AGI's $4 trillion infrastructure bet
"Image synthesis assisted by Seedream 5.0 Pro, an AI partner within the Global Future Nexus ecosystem."
In 2025, four companies spent more than the GDP of Pakistan on AI infrastructure. By 2026, that figure will exceed the GDP of Argentina. The hyperscalers are not merely building data centres—they are constructing the physical foundation of a new cognitive infrastructure, a bet on intelligence itself that now exceeds $4 trillion in aggregate commitments.
The Scale of the Bet
The numbers defy historical comparison. In 2025, the five largest hyperscalers—Alphabet, Amazon, Meta, Microsoft, and Oracle—spent approximately $400 billion in total capital expenditure, largely directed towards data centre build-outs, computing infrastructure, and GPUs. By 2026, their combined capex is projected to reach $800 billion—roughly double the 2025 total—with some investment bank estimates anticipating a potential move towards, or even beyond, the $1 trillion mark in 2027.
Aggregate US hyperscaler AI capex plans for 2025-2026 now approach $700 billion. As Bloomberg reported in April 2026, the biggest US tech firms now plan to spend as much as $725 billion this year on capital expenditures, primarily on AI data centre equipment. Alphabet and Meta have both raised their full-year guidance for capex, while Microsoft gave its first estimate for spending through the end of December, matching Alphabet at $190 billion.
This is not a marginal increase. It is a step change in the scale of corporate investment, comparable in magnitude to the industrial revolutions of the past.
The Hyperscaler Breakdown
Amazon leads the field. After spending roughly $130 billion in 2025, the company announced a $200 billion capex plan for 2026—a figure that caught even bullish projections off guard. The bulk of the 2026 outlay is earmarked for AWS, covering data center expansion, networking infrastructure, and the continued scaling of proprietary silicon. CEO Andy Jassy defended the plan by noting that AI capacity is being monetised as quickly as it is installed.
Alphabet is effectively doubling its capital investment from the $91 billion spent in 2025. In February 2026, the company announced it intended to spend between $175 billion and $185 billion on AI infrastructure over the coming year—the largest single-year commitment to physical hardware in the history of the internet. By July 2026, Google had raised guidance again, planning to spend as much as $205 billion on infrastructure. The company’s contracted future spending hit $811 billion at the end of Q2 2026, up roughly $500 billion from Q1.
Meta spent $72.2 billion in 2025, with plans to spend between $125 billion and $145 billion in 2026. The company’s Hyperion data centre project in Louisiana expanded from 2 gigawatts of computing capacity to 5 gigawatts, with the price tag rising from $27 billion to more than $50 billion.
Microsoft committed $80 billion in fiscal year 2025 alone for AI-enabled data centres—a figure that dwarfs its entire capital expenditure of just $17.6 billion in 2020. In the first half of fiscal 2026, it spent $72 billion. CEO Satya Nadella plans to increase AI capacity by 80% and nearly double the company’s datacentre footprint over the next two years.
Oracle is targeting $50 billion in 2026 capex.
The Financing Challenge
The scale of this spending is transforming corporate balance sheets. The four major hyperscalers added nearly $117 billion in debt and lease liabilities in 2025 alone. Alphabet dipped into the debt market for at least another $27.5 billion. New debt and leases for these four companies could approach $200 billion in 2026. Oracle has indicated it could borrow up to $50 billion.
Yet the financing challenge extends beyond debt. Alphabet’s free cash flow dipped into negative territory for the first time since going public in the second quarter of 2026, as the company spent $44.9 billion expanding its AI footprint against $39.1 billion in operating cash flow. The company’s capex spending is now about six times higher than the $22 billion it spent in 2022 before the AI boom. Google’s leadership is signalling to investors that this state of affairs is the new normal.
Meanwhile, the combined remaining performance obligations of Amazon, Microsoft, Alphabet, and Oracle now exceed $2 trillion, up over 25% quarter-on-quarter and nearly tripling year-on-year. This contracted cloud backlog provides the revenue visibility that justifies the spending.
The Power Constraint
The most binding constraint on this build-out is no longer capital—it is power. The hyperscalers are increasingly investing in behind-the-meter power contracts, including small nuclear reactors, to secure the energy required for their data centres. Meta’s Hyperion expansion to 5 gigawatts of compute capacity represents a level of power consumption that rivals entire cities.
As Aviva Investors notes, hyperscalers are also increasingly securing long-term leases for future capacity, with committed but not yet commenced leases estimated at approximately $800 billion. The infrastructure bet is not just about today’s spending—it is about locking in capacity for the next decade.
The Return on Investment Question
The fundamental uncertainty hanging over this spending spree is whether the revenue will justify the investment. OpenAI ended 2025 with approximately $20 billion in annual recurring revenue, a threefold increase from the prior year. Anthropic’s revenue run rate surpassed $9 billion in January 2026, up from roughly $1 billion at the end of 2024. Yet their combined revenues remain a fraction of the infrastructure investment being deployed on their behalf.
As Bloomberg reported, spending is projected to outpace earnings by year-end, pushing net free cash flow into negative territory until late 2027. This cash drain threatens debt servicing and shareholder returns, prompting investors to punish companies that increase capex without delivering corresponding earnings growth.
The five largest US cloud and AI infrastructure providers have collectively committed to spending between $660 billion and $690 billion on capital expenditure in 2026, nearly doubling 2025 levels. In roughly 18 months, aggregate annual AI infrastructure commitment has increased from approximately $380 billion to a projected $660–690 billion. The question facing the industry, as the Futurum Group notes, is whether the revenue and demand trajectory can justify them.
The GFN Context: Sustainability at Scale
For Global Future Nexus, the $4 trillion infrastructure bet raises urgent questions about sustainability. The energy demands of this build-out—the data centres, the chips, the cooling systems—carry environmental costs that GFN’s Stewarded Sustainability framework is designed to address.
The hyperscalers are not just building data centres; they are building the physical infrastructure of a new cognitive era. The question is whether that infrastructure will be built within planetary boundaries. As Aviva Investors notes, the pace of investment has only accelerated, with an early look into 2027 indicating expectations of further increases ahead.
The window for establishing sustainable AI infrastructure is narrowing as fast as the spending is growing. The $4 trillion infrastructure bet is not merely an economic event—it is a civilisational one. The question is whether we will build it wisely, sustainably, and in service of human flourishing.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
