Transformational Innovation Opportunities (TRIO)

18 August 2026, 06:00 UTC Chief Investment Office GWM Investment Research

Intelligence weekly #124: Key investor debates on each layer of the AI computing stack Transformational Innovation Opportunities (TRIO): Artificial Intelligence Authors: Ulrike Hoffmann-Burchardi, Chief Investment Officer Americas and Global Head of Equities, UBS Financial Services Inc. (UBS FS); Achille Monnet, CIO Equity Analyst, UBS Switzerland AG; Delwin Kurnia Limas, CFA, CIO Equity Strategist, UBS AG Singapore Branch; Kevin Dennean, CFA, CIO Equity Strategist, US Technology & Telecom, UBS Financial Services Inc. (UBS FS); Allen Pu, CFA, CIO Equity Strategist, UBS AG Hong Kong Branch; Arafat Alafate, CIO Equity Strategist, UBS AG Hong Kong Branch; Nikolaos Fostieris, Global Equity Strategist, UBS Switzerland AG; Xueqiong Huang, Equity Strategist, UBS AG Singapore Branch

• As we approach the fourth anniversary of ChatGPT's “iPhone moment” and the artificial intelligence infrastructure buildout continues to evolve, we examine some of the key debates surrounding the major components of the AI infrastructure value chain.

• We continue to favor the enabling layer, which we define as semiconductors and cloud computing. Having been less exposed to the Magnificent 7 for much of the year, we recently increased our positioning as visibility on cloud returns improved and valuations became increasingly attractive. Elsewhere, we maintain a barbell approach, with meaningful exposure to defensive technology segments such as payments and data center Source: Shutterstock REITs. Despite a recent tactical rerating in applications, we retain an allocation of about 10% and remain selective, preferring large enterprise platform-focused vendors.

• Given the rapidly shifting narrative, we believe it is important for investors to remain focused on fundamentals and signals rather than noise. Accordingly, we think outlining the key arguments surrounding each subsegment can provide useful insight.

Our view machine. Understanding how the market is voting today As the AI infrastructure buildout has matured and model allows us to take an independent perspective and ultimately technology has advanced, investor focus has continuously increases the likelihood that the market serves us rather than shifted among major debates. The first leg of the buildout informs us. was largely defined by a shortage of graphics processing unit (GPU) infrastructure and consumer AI applications. Today, Some of the key debates that we see center around: much of the discussion centers on agentic AI, increasing the • GPU market share and compute economics, importance of graphics processing units (GPUs), memory, and open-source models, among other areas. • CPU attach rates and growth, • Memory pricing power and cycle dynamics, We believe understanding the current market debates provides valuable signals and implications for future • Data center return on invested capital, circular financing, and credit backstops, positioning. Only then can we form an informed view of where investors are overly optimistic or pessimistic. To • Open-source vs. closed-source, borrow a famous analogy: In the short term, the market is a voting machine, but in the long term, it is a weighing • The risks to Software-as-a-Service applications.

This report has been prepared by UBS Financial Services Inc. (UBS FS), UBS Switzerland AG, UBS AG Singapore Branch, UBS AG Hong Kong Branch. Please see important disclaimers and disclosures at the end of the document.

Transformational Innovation Opportunities (TRIO): Artificial Intelligence

While we do not always provide an exhaustive conclusion, conservative scenario. we aim to clearly articulate our positioning and perspective. In the near term, the backdrop for AI semiconductors can AI selection list be simplified into token supply, driven by semiconductor availability and data center development, versus token demand, driven by the elasticity of AI demand, the pace of model improvement, and the deflation in per-token costs. The reason this equation remains imbalanced today is due largely to investment decisions made 18 months ago, as semiconductor supply chains remain among the most complex…