From the studio

Podcast: Signal over Noise | What US history tells us about AI capex, on Apple and Spotify (8 mins)
Podcast: Jump Start | US-Iran latest, jobs data, and tech volatility (4 mins)
Video: Investors Club | Oil, US dollar, AI cycle, and China tech (10 mins)
Video:Top of Mind in Asia Pacific | Min Lan Tan: Asia's second-half outlook (7 mins)

Thought of the day

Global equities snapped their two-week gains last week amid renewed concerns over the sustainability of AI capex growth and whether it can continue to fuel the semiconductor complex that has rallied strongly this year. The Philadelphia Semiconductor index fell nearly 8%, underperforming the MSCI All Country World’s 2% loss.

We still favor the “picks and shovels” of the AI buildout in our tactical positioning, as demand visibility, pricing power, and earnings momentum remain strong. But recent market volatility has also pointed to certain risks worth monitoring.

We offer our thinking on the AI trade and discuss how investors should position.

Risks of slower capex growth have risen. The recent equity capital issuances by megacap tech companies suggest that hyperscalers remain in build-out mode, and we expect annual AI-related capex to rise to nearly USD 1tr next year. But the hyperscalers’ average 20% share price decline this month points to increasing shareholder pressure to justify their spending, and we acknowledge that the risk of slowing capex growth has risen at the margin. For example, Microsoft could slow the pace of its capex expansion by using lower-cost, open-source models for its AI tools, while Meta may potentially lease out part of its in-house GPU capacity to external customers to improve free cash flows. These moves could ease the current compute constraints, but raise questions around future capex spending.

Advances in lower-cost models may inject volatility, but growing demand should ultimately offset the concerns. The capability advances in lower-cost AI models have been rapid, with some models outperforming their frontier competitors in certain benchmarks at a fraction of the costs. Regulatory restrictions on certain frontier models have also dampened monetization prospects for big AI spenders. Taken together, investors may question whether higher adoption of lower-cost models could slow AI capex further. For the medium term, we believe the Jevons paradox will hold—lower AI token costs should support increasing demand for agentic AI workloads, more than offsetting lower per-unit economics. Although there is risk of heightened volatility if AI token supply scales faster than end-demand absorption, or if enterprise deployment, workflow integration, or monetization lags model-level efficiency gains, we believe the medium-term demand story remains intact.

Supply bottlenecks have yet to ease meaningfully, while monetization should accelerate further. Without taking any single-security views, Micron’s results last week suggested that supply constraints in key segments of the AI hardware supply chain remain, and we have yet to see any meaningful evidence that points to relief in the supply-demand dynamics. Meanwhile, we continue to expect cloud revenue to accelerate at major platforms through the rest of 2026, from the 40% growth reported for the first quarter of this year. These underscore the solid fundamentals of the AI growth story that we believe should remain a key driver of the broader market.

For investors, we believe that exposure to AI-related stocks will remain a key differentiator for equity market performance over the long run, but we also believe diversification, both within and beyond AI, is essential. This means that investors can consider more defensive areas within the AI complex, such as data center operators and select payment companies, as well as other structural trends such as Power and resources and Longevity. Substituting a portion of direct equity exposure with capital preservation strategies can also help investors navigate potential volatility in the near term.