UBS has moved to reassure investors rattled by a chip-stock wobble, arguing that fresh calls to slow the development of the most advanced artificial intelligence models are unlikely to derail the industry's spending spree.
Semiconductor shares came under pressure on Monday after several leading US AI labs publicly backed efforts to pace the rollout of their most powerful systems. Memory names in North Asia fell by mid-single digits and America's Philadelphia semiconductor index dropped 5.9%, though the selling had eased by Tuesday.
The trigger was a weekend push by frontier lab bosses for tougher safeguards and closer coordination, following reports of behind-closed-doors talks to set up an industry-led safety standards body. The debate has sharpened as departing safety researchers warn that advanced AI could threaten humanity. A cross-party effort in the US Senate to write national rules is also gaining ground, while California has brought in standards for independent AI auditors. President Donald Trump, for his part, dismissed the misuse fears as a "hoax".
UBS, led by chief investment officer Mark Haefele, urged caution before reading any of this as the end of the AI investment cycle.
Its central point is that pacing need not mean less spending. Roughly two-thirds of computing demand, the bank reckons, comes from running AI models rather than training them, and that is driven by real-world take-up rather than by how clever the latest model is. The safety proposals also draw a clear line between slowing and halting training. Elon Musk confirmed at the weekend that work continues on xAI's Grok 4.8. UBS is sticking with its forecast for industry capital spending to hit US$1.2 trillion in 2027, a 33% jump on this year.
Tighter rules, it added, may reshape who wins rather than choke off development, potentially favouring deep-pocketed incumbents over startups and overseas rivals.
The wobble also arrives with valuations looking less stretched. The SOXX index trades on around 21 times forward earnings, down from 33 in June.
The real question, UBS concludes, is not whether frontier development slows but whether AI demand keeps growing. Its answer remains yes.
