The AI trade is crowded, pushing legacy software back into favour

Last year there was a surprisingly narrow rally in global markets. According to Alex Tedder, group chief investment officer for equities at Schroders, more than 50% of the total return of the S&P 500 came from just seven companies.

However, there is a widening performance gap. Tedder noted that only Nvidia and Google remained in that top-performing group, while the other five members of the “Magnificent Seven” lagged behind. Even among the world’s most powerful stocks, the winners’ circle is getting smaller.

This market concentration is driven by what Nick Balkin, chief investment officer at Foord Asset Management, described as a “fear of missing out”. Speaking at a Nedgroup Investments event, Balkin noted that cash holdings among traditional global investors have dropped to their lowest levels since 2007 – a statistic he interprets as a signal that the market is effectively “all in”.

What is the Magnificent Seven?
The Magnificent Seven is a nickname for the elite group of US tech giants – Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla – that have dominated the stock market in recent years. These companies represent roughly 35% of the S&P 500’s total value, giving them enough power to pull the entire market up or down.

Read more: Big Tech's AI ambitions strained as investors weigh high spending against stock performance

The Magnificent Seven are also responsible for a dominant share of global capital expenditure, Tedder said. Essentially, the same companies driving market returns are also spending on an unprecedented level to build AI infrastructure.

As scepticism grows over AI’s profitability, fund managers are moving their focus to legacy software like SAP and Salesforce. These entrenched systems are often cheaper and too critical to be replaced by new, unproven technology.

ROI reality check

One of the clearest fault lines runs through the question of return on investment. Historically, Tedder said, the technology companies that spend the most have often gone on to perform the worst in subsequent years.

Fund manager at Truffle Asset Management, Saul Miller, said that large language models such as ChatGPT and Gemini are becoming commoditised, a process that makes it difficult to justify valuations, even if use continues to grow.

Miller described a “circularity that adds risk to the system”, specifically citing the relationship between chip manufacturers and AI companies. “When Open AI buys chips from Nvidia, one of the ways that they do that is (by) issuing equity, because they can’t pay for it.”

Read more: Hard refresh – the truth behind the push for you to get a new computer

He warned that this structure relies heavily on the value of that equity. “If that equity is overvalued, then maybe there’s a bit of a hole in some of Nvidia’s earnings,” Miller noted.

To justify current US equity ratings, Balkin said companies would need to deliver real earnings growth of 5.5% annually, a figure he described as a “hard base” to achieve compared with the 100-year historical average of just 2.5%.

Flight to reliability

Investors are relying on the digital bedrock of the global economy. Andrew Headley, global fund manager at Veritas Asset Management, said his fund is building positions in what he called legacy software companies such as SAP and Salesforce.

He said that fears of wholesale AI replacement misunderstand how deeply these enterprise resource planning systems are embedded into business operations.

Headley explained that, unlike creative AI models, corporate financial controls require “five nines reliability” – or 99.999% accuracy – to satisfy auditors and regulators.

Because these systems feed directly into compliance and financial statements, he dismissed the idea of a near-term swap, predicting that AI won’t displace these core systems “certainly not in the next five years, probably not in the next 20 years”.

While AI-linked stocks are trading in a state of euphoria, legacy software firms are “insanely cheap”, Headley said, with free cash flow yields between 4% and 7.5%. Those yields sit alongside projected free cash flow growth of 10% to 20% a year, a combination he regards as mispriced risk.

Benjamin Alt, head of global private equity portfolios at Schroders, echoed this view from the unlisted side of the market.

He said that small and mid-market clients “extremely value the kind of human interaction, having the feeling that someone listens to them, understands the problem, then acts in a way customised”. This means AI agents are unlikely to replace specialised software overnight.

Read more: AI won’t replace you – but it will redefine what makes you valuable at work

How this affects you
🤖 For most South Africans with retirement savings, AI exposure is already built into global funds as US technology shares dominate major indices;
🤖 This high concentration means portfolios are vulnerable if a few tech giants stumble;
🤖 Moving focus to legacy software may offer more stable cash flow for long-term capital preservation; and
🤖 The huge energy needs of AI infrastructure could create a tailwind for South African mining, as global demand for copper and platinum increases.

The infrastructure play

For investors wary of betting on a single AI winner, looking to the infrastructure and service layers that support the entire AI ecosystem provides an alternative.

“The supply chain is massive and it goes through tech, to industrials, to services, to materials – the whole spectrum,” Tedder said, highlighting how investors can gain exposure to the theme without facing concentration risk.

Alt described a similar approach favouring the service provider layer – companies that help businesses implement AI regardless of which underlying system dominates.

Read more: The great schism — Where does AI go from here?

“It’s not necessarily important who is the winner,” he explained, noting that by investing in facilitators that work with “almost all the language models”, investors can bypass the risk of backing the wrong horse while still capitalising on widespread adoption.

The cloud and copper

In South Africa, the most direct line to AI returns may lie in the ground (literally) rather than the cloud. Miller pointed out that the “incredibly energy-hungry” nature of AI data centres has created a structural deficit for copper, which is essential for both the centres and the electrical grids powering them.

Rising costs for new mine shafts are pushing mining companies towards mergers and acquisitions to secure supply, Miller said. This creates a potential opening for South Africa, where major mining groups could become targets for consolidation within the copper sector.

Read more: S&P sees looming copper shortages posing ‘systemic risk’ to global economy

Market momentum is performing at its highest since the dotcom bubble, Headley said, driven by the “insane performance” of AI stocks since late 2022.

While the long-term outlook for AI remains positive, Tedder said, the path “is not going to be a straight line”. Between geopolitical risks and execution hurdles, investors may find more reliable safety in the physical supply chains and the legacy software that remains the true brain of global business. DM