Slide 1: Artificial intelligence: Sizing and seizing the investment opportunity
Public Artificial intelligence Sizing and seizing the investment opportunity GWM Chief Investment Office Ulrike Hoffmann-Burchardi, CIO Global Equities Linda Mazziotta, CFA, Strategist Nikolaos Fostieris, Strategist 29 August 2024 This report was prepared by UBS Financial Services Inc. and UBS Switz
Slide 2: UBS CIO AI Views
UBS CIO AI Views 1 AI will be the most profound innovation and one of the largest investment opportunities in human history The ratio of monetization potential of the AI application layer to the costs of the enabling and intelligence layers 2 will become a key metric for investment returns 3 AI will
Slide 3: CIO View 1: AI the most profound innovation
CIO View 1 AI will be the most profound innovation and one of the largest investment opportunities in human history Value creation in past cycles Real US GDP per worker during innovation cycle, USD +15.7% +17.9% +26.8% +16.8% >+30%? 140,000 90,000 40,000 1964 1969 1974 1979 1984 1989 1994 1999 2004
Slide 4: CIO View 2: AI monetization ratio and market opportunity
CIO View 2 The ratio of monetization potential of the AI application layer to the costs of the enabling and intelligence layers will become a key metric for investment returns The AI market opportunity: a bottom-up perspective Source: UBS estimates, as of June 2024 Please read important disclaimer a
Slide 5: CIO View 3: AI data center capex cycle
CIO View 3 AI will kick off a data center capex cycle that will dwarf general purpose data center capex in the next years Annual capex for the enabling layer is expected to grow significantly by 2027 Split of estimated data center (DC) capex between AI and general purpose data centers, in USD bn and
Slide 6: CIO View 4: The AI silicon moment
CIO View 4 The AI silicon moment: AI chips will capture a large part of the value creation Over 70% of the bill of materials of an AI server is made up of GPU costs. This makes it the largest beneficiary, followed by memory and interconnects, which make up 15% and 9% of the total costs, respectively
Slide 7: CIO View 5: AI enablers first adopters
CIO View 5 The AI enablers will be the first adopters of AI, driving both revenue and margin upside We expect margin expansion for enablers of genAI, such as hyperscalers, as they capture revenue from AI products and productivity gains Average profit margin of select AI chip companies and cloud hype
Slide 8: CIO View 6: Oligopoly of vertically integrated AI foundries
CIO View 6 Monolithic players will emerge along the AI value chain and over time, the AI market will be dominated by an oligopoly of vertically integrated “AI foundries” With an attractive margin structure along the value chain, it is likely that an increasing number of companies will try to expand
Slide 9: CIO View 7: Software will become ubiquitous
CIO View 7 Software will become ubiquitous Productivity gains in coding development will accelerate the creation of software code Total number of GitHub repositories*, assuming a growth rate 55% higher than the previous 5-year average 800 Impact of co-pilots? 600 400 200 Education EnergyEnergy manag
Slide 10: CIO View 8: Data assets as competitive differentiators
CIO View 8 Data assets will emerge as the competitive differentiators for AI adopters Proprietary data is a strategic asset. Data from ….gives the ability to Google Waymo Millions of miles of autonomous driving improve AI algorithms for navigation and safety Amazon Customer transaction history, revi
Slide 11: CIO View 9: Proprietary models remain top performers
CIO View 9 Despite the increasing number of new open-source models, proprietary models will remain the top performers The top performing models are closed source proprietary models as of Aug 2024. Increased capex for foundational models will limit open-source competition. Sources: Hugging Face, as o
Slide 12: CIO View 10: Application and intelligence layers merge with AGI
CIO View 10 The application layer and intelligence layers will merge with artificial general intelligence (AGI) Single language models achieve above average scores across domain-specific exams GPT-4 performance on academic and professional exams. Scores in performance percentile 100 80 60 40 20 0 Un
Slide 13: Disclaimer/Risk information (1/2)
Disclaimer/Risk information (1/2) UBS Chief Investment Office's ("CIO") investment views are prepared and published by the Global Wealth Management business of UBS Switzerland AG (regulated by FINMA in Switzerland) or its affiliates ("UBS"), part of UBS Group AG ("UBS Group"). UBS Group includes for
Slide 14: Disclaimer/Risk information (2/2)
Disclaimer/Risk information (2/2) This material may not be reproduced or copies circulated without prior authority of UBS. Unless otherwise agreed in writing UBS expressly prohibits the distribution and transfer of this material to third parties for any reason. UBS accepts no liability whatsoever fo