The four largest technology companies in the world — Amazon, Alphabet, Meta, and Microsoft — are on track to spend a combined $650 billion on artificial intelligence infrastructure in 2026, a figure that represents a 67% increase over the $381 billion they spent collectively in 2025 and a level of capital expenditure without parallel in corporate history this century.¹² Bloomberg described the scale of the spending as "a boom without a parallel this century," and for each of the four companies, their 2026 capex budget is expected to match or surpass their total investment across the previous three years combined.¹

The numbers company by company

Amazon has committed approximately $200 billion in capital expenditure for 2026, the largest individual sum among the four, driven primarily by AWS data centre buildout and the compute infrastructure required to support its growing AI services business.⁹³ Alphabet, Google's parent company, has guided for capex of $175 billion to $185 billion for the year — a figure that includes substantial investment in new data centres, custom AI chips (TPUs), and the network infrastructure underpinning Google's AI products.⁹⁴

Microsoft is running at an annualised capex pace equivalent to approximately $145 billion for its fiscal year 2026, underpinned by its deep partnership with OpenAI and the infrastructure demands of serving Azure AI customers globally.⁹⁶ Meta has told investors it will spend between $115 billion and $135 billion in 2026, channelled into data centres, AI research, and the compute clusters needed to train and run its Muse Spark model family and power AI features across its apps.⁹⁴

What the money is being spent on

The vast majority of this capital expenditure is being directed at three categories: data centres, semiconductors, and the power infrastructure to run them.²³ Building a modern AI data centre involves acquiring land, constructing specialised buildings capable of handling the cooling loads of thousands of GPUs running continuously, purchasing the chips themselves — primarily from Nvidia, though all four companies are also developing proprietary AI accelerators — and securing long-term energy contracts to power the facilities.⁷⁸

The sheer scale of procurement involved is reshaping supply chains and energy markets. All four companies have been negotiating directly with utilities and in some cases developing their own power generation capacity — including nuclear and solar agreements — to secure the gigawatts of electricity required to keep their AI infrastructure running.⁵⁶ Construction companies, fibre manufacturers, and cooling equipment suppliers have all reported surging order books driven by hyperscaler demand.²

Why the spending is accelerating now

The acceleration from $381 billion in 2025 to $650 billion in 2026 reflects a specific competitive dynamic. Each company fears that underinvesting at this moment — when foundation model capabilities are advancing rapidly and enterprise AI adoption is reaching an inflection point — will leave it permanently behind rivals who locked up compute, talent, and customer relationships during the crucial 2025-2027 window.¹⁸

The logic is partially self-reinforcing: more compute enables more capable models; more capable models attract more enterprise customers; more enterprise customers justify more compute investment. All four companies are simultaneously racing to establish dominant positions in enterprise AI before the market consolidates around a small number of preferred providers.³⁷

The scrutiny the spending is attracting

Not everyone is convinced the bet will pay off at this scale. Investors have pressed all four companies' management teams on whether the return on invested capital from AI infrastructure will materialise fast enough to justify spending levels that each company's finance chiefs acknowledge are extraordinary even by their own historical standards.⁵⁹

The debate is sharpest at Meta, where the top end of the $115-135 billion 2026 capex range represents a near-tripling of the company's infrastructure investment from just three years ago.⁴⁹ Critics argue that until AI-generated revenue streams grow substantially faster than AI infrastructure costs, the companies are running an arms race where the primary beneficiary is Nvidia. Advocates counter that failing to invest now would be the more costly long-term mistake — and that the revenue from cloud AI services is already growing sharply.²⁸

#big tech

The $650 Billion Bet: How Amazon, Google, Meta, and Microsoft Are Reshaping the Global Economy With Their 2026 AI Spending