From the studio

Video:Investors Club | your questions answered on gold, the Fed, and semis (9 mins)
Podcast: Across the Pond | Europe's overlooked innovators, on Apple and Spotify (17 mins)

Thought of the day

While the trigger for recent market swings has varied—from Iran and geopolitics to the Federal Reserve, valuations, or leverage—large-cap tech stocks have consistently remained at the center of market moves. This has been evident not only in broad index moves, as with the Philadelphia Semi and Kospi indices, but also in the sharp rotations seen within the AI value chain and tech sectors.

Alongside Iran this week, another risk emerged in the form of Oracle’s latest AI spending plans, reviving the broader debate over the sustainability of AI infrastructure capex. The company’s above-consensus capex guidance, to be partly funded through equity issuance, contributed to an 8.5% share price decline overnight. The move follows similar issuance from Alphabet, and has added to speculation that other listed hyperscalers could do the same.

Without taking any single-stock views, we think that the pickup in AI sector volatility may endure in the near term. But, we think investors should not see this as a signal of rising risks to the wider tech and AI investment cycle:

Compute demand isn’t slowing down. While investors are debating whether token costs could curb AI adoption, the underlying trends still point to firm demand. Token prices have been falling structurally for some time, so this itself is not a risk signal. Newer frontier models tend to complete tasks with fewer tokens than predecessor models, supporting broader usage. Meanwhile, hourly GPU prices are up nearly 25% this year and have reached a 19-month-high, according to Silicon Data’s H100 Index. If anything, we see the shift toward agentic AI as likely to increase token usage from here.

Funding jitters are not fatal. Concerns around free cash flow, dilution, and return on invested capital are legitimate and warrant monitoring, especially for companies relying on external funding. But we believe those concerns matter more for stock selection than for the durability of the theme itself. We forecast AI capex at USD 821bn in 2026 and USD 986bn in 2027. The renewed capital-raising push confirms the industry is still firmly in buildout mode, and will be for longer.

“Picks and shovels” look stronger. Recent market action suggests investors are no longer willing to pay for everything equally, with the bar now higher for prior platform leaders. As the industry moves into the agentic era, we think the next phase of spending is likely to be distributed more widely across semiconductors, memory, optics, power-related components, and semiconductor capital equipment, all of which can offer more demand visibility, pricing power, and earnings support.

So, we remain constructive on the AI theme, but we think this is a phase that calls for a more selective and diversified approach across the enabling, intelligence, and application layers. Alongside reducing excessive single-stock concentration, investors may consider tilting some exposure to semiconductors, infrastructure, memory, optics, and semiconductor capital equipment, where the link to current spending remains clearer.

Stepping back, AI remains one of the three pillars of our transformational innovation approach. We continue to recommend diversified exposure across these TRIO themes, which also include Longevity and Power and resources. Longevity, in particular, has shown diversification benefits for portfolios with significant exposure to growth-sensitive technology companies.