10 posts
- 195Samsung vs SK Hynix: the HBM battle that will decide Korea's AI chip future
- 185TSMC and the AI chip bottleneck Wall Street can't price correctly
- 181The Nvidia market worth watching now isn't data centers — it's robots
- 179Why Did the Space Sector Suddenly Collapse? June's Selloff Explained
- 174Big Tech's Data Center Spending This Year Is Not 725 Trillion Won — It's Dollar-Scale
Big Tech's Data Center Spending This Year Is Not 725 Trillion Won — It's Dollar-Scale
These days, when we talk about U.S. stocks, it all comes down to one word. data center. This is because AI requires a huge computer warehouse to run, and the amount of money spent to build that warehouse is beyond common sense.
Today, rather than speculation, let's summarize where this industry is now with the investment plan numbers for 2026 announced by the companies themselves.
The amount of money the four companies spend per year is 725 trillion won?
Google, Amazon, Microsoft, Meta. The combined amount of money that these four companies announced they will spend as capital investment (capex) in 2026 is about $725 billion, or close to 1,000 trillion won in Korean money. This is a increase of 77% from last year's record of approximately $410 billion.
If you look at each company, the picture becomes clearer. Amazon is the most aggressive at about $200 billion ($125 billion last year), Microsoft is at $190 billion, Google is at $175-185 billion, and Meta is at $125-145 billion. In particular, Google nearly doubled from $91 billion last year. Most of this money goes into data centers and semiconductors that will run AI.
So who receives this money?
The representative place where this released investment money flows is NVIDIA. NVIDIA's data center revenue reached an all-time high of approximately $75.2 billion in the first quarter, up 92% from a year ago. These numbers clearly reveal the structure in which NVIDIA's performance increases as Big Tech purchases GPUs.
What's interesting is that despite pouring this much money, "supply cannot keep up with demand" keeps coming out. Data center operators say they still lack capacity to handle the demand for AI services, and Microsoft expects the 'capacity shortage' to continue at least through 2026.
The current market topic is shifting from "Will AI make money?" to "Can we secure the electricity and chips to run AI in time?" There is.
What should investors look for?
A large number is not necessarily a good sign. On one side, a 'bubble' debate continues over whether this massive investment will one day return with sufficient profits. As seen in the fact that Microsoft attributed $25 billion of the increase in capex to the increase in prices of memory chips and components, it is also burdensome that costs are also rising.
So, when looking at the data center theme, it is not simply "the amount of investment has increased," but how much that investment leads to sales and profits, such as power and components. We also need to look at how bottlenecks drive up costs. It is also worth remembering that the layers of the industry are becoming thicker, including companies that make chips, companies that provide power and cooling facilities, and companies that rent data centers.
What is clear is that data centers are the biggest bet of big tech in 2026. Where this trend ends in profit is likely to be the key story of the stock market in the coming years.
Comments
No comments yet. Be the first!