Observing institutional investment logic from the UBS Global Research system: AI cost curves, quantum computing, and a long-term perspective on global asset allocation

Observing Institutional Investment Logic Through the UBS Global Research System: The AI Cost Curve, Quantum Computing, and a Long-Term Perspective on Global Asset Allocation

Against an environment in which the interest rate path remains unclear, inflation stickiness repeatedly resurfaces, and geopolitics and technological competition are intertwined, the way institutional investors make decisions is undergoing structural change: from judgments based on a single asset class toward a cross-asset, cross-region research system capable of sustained tracking. Public materials from UBS Global Research show that its team consists of approximately 720 analysts, strategists, and economists, covering more than 50 markets and about 4,000 stocks. This article does not take short-term price movements as its subject; rather, it uses the capability composition and recent topics of this research framework as threads to observe how global investment research is being redefined, and what this shift means for long-term asset allocation.

Market Background: Macroeconomics and Microeconomics Reconnect

Over the past decade-plus, the low-rate environment substantially reduced the marginal impact of macro judgments on portfolio outcomes, and asset allocation relied more on relative valuations and risk premia across asset classes. As the interaction among interest rate cycles, inflation structure, and fiscal policy becomes more complex, relying solely on mean-reversion logic at the asset-class level can no longer explain the enormous dispersion within the same asset class.

UBS’s positioning of its economics and strategy teams precisely reflects this change: UBS Economics draws on the judgments of market strategists and works closely with company and sector analysts to gain a deeper understanding of how the world economy operates; UBS Global Strategy helps investors identify drivers and connect them with opportunities across asset classes and regions. In other words, macro judgments and micro evidence are viewed as two ends of the same analytical chain, rather than two independent professional domains.

The long-term historical perspective is also being reemphasized. UBS public materials mention the Global Investment Returns Yearbook 2026, whose purpose is to examine the drivers of long-term real asset returns and extract lessons that can be used to understand the future. In periods of heightened volatility, such long-term data spanning decades and covering multiple asset classes has direct research value for long-horizon allocators such as pensions, sovereign wealth funds, and insurance capital: it answers not “what to buy next quarter,” but “which factors truly determine real returns over the long term.”

At the same time, the research agenda itself is changing. UBS Global Research has made “generative AI” one of its key themes, and explicitly states that its purpose is to showcase the breadth, depth, and originality of its economists, strategists, and analysts. This formulation is noteworthy: AI is no longer merely an industry research topic, but more like an analytical variable that runs across multiple industries and asset classes.

Current Capital Flows: Where Attention Is Concentrating

需要强调的是，以下讨论基于UBS公开披露的研究议题与能力构成，而非对具体资金规模的推断。研究议题的选择，本身是观察机构注意力流向的一个可靠窗口。

第一，生成式AI从“概念讨论”转向“成本与效率讨论”。 UBS全球研究播客中，UBS Research 的 Caroline Li 与 Wei Xiong 讨论了AI采用趋势、推理成本下降，以及中国AI生态可能揭示的下一阶段竞争格局。当组织从实验阶段走向可衡量的回报阶段，成本效率正成为部署决策中的重要因素。这一转变对投资研究的含义是：AI叙事的验证标准，正在从“能力上限”转向“单位经济性”。

第二，量子计算被放置在“长期期权价值”的位置上。 在另一期讨论中，UBS量化研究负责人 Paul Winter 与 Elena Ynduráin 教授探讨了该技术的现状、“量子优势”的实际含义，以及其与AI的融合可能如何影响创新。讨论同时指出，量子计算目前主要仍处于测试阶段，AI的进展可能有助于在更长时间维度上释放其潜力。对于长期配置者而言，这类主题的价值不在于短期催化，而在于它是否可能改变某些行业的成本结构与竞争壁垒。

第三，对私营市场的系统性研究需求上升。 UBS的私募公司研究（Private Company Research）以系统化方法对私营公司进行画像，并评估其竞争定位。随着创新活动与资本形成在私营市场中的比重上升，公开市场研究若缺少私营部门的对照坐标，对行业格局的判断将出现缺口。这也是机构投资者近年来持续关注私募股权与另类资产的一个研究方法论层面的原因。

第四，另类数据与量化工具成为研究基础设施。 UBS Evidence Lab 提供覆盖55个以上专业领域的“可直接用于洞察的数据集”，资产库覆盖5,000多家公司；UBS HOLT 数据库包含超过20,000家公司，用于全球公司的比较与估值；Quant Answers 则提供组合风险预测、组合风格与基本面分析等模块。这些工具的共同指向是：研究结论需要可追溯、可复现、可被组合层面检验。Fifth, consumption and cultural topics enter long-term brand value analysis. UBS analysts Jay Sole and Michael Lasser discussed how sports and cultural moments are transformed by consumer companies into long-term brand value, as well as the growth opportunities brought by shifts in participation trends. This may seem unrelated to macro, but it is an important part of understanding the long-term profitability of the consumer sector—brand equity, user engagement, and cultural relevance constitute competitive barriers that are difficult to measure with short-term financial metrics.

Putting these topics together reveals a clear thread: institutional focus is shifting from “asset class rotation” to “structural themes + verifiable evidence.”

Investment Logic Analysis: Structural Factors Driving Change

First, the downward movement of the technology cost curve has changed the durability of investment themes. As AI moves from experimentation to large-scale deployment, inference costs, energy consumption, data governance, and organizational adaptation capabilities become key variables determining who can truly reap returns. This means the AI theme will not exist in the form of a single industry, but will permeate the cost structures of multiple industries in the form of “efficiency gains.” For asset allocators, such themes are closer to long-term infrastructure than to temporary hotspots.

Second, the boundary between public and private markets is blurring. The maturation cycle of innovative companies has been extended, and a mismatch has emerged between the industry coverage investable in public markets and the distribution of innovation activity. This explains why systematic private-company research, private equity, and alternative investments are being included in broader asset allocation discussions, rather than existing only as satellite allocations.

Third, data and model capabilities have become the “production function” of research institutions. When data coverage, quantitative frameworks, and valuation tools become reusable infrastructure, the marginal cost of research output declines, but verification requirements rise. The combination of UBS Evidence Lab, HOLT, and Quant Answers reflects exactly this division of labor: breadth of data, consistency of valuation, and quantification of portfolio risk—all three are indispensable.

Fourth, the liability structure of long-term institutional capital requires “explainable sources of return.” Pension funds, sovereign wealth funds, and family offices, when facing fiduciary responsibilities and internal governance, need frameworks that can explain sources of return and risk exposures. This is exactly why long-term research such as the Global Investment Returns Yearbook, as well as new content distribution methods such as podcasts, are valued—research must not only be correct, but also understandable, discussable, and able to be incorporated into decision-making processes.

Taken together, these factors are not short-term phenomena. They jointly point to a long-term trend: investment research is shifting from “opinion supply” to “evidence infrastructure,” while asset allocation decisions are shifting from asset class segmentation to frameworks centered on themes, risk factors, and structural drivers.

Risk FactorsMacro risk. The interest rate path and inflation structure remain uncertain. If the pace of disinflation and the rhythm of policy adjustment become mismatched, cross-asset correlations may change significantly within a short period, rendering portfolio assumptions based on historical correlations invalid.

Policy and regulatory risk. Changes in AI regulation, cross-border data flow rules, technology export controls, and tax policy may all reshape the competitive landscape and cost structure of the technology industry. For portfolios with technology themes as their core exposure, policy variables have become a source of risk as important as fundamentals.

Geopolitical risk. The fragmentation of technology ecosystems, supply chain restructuring, and restrictions on regional capital flows will affect how companies enter markets and the paths through which capital returns are realized. One reason the evolution of China's AI ecosystem has attracted research attention is precisely that it may reveal the regionalized characteristics of the competitive landscape.

Valuation and expectation risk. When a theme gains broad consensus, prices often adjust before cash flow improvements. The biggest risk of thematic investing is usually not being wrong on direction, but timing mismatches and excessive pricing.

Model and data risk. Rising reliance on alternative data, quantitative models, and systematic valuation frameworks also brings new risks: data quality, sample bias, model assumptions, and the limitations of historical extrapolation may all be underestimated. The more powerful the tools, the higher the requirement for methodological transparency.

Liquidity risk. Greater research depth in private markets and alternative assets does not automatically equate to improved liquidity. For long-horizon capital, valuation frequency, exit paths, and the pace of capital calls remain risk dimensions that need to be managed independently.

Timing risk of frontier technologies such as quantum computing. As has been made clear in discussions, quantum computing is currently mainly at the testing stage. The investment implications of such technologies are closer to long-term options than to near-term cash flow sources that can be linearly extrapolated.

Long-Term Outlook

Next 3 to 5 years: research industrialization and thematic frameworkization. Data infrastructure, quantitative tools, and private market research will gradually shift from "differentiated capabilities" to "basic allocation." At the same time, AI themes will more clearly shift from industry narratives to cost and efficiency metrics, and investors will focus more on the measurability of deployment returns rather than demonstrations of model capabilities.

Next 5 to 10 years: AI becomes infrastructure rather than a standalone theme. If inference costs continue to decline and AI's integration with business processes deepens, AI is more likely to become a background variable for analyzing other industries, much like electricity and the internet. By then, what truly determines return differences may be "non-AI" factors such as energy supply, data governance, talent structure, and organizational efficiency.

The long-term option value of frontier technologies. The prospect of integrating quantum computing and AI is a typical long-cycle, high-uncertainty theme. Its research value lies in identifying which industries may have their cost curves changed by leaps in computing power, rather than making precise predictions about commercialization timing.Further Evolution of the Asset Allocation Framework. With the integration of public and private market research and the standardization of alternative data, allocation decisions may be organized more around risk factors, thematic exposures, and long-term drivers rather than traditional asset class proportions.

Changes in Research Distribution and Decision-Making Processes. Podcasts, data platforms, and client portals (such as the macroeconomic and local market insight channels provided by UBS Neo) indicate that research is becoming embedded in more parts of the investment process. For institutions, this means that the criteria for evaluating research will shift from “whether the conclusion is appealing” to “whether the process is reviewable.”

Conclusion: The Long-Term Value of Research Lies in Verifiability

In the global investment landscape, changes in capital flows are often observed before price changes, and the ability to observe depends on the breadth and depth of the research system. The capability structure publicly presented by UBS Global Research—cross-asset economics and strategy teams, analytical coverage of approximately 4,000 stocks across more than 50 markets, data assets covering more than 55 specialist domains, a valuation database of over 20,000 companies, and systematic private company research—essentially reflects the same trend: institutional investors need not more opinions, but a more verifiable chain of evidence.

For long-term capital allocators, the question truly worth tracking continuously is not whether a theme will be in favor in the next quarter, but: Which structural factors are changing the cost curves and competitive barriers of industries? Do these changes have measurable data support? Have current prices already reflected these changes? And, with macroeconomic, policy, and geopolitical risks coexisting, does the portfolio have sufficient diversification and explanatory power?

In this sense, the value of investment research lies not in prediction, but in organizing uncertainty into a manageable form. This is both a long-term proposition for institutional investment strategy and the underlying logic of the continued evolution of global capital markets.