Artificial intelligence is often presented as a narrow investment theme centered on semiconductor companies, cloud platforms, and software developers. That view is incomplete.
AI is a physical and economic system. It requires minerals, electricity, transmission capacity, buildings, cooling equipment, chips, memory, servers, networks, software, data, applications, and ultimately machines that can act in the real world. Every layer depends on the layers beneath it.
The progression is:
Raw Materials → Power → Infrastructure → Semiconductors → Compute → Intelligence → Applications → Physical AI / End Users
This framework matters because some of the companies benefiting from AI investment may never develop a foundation model. A copper producer, transformer manufacturer, natural-gas pipeline, nuclear generator, cooling company, or data-center contractor may participate in the same long-term buildout.
Research from iShares describes AI as a value chain spanning hardware, infrastructure, software, and services. Goldman Sachs has emphasized that the scale of future AI capital spending depends on assumptions about chip life cycles, data-center complexity, architecture, power, and equipment bottlenecks. These are not forecasts or guarantees, but they demonstrate why investors should analyze AI as an economic ecosystem rather than a collection of technology stocks.
1. AI Materials
Includes: copper, uranium, steel, aluminum, rare earths, and specialty chemicals.
AI begins with physical inputs. Materials sit at the base of the chain because data centers, transmission systems, semiconductor fabrication, and advanced electrical equipment all require large volumes of metals, minerals, and high-purity industrial inputs. If materials are constrained, the rest of the buildout becomes more expensive or slower to execute.
In other words, AI demand does not begin at the software layer. It begins with the resources required to construct and power the system. That makes this the first layer in the progression and an important reminder that AI exposure is not limited to traditional technology classifications.
2. AI Power
Includes: nuclear, natural gas, utilities, electricity generation, and transmission.
Power is the next necessary layer because AI workloads are electricity-intensive. Training, inference, storage, and networking all depend on reliable generation and grid capacity. It is not enough for power to exist in the abstract; it must be deliverable, permitted, and dependable where compute capacity is being built.
That makes AI Power a foundational constraint. In many cases, access to generation, transmission, and interconnection may determine which projects move forward and on what timetable. Without sufficient power, higher layers in the chain cannot scale efficiently.
3. AI Physical Infrastructure
Includes: transformers, switchgear, cooling systems, construction, data centers, and fiber.
Physical infrastructure converts power availability into operational capacity. This layer includes the electrical equipment, cooling systems, facilities, and connectivity that allow AI hardware to function in the real world. It sits between energy supply and semiconductor-enabled computing because it is the bridge that makes both usable.
This matters because AI is not merely a digital abstraction. It requires land, buildings, thermal management, network connections, and specialized equipment. When investors overlook this layer, they may underestimate how much of the AI economy depends on industrial and utility-adjacent systems rather than software alone.
4. AI Semiconductor Supply Chain
Includes: semiconductor equipment, wafers and materials, foundries, advanced packaging, memory, GPUs, and ASICs.
Semiconductors are the computational engine of AI. This layer includes not only chip designers, but also the equipment, materials, foundries, memory, and packaging capabilities required to manufacture advanced processors at scale. It sits after infrastructure because chips still require powered, cooled, and connected facilities before they can be deployed effectively.
For many investors, this is the most familiar part of the AI theme. However, it is still one layer within a broader system. The semiconductor supply chain matters because it translates physical capacity into computational capability, but it does not operate independently of the layers beneath it.
5. AI Compute & Networking
Includes: servers, storage, networking equipment, and cloud infrastructure.
Compute and networking assemble chips into usable systems. This layer includes the servers, storage architecture, switching, optical connectivity, and cloud environments that allow organizations to train models, run inference, and distribute AI services. It follows semiconductors because processors must be integrated into complete platforms before they can create commercial utility.
This layer matters because raw chip supply alone does not deliver business outcomes. AI requires throughput, latency management, storage, orchestration, and reliable access pathways. In practical terms, this is where semiconductor capability becomes scalable enterprise infrastructure.
6. AI Intelligence
Includes: foundation models, databases, developer tools, data infrastructure, and cybersecurity.
The intelligence layer turns compute capacity into software capability. Models, data systems, development tools, and security architecture allow organizations to generate outputs, customize workflows, protect information, and build deployable AI solutions. It sits above compute because software intelligence depends on the infrastructure below it.
This is the layer most closely associated with the public conversation around AI, but it remains dependent on power, facilities, chips, and systems capacity. Understanding that dependency is critical. Intelligence is economically important, but it is not self-contained.
7. AI Applications
Includes: enterprise software, financial services, healthcare, advertising, and consumer applications.
Applications are where AI begins to show measurable economic utility for end users. This layer takes underlying intelligence and embeds it into workflows, decision systems, products, and sector-specific use cases. It sits above the intelligence layer because applications monetize AI through implementation rather than through model development alone.
This is where investors should look for evidence of actual adoption: productivity gains, process improvement, pricing power, customer retention, and better operating metrics. The key point is that application value depends on the software and infrastructure stack below it.
8. Physical AI
Includes: robotics, industrial automation, autonomous vehicles, drones, and defense applications.
Physical AI is the final layer in the chain. It applies digital intelligence to machines that perceive, decide, and act in the physical world. This includes robotics, automation, autonomous systems, and other machine-based end uses. It sits at the top of the chain because it depends on every lower layer: materials, power, infrastructure, semiconductors, compute, intelligence, and applications.
For long-term investors, this layer illustrates the full economic reach of AI. The technology is not limited to screens and software. Over time, it may reshape logistics, manufacturing, transportation, defense, and other real-world operating environments.
Avoiding Double-Counting AI Exposure
The eight-bucket framework can improve portfolio construction, but it also reveals a common mistake: believing a portfolio is diversified when it is concentrated in the same AI spending cycle.
An investor might own:
- A semiconductor ETF
- A technology-sector ETF
- A Nasdaq fund
- Several cloud companies
- Individual AI stocks
- A robotics fund
Those holdings may appear to represent different ideas, yet many may depend on the same hyperscaler capital-spending plans, chip demand, data-center construction, or valuation assumptions.
A more useful review asks:
- Where in the AI value chain is spending unavoidable?
- Where are genuine supply constraints rather than temporary excitement?
- Which companies have durable margins, backlogs, recurring revenue, or balance-sheet strength?
- Which businesses have measurable exposure rather than an AI marketing narrative?
- How much future growth is already reflected in the share price?
- How does the position fit the household’s tax, liquidity, retirement, and risk requirements?
AI exposure should be evaluated within the broader financial plan. A concentrated position may create tax consequences, employer-stock risk, or an inappropriate level of volatility even when the underlying theme is attractive. InSight Financial Planners’ comprehensive planning and investment management approach is designed to connect investment decisions with cash flow, taxes, retirement, estate planning, and risk management.
Conclusion
AI is not simply a group of software companies and semiconductor stocks. It is an interconnected economic value chain:
Materials → Power → Physical Infrastructure → Semiconductors → Compute → Intelligence → Applications → Physical AI
The objective is not to predict the single winning model, chip designer, or application. A more disciplined objective is to identify thoughtfully selected pieces of the infrastructure required for AI to exist and expand.
That may include technology companies, but it may also include miners, utilities, electrical-equipment manufacturers, construction firms, data-center operators, network providers, software companies, healthcare businesses, and industrial automation leaders.
As this series develops, each layer deserves its own analysis because the economics, competitive structure, capital intensity, and valuation drivers can differ substantially across the chain. What they share is interdependence. Higher-layer growth still rests on lower-layer capacity.
The investment decision is ultimately not whether AI sounds promising. It is whether a particular business has durable economics, credible demand, financial strength, competitive advantages, and a valuation that compensates investors for the risks.
Disclosure: This article is provided for general informational and educational purposes and does not constitute investment, tax, or legal advice, or a recommendation to buy or sell any security or adopt any investment strategy. References to companies, securities, ETFs, industries, and market research are illustrative only and do not represent a complete list of investments or current or future holdings. Investing involves risk, including possible loss of principal. AI-related investments may be subject to significant volatility, concentration risk, technology and obsolescence risk, regulatory and geopolitical risk, commodity-price risk, interest-rate risk, and cyclical demand. Forward-looking estimates and projections may not come to pass. Investors should consider their individual objectives, risk tolerance, time horizon, tax situation, and overall financial plan and consult qualified financial, tax, and legal professionals before making investment decisions. InSight Financial Planners is a Registered Investment Advisor. Past performance is not indicative of future results.
