How do we interpret the news of Google’s massive spending on data centers… without getting lost in the numbers?

The news seems simple: Alphabet (the parent company of Google) states that it will increase its capital expenditure in 2026 to $175–185 billion for building data centers and purchasing servers and equipment. However, the cognitive value lies not just in the number, but in the “story” it hints at: artificial intelligence is no longer just a program—it has become an infrastructure industry akin to electricity networks, roads, and ports: it requires land, energy, and supply chains, and its costs are paid today to yield returns later.

This article is not to provide an opinion on Google’s stock. Its goal is simpler and more important: to provide a lens that explains any similar news—today with Google, tomorrow with Microsoft, Amazon, or even a local company building computing capabilities.

The Big Idea: “Computing” is the New Fuel

When you hear “artificial intelligence,” you might envision an application or a chatbot. But the reality is that the intelligent model needs something very costly before it “speaks”: computing power.
And computing power ultimately means:

  • Data Centers (equipped buildings)
  • Servers and Processors (like Nvidia’s GPUs or Google’s TPUs)
  • Networks that connect all of this quickly
  • Continuous Electricity and Cooling

Therefore, when a company says it will spend nearly $180 billion, it is practically saying: we will expand the computing factory.

Translating Terms Without Complexity

1) What is “Capital Expenditure” CapEx?

It is the money spent on assets used for years: building a data center, purchasing servers, extending networks…
Not like salaries or monthly electricity bills. Those are usually called Operating Expenses (Opex).

2) What does “Rapidly Depreciating Assets” mean?

Some assets “last” less or become obsolete faster—like servers.
Accountingly, the cost of the server is not recorded as a full expense on the day of purchase; rather, it is spread over the years of its use. This is depreciation.
So you will hear companies say: “A large part of CapEx will go to rapidly depreciating assets,” meaning: servers will appear as a cost on profits over several years, but they consume cash from now.

3) GPU and TPU: Why are they frequently mentioned?

These are “heavy” processors designed to efficiently train and run artificial intelligence models.

  • GPU: Common in the market and often from Nvidia.
  • TPU: Chips developed internally by Google to reduce costs and improve performance for certain workloads.

What is Google Actually Saying in Numbers? (A Very Small Numbers Section)

  • The annual CapEx target for 2026: $175–185 billion.
  • Management stated that about 60% of this spending will go to servers (rapidly depreciating assets), and 40% to construction and networks for new data centers.
  • In the fourth quarter results, the company reported that advertising revenues from search and YouTube and the ad network exceeded $82.28 billion, and cloud revenues reached $17.66 billion with strong annual growth.

(The numbers here are not for boasting, but to solidify the idea: growing demand drives companies to expand the “computing factory”).

Why Might Investors Be Concerned Despite Huge Profits?

It’s natural to say: “A company making tens of billions… why is anyone worried?”
The reason is that investors do not look at profits alone; they look at cash. Because a company may report good profits, but it is spending massive cash on expansions that make free cash flow lower.

The idea is simple:

  • Profits are calculated according to accounting rules that include depreciation and distribution over the years.
  • Cash goes out immediately when building a data center and purchasing servers.

And here appears the new “artificial intelligence equation”:

Will this spending today translate into strong revenue growth tomorrow… or will it turn into a heavy cost if demand cools?

There is no single answer; but there is a way to read.

“Non-Technical Constraints”: Electricity, Land, and Supply Chains

The most educational part of such news is that it reveals that the challenges of artificial intelligence are not all algorithms.
When a CEO says that what worries him is “expanding computing capacity with constraints of energy, land, and supply chains,” it means:

  • Electricity: Running data centers requires massive energy and high stability.
  • Land and Permits: Building new centers is not a “turn on” button.
  • Supply Chains: Servers, processors, and networking equipment are not available without limits.

These constraints make expansion “economic and logistical” before it is “technical”.

Where Will the Returns Come From? From the Cloud… and from Ads Too

It’s easy to assume that returns will come only from “selling AI services.” But major companies have two engines:

  1. Cloud:
    Companies and institutions pay for computing, storage, and artificial intelligence as a service. If demand continues to grow, data centers become a recurring source of income.
  2. Advertising:
    If Google uses models like Gemini to improve understanding of “search intent” and ad relevance, it may increase the value of each search (indirectly), even without the user paying a single riyal.
    And here is an important idea for the reader:

Artificial intelligence is not always a “separate product”; it may be an “enhancer” of the core revenue engine.

Practical Box: How to Read Any Similar News?

The next time you read: “A company will spend 50 or 150 or 200 billion on artificial intelligence,” ask these six questions:

  1. Why now? Is there clear growth in the cloud or enterprise demand?
  2. Where is the spending going? Servers (rapidly depreciating) or buildings and networks?
  3. What is the impact on cash? Is free cash flow temporarily decreasing?
  4. Is the return linked to a strong revenue driver? (like advertising at Google)
  5. What are the mentioned constraints? Electricity/Land/Supply Chains = may mean that expansion is harder than planned.
  6. What indicator will prove the success of the bet? Cloud growth? Margin improvement? Increased advertising revenues?

These questions do not require an economics degree—just make the news “understandable” instead of “exciting”.

Artificial Intelligence Has Become Like Infrastructure Projects

Google’s massive spending encapsulates a new phase: major tech companies are treating artificial intelligence as a productive capacity that must be built, not just a “feature” within an application.
This changes the way we read the news: instead of asking “Is the model smart?” we will ask more:
Do they have electricity? Do they have supply chains? Can they turn computing into sustainable income?