- Semiconductors are the backbone of the 2026 AI rally. Capital investment is overwhelmingly directed at AI infrastructure, and the data centers doing the heavy lifting run entirely on chips. The numbers speak for themselves: the Philadelphia semiconductor index is up +90% YTD – twelve times the S&P 500's +7.5% - despite a couple of recent sell-off episodes in June. In emerging markets, which concentrate most of the supply chain, markets like South Korea and Taiwan are up +110% and +60% this year.
- Concentration is the flip side of the boom but fundamentals are strong. Concentration is most acute in EMs, where dependence on a single industry is now extreme. If the chips sector disappoints, the re-rating could be brutal. But the earnings are real and lopsidedly so: chips deliver 13% of S&P 500 profits on just 5% of sales, half of South Korea's profits on a fifth of its sales and over 70% of Taiwan's earnings and revenue alike. Ferocious demand colliding with supply bottlenecks from Middle East tensions are keeping prices sky-high.
- The bull case still stands on five pillars. First, hyperscalers show no sign of easing off: Upward capex guidance in Q1 points to a still-aggressive stance that should lift Asian chipmakers' profits by a remarkable +70% CAGR through 2026-2029. Second, the demand pipeline looks structurally solid: Broader AI diffusion, the corporate shift to agentic platforms and the memory-hungry new chip generation all reinforce one another. Third, the moat is real: capital and expertise are needed to build a modern foundry, keep competition at bay and pricing power intact. Fourth, valuations are compelling as based on solid earnings trajectory, and recent sell-offs offer new entry point on top picks. And fifth, in the over-concentrated EM space, semiconductors ensure geography and supply-chain position diversification, with an attractive trade-off compared to developed markets.
- But strong convictions do not mean complacency: Four risks could unsettle the rally over the next 12-24 months. First, a Middle East re-escalation could disrupt energy routed to Asia, forcing production halts – temporary or not. Second, bottlenecks in the Asian supply chain for inputs critical to data-center fleets could lift costs and lengthen lead times, throttling AI capex momentum. Third, a timing mismatch looms: New supply could come online just as capital allocation decelerates, should a fresh rate cycle take hold in the US or Europe. Fourth, competition is sharpening fast – Beijing is piling on funds and pressure to drive chip self-sufficiency (70% target in AI chips in 2026) and a bigger global footprint for its champions. Contrary to what market volatility suggest, we do not see lower inference cost to be a negative driver for memory chip business, but rather a guarantee of a sustainable book order if it comes with stronger AI diffusion and usage.