Artificial intelligence is reshaping industries fast. Felix Prehn from Goat Academy breaks down a simple way to understand where the value is created in AI and how public markets map to it today. The approach organizes AI into seven practical tiers. Each tier explains a different way companies make money from AI and where public investors can find exposure.
Tier 0: Energy infrastructure
AI needs huge amounts of electricity. One large AI data center can consume as much power as a city. That is why energy producers and power equipment companies benefit as demand grows. Nuclear power and advanced cooling systems are especially important because AI servers run 24/7 and get very hot. Example ideas often discussed in markets include nuclear-heavy utilities and data center cooling vendors. These can be lower risk than pure software, but they still move with interest rates and power prices.
Term explained: data center cooling. These are liquid or air systems that remove heat from servers to prevent failure and keep performance high.
Tier 1: Semiconductors
Chips are the “engines” of AI. Better chips mean faster training and cheaper inference (running AI models). The discrete GPU market is highly concentrated today, with one company holding the lion’s share, and fast followers building competitive products. Fabrication giants that manufacture chips for many designers provide diversified exposure and are often sold out years in advance.
Term explained: inference. This is when a trained AI model is used to produce outputs, like answers or images, for users.
Tier 2: Data centers
Think of data centers as the “houses” where AI lives. They are specialized buildings with power, networking, and cooling. Many are owned by real estate investment trusts (REITs) that sign multi-year leases with cloud providers, which can make cash flows more predictable. Yields vary, and valuations can be sensitive to interest rates.
Term explained: REIT. A company that owns income-producing real estate and pays out most of its profits as dividends.
Tier 3: Foundation models
These are the core AI “brains” such as large language models (LLMs). Big tech firms fund them because they are expensive to build and train. Public exposure typically comes through diversified platforms that bundle models with search, cloud, office tools, or social apps. The upside is large distribution and cash flow. The trade-off is size: large platforms can move slower than startups.
Tier 4: Software infrastructure
This layer provides the tools to deploy, manage, and secure AI across enterprises. It includes data integration, orchestration, safety, monitoring, and governance. A key challenge is that model providers increasingly build parts of this tooling themselves. Still, some specialized platforms have carved strong positions by helping large organizations connect messy real-world data to AI systems.
Term explained: ontology (in business software). A structured map of how a company’s operations, data, and processes fit together so software can reason about them.
Tier 5: Applications
These are the “furniture” people actually use—AI products that solve clear jobs: writing code, answering support tickets, drafting legal text, or powering voice assistants. Many leaders here remain private. Public exposure exists through a few focused application companies and through established software firms that have embedded AI into their suites. Risk and reward vary widely: pure-play applications can be volatile; incumbents can be steadier but may grow slower.
Tier 6: Traditional companies adopting AI
Legacy firms are upgrading operations with AI to cut costs and boost output. Examples include manufacturers improving factory throughput, retailers optimizing inventory, and banks enhancing fraud detection. This can be the lowest-risk way to benefit from AI, since the core business already works and AI expands margins.
Term explained: margin. The share of revenue left after costs—higher margins often mean stronger profitability.
Three simple portfolio postures to study
- Conservative: More weight to energy infrastructure and data center REITs, plus broad AI ETFs and one or two large platforms. Lower growth, steadier cash flows.
- Balanced: Mix of leading chips, one or two platforms, one infrastructure name, and a broad ETF. Aiming for growth with diversification.
- Aggressive: Concentrated positions in chips, select infrastructure, and application pure plays. Highest upside and drawdown risk.
Key risks to track
- Valuation bubbles in popular names
- Antitrust actions against large platforms
- Rapid technology shifts that change model or chip leadership
- Supply constraints in power and advanced chip manufacturing
- Interest rate moves that affect REIT valuations and growth stocks
The big picture
Estimates suggest the AI economy could reach several trillion dollars within a decade. Chips, power, and data centers build the base. Models and tooling make AI usable. Applications and incumbents capture value in workflows people pay for every day. For most public investors, a diversified mix across tiers can reduce single-point failure while keeping exposure to growth.
For more on Felix Prehn and Goat Academy, readers can learn about his work here: About Felix Prehn Goat Academy.