Bloomberg Tech→ original

Meta, Amazon, Microsoft and Alphabet to Spend $750 Billion on Capex — More Than Ireland's GDP

Bloomberg estimates that Meta, Amazon, Microsoft and Alphabet will spend about $750 billion on capital expenditures this year. That is more than the GDP of Ireland or Belgium, and more than half of what all S&P 500 companies invested last year. Split evenly, it comes to $5,500 for every US household. In Businessweek Daily, Bloomberg reporters Ian King and Lynn Doan break down what goes into the price of a single data center, from land to chips. • $750 billion: the four companies' capex for the year. • More than half of last year's capital spending by the entire S&P 500. • The key question: how much of this money goes to chips.

AI-processed from Bloomberg Tech; edited by Hamidun News
Meta, Amazon, Microsoft and Alphabet to Spend $750 Billion on Capex — More Than Ireland's GDP
Source: Bloomberg Tech. Collage: Hamidun News.
◐ Listen to article

About $750 billion. That is how much Meta Platforms, Amazon, Microsoft and Alphabet, Google's parent, will spend on capital expenditures this year, according to Bloomberg's estimate. The sum is larger than the GDP of Ireland or Belgium. It exceeds half of what all companies in the S&P 500 index combined spent on capital investment last year. Split evenly, it would be a check for $5,500 to every household in the US.

Bloomberg breaks this money down to the level of a single facility. In an episode of Businessweek Daily, semiconductor reporter Ian King and managing editor for technology Lynn Doan explain what goes into the cost of one data center, from land to chips. The episode announcement gives no detailed line-by-line breakdown, so below we look at the scale of the numbers and the logic behind them.

A Sum That Is Hard to Compare

The comparisons in the piece are chosen to give readers a sense of the order of magnitude. The GDP of two developed European countries. More than half of the annual capex of the 500 largest publicly traded US companies, a group that includes oil producers, utilities, industrial firms and telecom giants. Check the math: $750 billion divided by $5,500 comes to roughly 136 million households. That is the right order of magnitude for the US, so the comparison holds up.

One important caveat: the sum covers only the four named companies. Other major builders of AI infrastructure are not included, so total industry spending on computing capacity is higher.

From Land to Chips: What Goes Into the Bill

The title of the episode sets the order of the cost items. First comes the site and the grid connection. Then the building, cooling systems and electrical equipment. Finally, the servers with accelerators and the networking infrastructure. By most public estimates, computing hardware takes the largest share of the budget, but Bloomberg gives the exact proportions for specific sites in the episode itself, not in the announcement.

A few reference points are enough to grasp the scale:

  • $750 billion — expected annual capex of Meta, Amazon, Microsoft and Alphabet;
  • more than 50% — share of last year's capital investment by all S&P 500 companies;
  • $5,500 — the hypothetical check for every US household;
  • 4 companies — that is how many stand behind this number.

The main takeaway: the cost of a data center is no longer determined by land and concrete. Access to electricity and chip supply now decide it, and both resources are limited.

What to Watch Next

Investors and market watchers should keep three things in view.

1. Capex guidance in quarterly reports. Any raise or cut in the four companies' targets directly affects the revenue of suppliers of chips, servers and electrical equipment. 2. The spending mix. If the share of hardware grows, accelerator makers win. If the share of energy and real estate grows, utilities and builders win. 3. Payback. Investments this large require matching revenue from AI services. For now, that revenue is noticeably smaller than the spending, so the market will be watching closely to see when the money comes back.

Startups and developers who depend on cloud GPUs should consider the other side. The more the giants invest in capacity, the better the chance of cheaper compute. But the same demand for chips and electricity may keep prices up, so it is wiser to plan inference budgets with a cushion.

ZK
Hamidun News
AI news without noise. Daily editorial selection from 50+ sources. A product by Zhemal Khamidun, Head of AI at Alpina Digital.
✦ Enterprise AI & Advisory

Executive AI Consulting, Custom Agents & Team Workshops

Work directly with Zhemal Khamidun at hamidun.com — enterprise AI architecture, production RAG & multi-agent systems, and executive training.

Book AI Consultation →Explore NomadNet AI
What do you think?
Loading comments…