What separates AI winners from AI exposure now?

The AI trade has driven most of the United States (US) equity market's gains since 2023. The "Magnificent Seven" technology stocks, most of them AI-linked hyperscalers and chipmakers, now account for close to a third of the S&P 500's total market capitalisation, and their combined capital expenditure on data centres and AI infrastructure exceeded $400 billion in 2025, a figure the International Energy Agency (IEA) projects will rise a further 75% in 2026.

That spending survived a software sector sell-off earlier this year that wiped hundreds of billions from software-as-a-service valuations. HSBC Private Bank emerged from the correction recommending clients widen AI positioning rather than narrow it, extending its guidance from chipmakers and software into infrastructure and energy, one of four stated priorities in its third-quarter 2026 outlook.

Willem Sels, global chief investment officer at HSBC Private Bank and Premier Wealth, argues the case for staying invested rests on earnings rather than sentiment. Since March 2023, S&P 500 earnings per share have risen roughly 30% on a rebased basis, technology earnings around 90% and a cohort Sels labels AI winners have more than tripled, while valuation multiples on US technology stocks have eased back over the same period. He has pushed back on suggestions the theme is overextended, arguing its productivity gains are only beginning to show up in earnings.

What does the earnings argument say about the rest of the market?

BlackRock's own research backs up the sector-level version of this argument, stating that most of the recent technology rally has come from earnings growth rather than rising multiples, and that current multiples are justified based on growth expectations.

HSBC and BlackRock's case is specifically about AI-linked names, not the index as a whole. Of note, most stocks in the S&P 500 are not keeping pace with the index itself. Northwestern Mutual's own research shows that in May 2026, only 27% of S&P 500 companies outperformed the index, describing market leadership as narrow.

Are private banks becoming more selective within the AI theme?

Barclays Private Bank has published research describing that shift. Its analysis, published in May 2026, argues the AI cycle has moved from a phase that rewarded almost any AI-linked exposure to one that separates companies already converting AI investment into earnings from those where the payoff remains delayed or uncertain. Barclays reports the performance gap between those two groups widened through 2025 and into early 2026.

HSBC's own guidance, recommending exposure into infrastructure and energy alongside chipmakers, echoes that move toward selectivity. Rather than distinguishing individual companies by earnings quality, HSBC encourages clients to spread exposure across layers of the AI supply chain, chips, data centres, electricity, models and applications, based on how much risk they want to take.

Why does HSBC treat AI, energy and defence as one investment cycle?

Sels points to the United States, China and Europe pursuing what amounts to the same policy priorities under different national labels. "They all want to win in AI, they all want to have energy independence, and unfortunately ultimately they all need to invest in defence as well," he said, arguing that when the three largest economic blocs compete on the same three fronts, it creates urgency, investment and capital flows regardless of which bloc is ahead at any given time.

That framing also extends to monetary policy. HSBC's own view is that the US Federal Reserve will hold interest rates steady through the rest of 2026, a call Sels linked partly to new Fed chair Kevin Warsh's belief, as Sels characterised it, that AI's productivity gains are ultimately deflationary, allowing the economy and labour market to strengthen without companies needing to pass higher costs on to consumers.

What should change in client conversations about AI?

HSBC treats AI investment as a standalone Q3 2026 priority, distinct from energy independence and defence spending, which it groups together as a second priority tied to the same government spending cycle.

The practical shift for wealth managers is in where the burden of proof now sits. Holding any AI-linked name may have been enough to capture returns through most of the past two years. Barclays' research suggests that is no longer true at the company level, where earnings conversion now separates winners from laggards. Northwestern Mutual's data suggests the broader index has not caught up to the AI names driving it either. Advisers reviewing AI exposure need to take a nuanced approach, distinguishing holdings that are actually converting AI investment into earnings from those merely riding the market's optimism.