The investment cycle in artificial intelligence infrastructure is gaining unprecedented momentum. According to my analysis based on MacroMicro data, the combined capital expenditures (CapEx) of the four largest cloud service providers (CSPs)—Amazon, Microsoft, Alphabet, and Meta—will exceed the $700 billion mark by 2026. This is not just a new record, but a clear signal of the beginning of a supercycle in the semiconductor industry.

The latest earnings season only confirmed this trend. Instead of the expected peak in AI investments, we see hyperscalers revising their forecasts upward. They are increasingly investing in data centers, developing their own chips, and computing infrastructure. This indicates that companies see far more monetization opportunities than previously anticipated.

Who is spending how much?

Amazon (AWS) remains the largest investor with projected costs of around $200 billion. The funds will go toward expanding its data center network and developing its own Trainium and Inferentia chips. Microsoft (Azure) has raised the bar to approximately $190 billion, deepening its partnership with OpenAI and expanding infrastructure for Copilot. Alphabet (Google Cloud) has increased its target to $180–190 billion, focusing on the TPU network and developing Gemini models. Meta, in turn, has significantly raised its CapEx target to $125–145 billion, directing resources toward training open-source Llama models and improving recommendation algorithms.

Tesla has also joined the race, raising its 2026 target to over $25 billion. These funds will go toward deploying AI servers and expanding the Dojo supercomputer. Apple and NVIDIA are pursuing different strategies: the former is betting on integrating AI into its own chips and devices, while the latter is focusing on developing next-generation chips and investing in AI providers such as CoreWeave.

Impact on the supply chain

Behind the sharp rise in semiconductor sales lies the entire value chain in AI computing. The top tier includes advanced chip manufacturing (TSMC), CoWoS packaging, and HBM memory from SK Hynix. The middle tier consists of the cloud providers themselves, linking suppliers with AI developers. The bottom tier comprises large language models, which are the ultimate source of demand. Both training and inference require enormous computing power, sustaining steady demand across the entire chain.

Global semiconductor sales show a strong positive correlation with the capital expenditures of tech giants, and both indicators are hitting new records. A key indicator to track for a reversal is the AI monetization coverage ratio, which compares potential revenue to costs. According to my calculations, the ratios for Meta (1.8), Microsoft (1.5), Google (1.1), and Amazon (0.7) confirm that these companies' investments are justified.

My expert opinion: The market is entering a phase where the volume of investments becomes self-sustaining. As long as the monetization ratio remains above one for most players, the semiconductor supercycle will continue. However, investors should closely monitor the dynamics of this ratio—a drop below 1 would be the first warning sign of a correction.