From the studio

Podcast: Jump Start | Fed comments and oil prices in focus (5 mins)
Video: Mid-year Checklist | Three portfolio drivers for the second half (8 mins)
Video: AI second-half outlook and CIO’s AI pecking order (5 mins)

Thought of the day

Semiconductor stocks posted a second consecutive week of declines last week, with the Philadelphia Semiconductor Index now nearly 14% below its all-time high in June amid investor concerns over the durability of AI capex growth.

But Micron over the weekend broke ground on the expansion of its factory in Japan to produce more advanced memory chips, while Samsung Electronics is reportedly planning to hike average third-quarter memory chip prices by about 20%. Hon Hai Precision Industry also posted a 40% jump in quarterly sales on continued AI demand.

Tech volatility is likely to remain elevated, in our view, as investors weigh strong agentic AI adoption against the growing risk of a slowdown in capex growth. We maintain a positive outlook on the AI investment theme as the demand story remains intact, but we also believe an active approach is warranted to navigate emerging risks.

AI adoption trends have been robust. Weekly consumption of tokens—the basic unit of AI computing—has risen roughly eight times this year based on OpenRouter data, pointing to strong AI adoption trends especially as agentic workloads continue to increase. Recent cloud compute agreements between hyperscalers and frontier AI labs, reportedly priced at four times the prevailing market rate, also suggest robust demand.

Semis valuations remain far from bubble territory. Despite the strong semiconductor rally this year, forward price-to-earnings multiples for the Philadelphia Semiconductor index at around 26x today are well below the 150x level at the peak of the dotcom cycle. With earnings largely keeping pace with share prices, we remain constructive on the semis complex. Consensus estimates currently imply earnings growth of 92% for the index this year and a further 40% in 2027.

A front-loaded spending cycle means the value creation may continue to shift along the AI value chain. As capex continues to grow, we estimate hyperscalers’ operating cash flows will be overtaken by their spending requirements in the second half of this year, with incremental spending increasingly funded through debt and equity issuance. Growing investor pressure for greater capital discipline could eventually pressure valuation multiples across semis and AI hardware. Meanwhile, supply bottlenecks in the AI value chain have continued to move upstream—from GPU systems in 2023-25, server components over the past year, to potentially semiconductor equipment over the next six months. These, in our view, point to the importance of diversified exposure as well as a dynamic approach.

So, with the investment narrative gradually shifting from near-term growth to the durability of spending beyond 2027, we think diversification and selectivity are key. We favor semiconductor equipment, foundries, CPU-related compute infrastructure, and memory within the “picks and shovels” of the AI build-out, but we also see value in defensive areas such as payment networks and data center REITs. Investors can also consider other structural trends such as Power and resources and Longevity, or substituting a portion of direct equity exposure with capital preservation strategies.