality for a country whose equity market is in desperate need of a growth narrative.

The scale of the divergence is striking even by historical standards. During the dotcom frenzy of the late 1990s – the closest parallel most investors can recall – the leading tech subsectors accounted for about 30% of index weighting and 60% of market gains at its peak.

This pales into comparison with what is happening now. AI-related companies now represent about 45% of the S&P 500’s market cap and have been responsible for about 62% of the market performance since mid-2024, according to Barclays data.

Even more worrying, unlike in the Dotcom era when tech-fuelled mania spilt across asset classes and sectors, today’s boom is precisely zero sum; capital flowing into AI is being sucked out of other sectors which are unceremoniously crushed. Examples are the “SaaSpocalypse” of software companies and “jobsageddon” in recruiting, media and financial services. Even in the US, real investment outside of technology is declining. Capital is not broadening, it is concentrating.

The macroeconomic effects of this are becoming apparent. GDP growth expectations have held up for AI-exposed economies such as the US and Taiwan, while for the peripheral economies they are falling, hurt, among other factors, by higher energy prices. For countries such as South Africa, which lacks the industrial base and energy infrastructure needed for AI investment, the boom might as well be on another planet.

Doubts creep in

What is less clear is how sustainable this whole dynamic is. Can the AI boom sustain itself if it is sucking life out of the broader economy? There are early signs, tentative – even premature – that this concentration momentum is losing steam and doubts are creeping in.

The first crack appeared in the so-called Magnificent 7. For the better part of five years, the performance of the S&P500 has been essentially a function of how a handful of mega-cap technology companies have done. Since mid-May, that relationship has broken down. The Mag 7 has foundered, yet the broader index has held up.

In their place, semiconductor companies have surged; investors have bet that these will be the natural beneficiaries of the enormous capital expenditure spends being made by the “hyperscalers” of Alphabet, Microsoft, Meta and Amazon. Whatever you might think the return on these investments turns out to be, the chip suppliers will collect their bounty regardless.

The ‘chip wreck’

But now the semiconductor boom is fading too. The Philadelphia Semiconductor Index has moved sideways since the beginning of June, while the Kospi Index in South Korea has had a torrid time in the past week, with chip heavyweights Samsung and SK Hynix crashing.

Dubbed the “chip wreck”, it ended last week roughly 12% lower. Instead, what has replaced chipmakers at the top of US market performance is a notably unglamorous collection of banks, industrials, staples, insurers and pharmaceutical companies – sectors with little obvious connection to the AI narrative. Whether they have the weight, or the earnings growth, to sustain the market if semiconductor enthusiasm truly evaporates is unclear.

The broader dynamic is like a relay race with each runner slower and more exhausted than the last. Since the pandemic, global markets have been animated by a rolling parade of speculative manias which have all petered out; cryptocurrencies, NFTs, meme stocks, precious metals, the Mag 7 and most recently semiconductors. Each has generated handsome returns for traders, who have traded their entry and exit points skilfully. Each then ran out of steam.

The technological revolution underpinning AI is real, and its economic consequences will almost certainly be profound and outlast the current speculative mania.

This has played out with other bubbles; for example, the great railroad boom of the 19th century and dotcom craze in the 2000s. Both collapsed, but in the end real structural change occurred, albeit after the investment frenzy had turned and many speculators had been washed out.

AI will probably follow a similar arc. Investors who remember how the last few manias ended will already be eyeing the exits. DM