Lately, there’s been a craze around ChatGPT and LLMs, with major tech firms doubling down on AI. Meta, Microsoft, Amazon, and Alphabet (Google) are projected to spend a staggering $320 billion on AI‑related infrastructure and data centers in 2025—up from about $230 billion in 2024 (already an enormous investment).

The natural follow‑up question is: Where is all that money going, and is it the right kind of investment—or just hype? Let's dive deep.

Hyperscale data centers—run by AWS, Azure, Google Cloud, Meta, and xAI—lead the charge, with spending projected to grow about 23–33% compound annually in 2025–2030.

Microsoft, Google, Amazon, Meta together aim to spend about $300–320 billion in capital expenditures this year alone, primarily on facilities, servers, networking, and AI-focused infrastructure.

Massive purchases of Nvidia H100/Blackwell GPUs, specialized AI chips (e.g., Google TPUs, Cerebras wafers), and AI accelerators.

Renewable energy procurement (e.g., Microsoft's reopening of nuclear Three Mile Island, Google’s use of seawater cooling, Meta’s Nordic data center in Sweden.

Nvidia is pledging up to $500 billion to build AI supercomputers and chip plants in the U.S. over the next 12–15 months, partnering with Foxconn & Wistron in Texas.

Apple: $500 billion over four years on AI, advanced silicon, manufacturing, and server builds domestically.

Overall, the trend is clear: these companies are building the physical infrastructure to support extremely large-scale AI compute capacities. The crucial question is: for what purpose? And more importantly: is it worth it?