Alphabet повышает прогноз капзатрат на 2026 год до $205 млрд ради гонки ИИ

On July 22, 2026, Alphabet holding company raised its 2026 annual capital expenditure guidance to $195–205 billion, lifting an already record-high estimate for the sake of its edge in the AI race. This was reported by Bloomberg Tech.

How much will Alphabet spend in 2026

Alphabet has set its 2026 capital expenditure guidance in the range of $195 billion to $205 billion. This refers to capex — money for building and equipping data centers, purchasing servers and AI accelerators, and expanding computing infrastructure. The previous estimate for the year was already considered record-breaking, and the new guidance raises it even further. Alphabet is the parent company of Google, YouTube, and Google Cloud, and almost this entire budget goes toward computing for artificial intelligence tasks.

*Alphabet's new 2026 capex guidance is $195–205 billion

*The estimate has been raised relative to the previous, already record-breaking plan

*Even the lower end of the range is unprecedented for the company

*The goal is to cement its edge in the artificial intelligence race

*Alphabet is the parent company of Google, YouTube, and Google Cloud

Where the capital expenditures go

Alphabet's capital expenditures go primarily toward the physical infrastructure of computing. The main line items are building and equipping data centers, purchasing servers and AI accelerators (GPUs and Google's own tensor processing units, TPUs), as well as power supply and cooling systems for these facilities. Unlike operating expenses, capex is a long-term investment in assets that the company depreciates over years. The $195–205 billion range means that even the lower bound of the 2026 plan is an unprecedented sum for Alphabet, directed almost entirely at scaling up for AI. The $10 billion spread between the bounds of the guidance is itself comparable to the annual capital budgets of many large corporations.

Why AI spending is growing

Alphabet is increasing capital expenditures to maintain its edge in the artificial intelligence race — that is how the company itself frames the reason. Training and serving large language models requires enormous computing power: accelerators, electricity, cooling, and data center space. Computing has effectively become a scarce resource that the largest technology companies are competing for, and whoever builds out infrastructure faster gains an edge in the speed of training and launching new models. The $195–205 billion range for 2026 alone shows that the infrastructure race has reached a new scale, and investment in computing has become a key front of competition.

The company is raising its already record-high spending guidance to cement its edge in the artificial intelligence race,

Bloomberg Tech reports.

What this means

The $195–205 billion guidance is a signal that for AI market leaders, capital investment in computing has become the main front of competition. Sums like these raise the barrier to entry: only companies willing to spend tens of billions of dollars a year on infrastructure can compete in training frontier models. At the same time, pressure on profits is growing — capex of this scale is depreciated over years, and investors are watching ever more closely for when these enormous AI investments will start paying off in revenue. The mid-2026 guidance increase shows that Alphabet has no intention of slowing down and is prepared to keep raising its spending bar.

Frequently asked questions

How much does

Alphabet plan to spend on capital expenditures in 2026?

Alphabet is forecasting capital expenditures in the range of $195 billion to $205 billion for all of 2026 — a raised estimate relative to the previous, already record-breaking plan.

What are capital expenditures (capex) and what does

Alphabet spend them on?

Capex is long-term investment in assets: for Alphabet, they go mainly toward building data centers, purchasing servers and AI accelerators, and power supply and cooling — that is, toward the infrastructure for training and running AI models.