Boom or bust: navigating AI in the markets

Artificial intelligence (AI) themed shares are still a good bet even as the sector goes on a $660 billion spending spree, managers reasoned at the Collaborative Exchange’s Investment Forum held in Cape Town and Johannesburg this week and last.

But investment experts warn it may be a bumpy ride as markets discover who the winners and losers in the AI revolution will be.

Global managers discussing investing beyond the seven technology stocks known as the Magnificent Seven, said they remained invested in many of these stocks but were diversifying into other shares expected to benefit from the boom.

A large Boston-based manager, Natixis Investment Managers International, owns six of the Magnificent Seven stocks because they still trade at discounts to their intrinsic value, Jonathan Copper, senior investment specialist at the manager, said.

These companies are durable, still have competitive advantages and the ability to compound free cash flow over the long term, he said.

Bubble or no bubble

However, managers took different views on the risks of being exposed to an AI investment bubble.

Copper said he is concerned that investors are crowding into the sector and overestimate how many companies will actually profit.

Sebastian Mullins, head of multi-asset and fixed income for Schroders, a large London-based manager, said AI shares are not trading at investment bubble levels. He noted that the average price-to-earnings (p:e) ratio of the Magnificent Seven stocks (27 times p:e) remains well below the dot-com bubble (52 times p:e) or Japan’s late-eighties bubble (60 times p:e).

Mullins warned against indiscriminately buying “everything tech” and stressed the importance of identifying which companies can continue to deliver earnings, particularly given the huge capital expenditures and rising debt levels across the sector. Companies termed hyperscalers have historically held more cash than debt, but some now find themselves with less net cash or even in debt.

Investors, Mullins advised, must analyse each company to determine which can sustain higher levels of debt and which are overextended. He pointed out that earnings revisions outside the US are accelerating, indicating that opportunities are becoming more global.

Investors have to think about which companies are going to win, who has better technology, who can pay off their debt, who has strong free cash flow, and who's using AI best to actually generate income, he emphasised. Active management matters more than ever, Mullins argued.

Strong earnings

Earnings growth in tech, particularly in semiconductors, is strong but investors need to be selective given the wide dispersion in AI-related performance, Natasha Sarkaria, a director at BlackRock, agreed.

Sarkaria said the tech sector had the strongest growth of nearly 20 percent in the fourth quarter last year, but the spread between best and worst performing sub-sectors is already 50 percent this year.

 She also noted the importance of diversifying exposure to US tech shares, suggesting that China’s tech stocks - while trading at a steep discount to US counterparts - offered compelling value and solid earnings.

Structural shifts that could trigger disruption

More than just diversification, the Chinese tech sector could trigger major financial market dislocations, Dr Michael Power, a financial consultant with Kazkazi Consulting, warned at the conference.

Power, who previously consulted to local manager Ninety One, said rapid, structural shifts in AI could trigger shocks, such as that which wiped hundreds of billions of dollars off major US tech stocks in a single day when China released its AI model DeepSeek in January 2025. He warned that an even larger shock may be imminent as China releases new AI models.

Power said the US and China are competing with two very different approaches to AI: the US is focussed on training AI while China is focussed on drawing inferences.

He said Chinese developers were collaborating openly with a fast, low‑cost model that has been adopted by many countries particularly in the global south. This is a direct challenge to AI in the US, which relies on expensive hardware, massive data‑centre buildouts and financing structures that could prove unsustainable in future, he said.

Power warned that many AI‑related investments in the US may be more vulnerable than markets assume, particularly given the extreme capital intensity, rising debt and energy constraints there.

Rather than defaulting to traditional tech giants, investors need to monitor emerging AI players, prepare for significant volatility and recognise that the next major repricing of technology may be driven not by the hyperscalers but by genuine shifts in where AI capability resides.

Is it a classic bubble?

One local manager at the conference had a more direct warning for investors banking heavily on US-based tech stocks, saying they were showing classic signs of a financial bubble.

Iain Power, chief investment officer of leading boutique manager Truffle, said investors may once again be ignoring the signs of another “inflection bubble” like the dot-com and Japanese bubbles of the past.

Iain Power explained that inflection bubbles emerge when a genuinely revolutionary technology captures the public imagination, unleashing vast investment long before the economic payoff is clear.

During the dot‑com era companies traded at extreme valuations and overbuilt internet capacity. Telecom firms laid 145 million kilometres of fibre, only for 95 percent of it to remain unused a year after the crash in the early 2000s, he said.

A similar phenomenon is occurring with AI, he argues. In addition to revolutionary technology, the AI boom has three things in common with other inflection bubbles:

• Hype at extreme levels

The powerful narrative has seen share prices of software giants “obliterated” in the market as investors fear AI will cannibalise their business models.

• Unprecedent capital expenditure

Hyperscalers, such as Amazon, Alphabet and Meta, are spending more on data centres and graphics processing units than they are expected to generate in profits.  Iain Power argues that while the excess fibre laid during the dot-com bubble had a 30-year lifespan and was eventually useful, AI chips last three to five years only. This raises the risk of poor long‑term returns, he says.

• Irrational behaviour and capital raising

Billions are being raised for companies without solid products, revenue or commercial plans, Iain Power argues.

Unlike Copper, Iain Power thinks the $4 trillion Nvidia could potentially trigger a market correction. The company holds 80 percent of the AI chip market, but companies like Google and Amazon are rapidly developing cheaper in‑house chips, he says.

Nvidia represents seven percent of the S&P 500. A steep fall in market share could trigger a massive share price drop with huge consequences for large global indices.

Power says AI now accounts for 75 percent of S&P 500 performance and the market is pricing in a flawless future, ignoring the lessons of past inflection bubbles.

Like the internet, AI will change the world, but the investment cycle may again be ahead of itself with investors overpaying for the benefits of a technology that may take years to materialise.

Like the global managers at the Investment Forum, Iain Power recommends looking for stock-specific AI winners rather than banking on the sector broadly.