Investing in the AI Revolution: Infrastructure, Applications, and Portfolio Strategy in 2026

Artificial intelligence has moved from a speculative theme to a genuine driver of corporate capital expenditure, productivity, and competitive positioning. The investment question is no longer whether AI is transformative — that debate is largely settled — but how to construct sensible portfolio exposure to a theme that is simultaneously one of the most genuinely exciting and most hyped in recent financial history.

The Infrastructure Layer: Where the Capital Is Going

The clearest and most immediate beneficiaries of the AI build-out have been the infrastructure companies: those that make, house, and power the hardware that AI workloads run on.

Semiconductors: Nvidia has become the most prominent example, with its GPU architecture proving to be the dominant compute substrate for training and inference of large language models. Demand for high-end AI chips has consistently exceeded supply, driving extraordinary revenue and margin expansion. But the semiconductor story is not a single stock: Advanced Micro Devices (AMD), Broadcom, Marvell Technology, and the broader chipmaking ecosystem all have roles. TSMC in Taiwan manufactures the chips; ASML supplies the lithography equipment. The supply chain for advanced semiconductors spans a handful of companies with significant pricing power.

Data centre REITs and infrastructure: Hyperscalers (Microsoft, Google, Amazon, Meta) are spending hundreds of billions of dollars on data centre capacity. This capital flows through to real estate investment trusts that own data centre facilities, to electrical equipment manufacturers (Schneider Electric, Eaton), and to specialist cooling companies. Data centre REITs such as Digital Realty and Equinix have benefited from power and space constraints driving higher contract values.

Utilities and power generation: AI data centres require substantial and reliable electrical power. US utilities with exposure to power purchase agreements with large data centre operators have re-rated. Nuclear power has attracted renewed institutional interest — both in the US (extensions of operating plant lives) and in broader discussions about small modular reactors (SMRs). In the UK, policy support for Hinkley Point C completion and new nuclear investment has increased.

Picks-and-Shovels Versus the Application Layer

The "picks-and-shovels" investment philosophy — backing the suppliers of essential inputs to a boom rather than the participants whose success is uncertain — has a strong historical precedent. During the California Gold Rush, those who sold equipment to miners often prospered more reliably than the miners themselves.

In AI, the picks-and-shovels approach favours semiconductor designers, data centre operators, and power infrastructure providers over individual AI software applications. The reasoning: many AI software applications will ultimately commoditise as models become cheaper to run, but the physical infrastructure they rely on retains its value.

The application layer — companies building AI into products and services (customer service automation, drug discovery, legal research, coding assistance, medical imaging analysis) — represents a broader and more heterogeneous opportunity. Some will generate durable competitive advantage from AI integration; many will not. The challenge for investors is distinguishing those where AI genuinely differentiates the product from those where it is a marketing narrative.

For most portfolio investors, access to the application layer is best achieved through diversified exposure — broad technology funds or AI-focused ETFs — rather than stock-picking individual AI software businesses.

AI Bubble Risk Assessment

The comparison with the dot-com bubble of 1999–2001 is frequently drawn. There are meaningful parallels: extraordinary valuation multiples for a small number of leading companies, enormous capital expenditure based on uncertain future revenue, and widespread investor enthusiasm that has attracted speculative retail participation.

However, there are also important differences:

• Revenue is real: The leading AI infrastructure companies — Nvidia, Microsoft, Amazon Web Services — have demonstrated actual revenue growth, not merely projected revenue. This is a substantive difference from many dot-com era companies that had no meaningful revenues at all.
• Enterprise adoption is underway: AI tools are being integrated into enterprise workflows at scale and at speed. The productivity case — reducing headcount, accelerating research, automating analysis — is being tested and, in many cases, validated.
• Concentration risk is real: A significant proportion of AI investment performance is concentrated in a small number of stocks. Any material disappointment in earnings, capex returns, or regulatory response could trigger sharp de-ratings.
• Valuation discipline matters: Paying very high multiples for any company, however high quality, reduces the expected return and increases downside risk. Position sizing and diversification remain essential.

The sensible conclusion is probably somewhere between "this time is different" and "bubble". AI will generate enormous economic value, but not all of that value will accrue to equity investors at today's valuations. Diversified exposure with disciplined position sizing is more appropriate than concentrated bets or wholesale avoidance.

UCITS ETF Options

For internationally mobile investors, UCITS-compliant ETFs domiciled in Ireland or Luxembourg are the standard access route. Tax treatment depends on the investor's country of residence; confirm the applicable withholding tax and reporting regime before investing.

Relevant UCITS ETFs with AI or technology exposure (this is not a recommendation; always verify current fund specifics):

• L&G Artificial Intelligence UCITS ETF (AIAI): tracks the ROBO Global Artificial Intelligence Index, weighting companies across the AI value chain — enablers, infrastructure, and applications.
• Global X Robotics & Artificial Intelligence UCITS ETF (BOTZ): primarily robotics and automation, with AI exposure.
• WisdomTree Artificial Intelligence UCITS ETF: broader AI theme, including data, cloud, and semiconductor companies.
• iShares Core MSCI World UCITS ETF: not AI-specific, but the MSCI World index has significant concentration in US large-cap technology, meaning global equity trackers carry implicit AI exposure.
• iShares MSCI USA Information Technology Sector UCITS ETF: direct sectoral exposure to US technology.
• Invesco QQQ/Invesco EQQQ (UK accessible): Nasdaq-100 tracker with very high concentration in AI-relevant large caps.

Investors should note that AI-focused thematic ETFs typically have higher total expense ratios (TERs) than broad index funds, and turnover in thematic indices can generate performance drag.

Portfolio Construction Considerations

• Position sizing: AI-related equities are volatile. An individual stock like Nvidia can move 10% or more in a single session on earnings results or macro sentiment. Thematic AI exposure should represent a portion of the total equity allocation, not the entire equity book.
• Existing exposure: Many global equity funds and indices already carry heavy technology weighting. Before adding an AI-specific allocation, review the overlap to avoid inadvertent concentration.
• Rebalancing: Thematic exposures require more active rebalancing decisions than broad indices, as themes can run far ahead of fundamentals or correct sharply.
• Currency considerations: Most AI infrastructure companies are US-listed, creating USD currency exposure for non-USD investors. This can be managed with hedged ETF share classes.

AI and the regulatory landscape

Artificial intelligence is increasingly attracting regulatory attention across major jurisdictions, and this is a material investment risk that is not always reflected in equity valuations.

EU AI Act: The EU's AI Act, which entered into force in 2024, establishes a risk-based regulatory framework for AI systems deployed in Europe. High-risk applications (healthcare diagnostics, credit scoring, hiring algorithms) face significant compliance requirements. This creates compliance costs for AI application layer companies operating in Europe and may slow adoption in regulated sectors.

US regulatory environment: The US has taken a more permissive approach than the EU, with executive orders rather than comprehensive legislation. However, FTC antitrust interest in AI market concentration — particularly regarding the dominance of a small number of foundation model providers and the vertical integration of cloud, model, and application providers — represents a longer-term regulatory risk to the concentration of market value in a small number of firms.

Data privacy: AI systems trained on large datasets are increasingly challenged on data privacy grounds (GDPR enforcement in Europe, state privacy laws in the US). Litigation and regulatory fines represent a cost and reputational risk for both model developers and enterprise adopters.

Regulatory risk is asymmetric: it is more likely to affect companies operating at scale in regulated sectors than pure infrastructure providers. This is another argument for a picks-and-shovels positioning bias.

Frequently asked questions

Is the AI theme already priced in? To a significant degree, yes. The leading AI infrastructure beneficiaries (Nvidia and others) were trading on very high earnings multiples as of mid-2026, reflecting considerable optimism about future revenue growth. The question is not whether AI will grow — it will — but whether the growth already embedded in valuations will be matched by actual outcomes. Valuations always matter; thematic enthusiasm does not override fundamental valuation discipline.

What is the difference between AI ETFs and simply buying a global tracker? A global equity index tracker (such as iShares Core MSCI World) already has very significant concentration in US large-cap technology, including the main AI beneficiaries. For many investors, adding a specific AI ETF on top of a global tracker creates unintended concentration in the same companies, not genuine additional diversification. Review the overlap before adding thematic exposure.

How should I think about the AI companies that are not yet profitable? Many AI application layer companies are in heavy investment mode — growing revenues rapidly but not yet generating positive net income. Valuing these requires assumptions about long-run margins, market share, and the competitive durability of their position. These are genuinely uncertain assumptions. The further out the expected profitability, the more sensitive the valuation is to the discount rate applied — meaning these companies are disproportionately affected by changes in interest rates and risk appetite.

Does AI pose a risk to the investments I already hold? AI creates risks as well as opportunities for existing businesses. Companies in sectors susceptible to AI-driven disruption — legal services, financial analysis, content production, customer support — face genuine competitive threats. Reviewing the portfolio for exposure to disruption risk is as important as assessing exposure to AI opportunity.

How Global Investments Can Help

Allocating to AI-related investments requires distinguishing between genuine long-term structural positioning and short-term trend-chasing. Our investment team reviews AI exposure within the context of your total portfolio, assesses the valuation discipline of any proposed allocation, and considers the tax-efficient wrapper structure most appropriate for your jurisdiction. Investments can fall as well as rise; past performance is not a guide to future returns, and thematic concentrations carry specific risks that must be understood. Contact us for a portfolio review.