When I added the years to these 5 layers, this is what formed in my mind. Energy and chips are already active, infrastructure is currently being built, models will mature by 2027, and applications will consolidate economic value in 2028-2030. This isn't a sector rotation, it's an accumulation that comes in waves. And the blockage of a single layer, each layer dependent on the one below it, means the entire stack slows down. Trillions of dollars worth of infrastructure hasn't even been built yet. Look closely at this picture.

Layer 1: Energy (2025–2028)

It all starts with energy. Artificial intelligence produces real-time intelligence, and this production relentlessly consumes electrons. Data centers no longer just store data; they function as intelligence factories. Understanding this difference is critical because a storage facility can shut down, but a production factory cannot. I've considered $CEG, $VST, $BE, $CCJ, $GEV. Nuclear, natural gas, and renewables are the only resources that can meet AI's energy hunger. $BWXT and $PWR are the companies building this energy infrastructure. The market still partially prices this layer as a "supporting sector." But as the question of who will pay AI's electricity bill becomes clearer, this valuation will be redefined. Since SMR will enter the picture after 2028, I think a reassessment is needed for 2028.

Layer 2: Chips (2023–2028)

If energy is the raw material, chips are the engine that transforms it into intelligence. This layer is actually the earliest to mature, and we've been seeing this since 2023. $NVDA already created a sector with a clear winner, but $AMD is trying to offer an alternative. $ARM architecture is spreading to mobile and edge. $AVGO plays a critical role in hyperscalers designing their own chips. But the real unsung heroes are $ASML, $MU, $ALAB, $TSM, and $COHR. Without ASML, none of these chips can be produced; it's a real monopoly. $MU continues to be the winner in the high-bandwidth memory AI segment. $ALAB and $COHR are solving the problem of how tens of thousands of GPUs communicate with each other within the data center. $CRDO, in which I am also an investor, is also involved, but I can't list all the companies.

Layer 3: Infrastructure (2025–2028)

This layer is where the discussion is most complex. $AMZN, $GOOGL, $ORCL, $MSFT are already there, but they're not alone; a lower class of smaller players like $NBIS, $IREN, $DOCN, $CIFR, and $CRWV go on and on. These companies play the game a little differently. Hyperscalers do everything, but sometimes they don't do anything perfectly. $CRWV offers pure GPU capacity and, along with $ORCL, is driving the industry crazy with aggressive borrowing. $DOCN plays a different role, a company that leases GPUs to smaller companies and also runs a production conference platform. The reason $CIFR and $IREN appear in this layer is because of their rapid deployment of energy and infrastructure; they don't have a particularly extraordinary story. The infrastructure layer is not yet clear, and this uncertainty creates both a great risk and the greatest opportunity for NeoCloud.

Layer 4: Models (2025–2029)

The most interesting aspect of this layer is that it is represented by only four names: $AMZN, $GOOG, $NVDA, $TSLA. As open-source models reach the frontier, the economics of this layer completely changes. When a powerful reasoning model becomes free, the demand for the layers below doesn't decrease, it increases. Because application developers build faster, more products are released, and more reference demand arises. Now you might ask what Tesla is doing in this layer, but considering it trains its own models for robotics and physical AI, its presence here makes more sense to me.

Layer 5: Applications (2028–2030)

My favorite layer. Economic value is created here, but this layer matures the latest. $GOOGL, $NET, $PLTR, $DOCN, $SNOW, $FSLY. Among these, $DOCN's dual-layer presence in both infrastructure and application layers reflects the company's unique position in the developer ecosystem. $PLTR is one of the rare companies that has found true product-market fit in defense and enterprise AI. As for $NET, I think the market hasn't fully grasped how strong its position is when it combines edge computing with AI. $META, $AXON, and $APP are included here, but I haven't considered them because they have reached a certain level themselves.

The information, comments and recommendations contained herein are not within the scope of investment consultancy. Investment consultancy services are provided within the framework of the investment consultancy agreement to be signed between brokerage firms, portfolio management companies, banks that do not accept deposits and customers. The comments in this article are only my personal comments and these comments may not be appropriate for your financial situation and risk return. For this reason, investments should not be made based on the information and comments in my articles.