Risk-off move driven by chips argues for broadening across the AI stack
Highlights: Semiconductor stocks are down 14 per cent from the peak as worries about their long-term earnings growth combine with concerns over higher rates and some profit taking ahead of the US midterms. Note that there has been no major earnings disappointment from the semis or spending cuts by the hyperscalers. Other parts of the AI stack are much more resilient, with hyperscalers down 5 per cent, and AI adopters still trading close to their highs. We spread our exposure across the AI stack but remain invested as we are confident of fast adoption and good monetisation.
- Global chip stocks are down sharply since early October. but hyperscalers just 5 per cent, software 4 per cent and adopters 1 per cent. Investors are rotating into adopters, energy, utilities and quality stocks
- This is about future earnings potential, not current profits of semiconductor and equipment companies, which remain strong. Chinese competition became a more immediate concern following CXMT’s progress in increasing memory-chip density. CoreWeave and Broadcom’s announcements added to the sell-off, as they suggested that funding costs could constrain expansion for smaller infrastructure buyers. However, we don’t see CoreWeave’s heavily debt-funded model as representative for the ecosystem: CDS levels for other players are stable
- The relative impact of these concerns is easily explained. Semiconductor companies could face slower order growth, increased competition and margin normalisation, while hyperscalers and adopters would benefit from cheaper computing. Meanwhile, productivity gains and new revenue streams lift adopters’ earnings (though laggards can be challenged by competitors)
- To be sure, there is plenty of chip demand, full order books and rapidly rising demand as AI training continues and inference is boosted by the acceleration of agents and physical AI. Scaling semi production is not cheap, straightforward or fast. As a result, while some short-term volatility could linger, we think markets will stabilise when investors see more evidence that semiconductor orders don’t deteriorate and hyperscalers guidance remains strong
- Rotation within the AI stack is not new. Market narratives have shifted between AI optimism, AI disrupting software, and concerns over excessive AI spending or returns on investment. The best approach is to take balanced exposure to the stack, to position for the next phase of AI monetisation, which broadens beyond chips to broader infrastructure and adopters