AI has changed from an imaginary ideation to a very real factor in investment discussions, corporate earnings, and positioning yourself competitively in the market. Starting in 2023, market earnings connected to AI have been largely condensed in a very small group of US based technology stocks, although there are multiple factors which decide the pricing of the shares. The speed of this shift has been rapid, and some investors describe it in multiple phases, viz. hardware, infrastructure, software, and adoption, though the boundaries are subjected to overlapping.
Every great technological revolution, be it- electricity, the internet, mobile phone was followed a recognisable arc. AI is doing the same. The question is not whether you believe in it. The question is: where are we in the cycle, and what does that mean for your money?

Phase 1: The True Enablers of AI
The first phase starts with chips. Companies which design and manufacture processors, advanced chips, and the equipment required for creating these chips, these companies are examples of the hardware layer which is the core of AI functioning.
In 2023-24, the market saw them as growth stocks, but they are very capital-intensive businesses which are heavily tied to the AI supply chain.
But Phase One carries its own particular danger: overbuilding. History rhymes here. The late nineties saw billions poured into fibre optic cable, so much that the resulting glut took nearly a decade to absorb. When everyone races to build the same infrastructure, supply eventually catches demand, then surpasses it. Margins compress. The stocks that tripled on the way up can give back a painful amount on the way down.
The investor who enters Phase One early and exits before the glut is the investor who does well. The investor who confuses a secular trend with a permanent growth rate is the one who learns the hard way.

Phase 2: Building the infrastructure

Once the chips come into existence, they need large amounts of energy and data storage capacities. Companies which provide large cloud storages have significantly increased their expenditure on data centers and AI infrastructure.
Energy availability and grid capacity are becoming real bottlenecks in certain parts of the world. Training large AI models demands significant power, though exactly how much depends on the model's size, the hardware being used, and how long the run takes — making headline comparisons tricky without knowing the underlying assumptions. That growing demand could channel investment toward clean energy, storage solutions, and grid modernisation, though how much actually materialises will depend on the regulatory environment, available financing, and the specific limitations of local infrastructure. These assets can still be volatile and sensitive to rates, regulation and sector risks.

Phase 3: The application wave

The application wave is where the wealth creation becomes most broadly distributed and most difficult to predict. In Phase 1, the winners are relatively obvious: whoever makes the hardware the AI runs on. In Phase 2, they are somewhat obvious: whoever builds the best platforms. In Phase 3, the winners could be in any industry. A healthcare software company that silently incorporates AI into its clinical workflow becomes drastically more valuable. A legal technology firm that automates document review builds a moat overnight. A logistics platform that uses AI for route optimisation starts compounding in ways its historical growth rate never suggested possible.
But Phase Three also produces the most visible failures. Not every application works at scale. Not every market is willing to pay what the model requires. The graveyard of the internet era is full of brilliant applications that arrived at the right moment but in the wrong market, or with the right idea but the wrong distribution. The same will be true here

Phase 4: The productivity harvest

Phase Four is not a product launch or a new capability. It is a slow, economy-wide recalibration of what it costs to do almost everything. When electricity became ubiquitous, the most consequential change was not a new electrical product. It was that every product, in every industry, became cheaper to make. The whole productivity floor shifted upward. Returns compounded quietly for decades.
AI's productivity harvest will look something like this. Not AI companies specifically outperforming, but every company in every sector running leaner, moving faster, and serving customers better because AI has become woven into the operational fabric of how business is done. The cost of writing code drops. The cost of customer service drops. The cost of analysis, translation, design, research, all of it compresses. Margins expand for companies that absorb these gains efficiently. Those margin expansions flow through to equity returns over time.

The question beneath the cycle
Understanding the four phases is useful. But there is a more important question sitting underneath all of it: which phase are you actually in right now?
The honest answer is that most markets are living through several phases at once. The infrastructure wave is not over. The enabler ecosystem is still taking shape. The first real applications are beginning to find their footing. And in small pockets of the economy, the productivity harvest has already started, quietly, without much fanfare.
That is the whole tension of investing in a technology cycle while it is still playing out. The investors who come out ahead are rarely the ones who perfectly called the top of Phase One or caught the bottom of Phase Two. More often, they are the ones who had a clear sense of where they were, sized their bets accordingly, and simply had the stomach to sit through the noise in between.