By Jing Peng
With AI capital commitments scaling to the level of national infrastructure, the winning edge belongs to asset managers who can translate massive, multi-asset complexity into resilient portfolios.
AI is no longer just a technology theme. It has become a global, cross-asset capital chain spanning computing power, data centres, energy, and transmission networks. Much of the news cycle remains fixated on the technology race, and understandably so. But from an investment perspective, success will no longer defined by picking the right product. It will come from executing complex, multi-layered portfolios with insight and discipline. That distinction is what separates winning asset managers from the rest as AI investing enters its next phase.
Why Isn’t “Which Product Should I Buy?” the Right Question Anymore?
The infrastructure required to scale AI is already reshaping global capital flows. Computing power, chips, data centres, electricity, power grids, cooling systems, cloud services, and long-duration capital are becoming interconnected parts of a rapidly growing investment ecosystem.
Bloomberg reported that US private data centre construction spending reached an annualised US$50.7 billion in April 2026, accounting for 2.3% of all US construction spending at the time. This shows that AI infrastructure investment is now approaching traditional infrastructure in scale.
What matters here, however, is not the size of the build-out, but what it implies for portfolio management. AI is no longer a technology theme. It is becoming a global capital chain that spans public markets, private markets, infrastructure, and alternative assets. It is inherently global, inherently cross-asset, and must be managed within portfolios rather than treated in isolation.
This is why the discussion around AI investment needs to move beyond theme selection and into the broader question of asset management capabilities. When faced with a complex theme such as AI infrastructure, it is no longer enough to ask which product should investors buy. A one-product lens obscures the more important portfolio question: what role is each exposure expected to play?
How Should Investor Break Down the AI Value Chain?
From an asset management perspective, the AI infrastructure value chain breaks down into multiple layers:
- Compute core — chips and processing capacity
- Data centres — physical facilities and real estate
- Energy supply — power generation and grid capacity
- Transmission and distribution networks — the infrastructure connecting power to demand
- Network connectivity and operating efficiency — the systems that keep it all running
These layers can each fit into one or more of three broader categories: digital infrastructure, energy systems, and physical infrastructure supporting AI deployment. Each group carries different asset characteristics which means they should be priced, managed, or exited differently:
This framework serves as a clear reminder: AI-related investments cannot be forced into a single product category. Underlying infrastructure, platform-layer assets, and early-stage venture opportunities need to be assessed against different risk-return profiles, liquidity conditions, and exit paths.
In practice, this demands a broader and more continuous portfolio management process. Asset managers need to observe how pricing, liquidity, and exit windows differ across the AI value chain, from energy and computing power through to platforms and applications. For wealth managers, it is not enough to assess whether a client “has AI exposure.” The first question should be what kind of AI exposure the portfolio owns, what role it plays, how liquid it is, and under what conditions it should be increased, reduced, or rebalanced.
It logically follows, then, that the real challenge for professional asset managers is not how to package a macro theme into a single product, but how best to break it down into portfolio exposures that can be allocated, monitored, and adjusted over time. This is where discretionary portfolio management becomes increasingly important: it connects mandates, risk budgets, portfolio construction, and rebalancing into one ongoing process. In practice, this means managing liquidity, risk exposures, correlations, and execution pace alongside asset allocation.
Is AI Replacing Investment Judgement?
AI’s significance does not stop at what sits inside a portfolio. It is also reshaping how asset managers themselves work, with a growing number of use cases across investment research, portfolio monitoring, and risk management, from tracking global market signals to flagging early portfolio deviations.
That said, I do not believe AI can currently replace investment judgment in its current form. Ultimately, the decisions that matter most — such as whether to act, how to make trade-offs, and how to manage portfolio exposure — still require human experience, accountability, and long-term judgment. AI can widen the field of view, but it cannot yet substitute for the responsibility of a decision.
In that sense, AI is better understood as a force multiplier, helping teams track global market signals more broadly, identify portfolio deviations earlier, and conduct scenario analysis and risk alerts more systematically than manually processes allow.
What Can Japan, Railways, and Telecom Networks Teach Us About Durable AI Value?
Our recent conversations with clients and partners in Tokyo reinforced this perspective. Japan may not be at the centre of the loudest AI narratives. But its strengths in precision manufacturing, industrial upgrading, energy-efficiency improvement, and trusted supply-chain networks position it across several foundational layers of the AI value chain — exactly the kind of structural exposure that investors there said they value for long-term portfolio resilience.
Historically, major infrastructure waves — such as railways and telecom networks — have rewarded investors who focused on how exposures were assembled, financed, and managed over time, rather than on predicting a single winning project. Japan reinforces the same pattern today: durable investment value often lies not in the most visible themes, but in the underlying structures that support long-term operation.
What Should Asset Managers Take Away?
As AI enters a capital-pricing phase, the takeaway is simple. If you don’t know how to already, learn how to translate complex, cross-asset AI opportunities into manageable, adjustable, and continuously executable portfolios. The concept of “AI fluency” is commonly used to describe how deep a user understands AI tools and their limitations. It’s time to extend this to asset managers and how deep they understand AI infrastructure layers and their asset characteristics.
In the age of AI infrastructure, the real test for professional asset managers is whether they can connect global perspectives, asset allocation, risk discipline, and continuous execution within one portfolio process. For long-term capital, owning the future matters more than trying to predict it.
