$205B Google, $200B Amazon — AI Capex 2026

The four largest hyperscalers — Amazon, Microsoft, Google, and Meta — plan to spend roughly $725 billion on AI infrastructure in 2026, a 77% increase from approximately $410 billion in 2025. This spending encompasses data center capex, acquisitions, minority investments, compute contracts, and talent across the industry's biggest players.

Amazon leads the charge with approximately $200 billion in 2026 capex, focused on AWS data centers, Trainium chips, and Bedrock. Microsoft follows closely at around $190 billion, more than doubling its record fiscal 2025 capex of roughly $80 billion, with spending directed toward OpenAI infrastructure, MAIA chips, and Copilot scaling. Alphabet raised its ceiling to $205 billion at its Q2 2026 earnings call, driving a 94% increase from roughly $91 billion in 2025, with resources allocated to TPU v7 development, Gemini training, and Google Cloud AI. Meta has raised guidance twice, now targeting $125–145 billion for 2026, a 74% jump from its $72 billion in 2025 capex, primarily for Llama training and superintelligence research.

Beyond the headline capex figures, broader AI spending — including infrastructure capex, acquisitions, minority investments, structured deals, and talent — grew dramatically from approximately $30 billion in 2020 to $381 billion in 2025, a 12-fold ramp in five years. Seven companies tracked across this period — Google, Amazon, Meta, Microsoft, Tesla, NVIDIA, and Apple — spent roughly $764 billion on AI from 2020 through 2025. When adding 2026 guidance of approximately $725 billion from the big four alone, cumulative AI spending exceeds $1.5 trillion before accounting for the $500 billion Stargate project.

Over the past six years, Amazon and Google have led on cumulative totals, each pursuing distinctly different strategies. Amazon has emphasized cloud infrastructure and model partnerships, accumulating $205 billion in AI spending with a 69% five-year compound annual growth rate. Google's full-stack vertical integration approach generated $190 billion cumulative spend with an 84% CAGR. Microsoft's partnership-first model drove $149 billion in spending with a 101% CAGR. Meta's open-source and infrastructure-heavy strategy totaled $141 billion with a 74% CAGR. NVIDIA, operating on a fabless model with R&D focus, spent $43 billion with only a 6% CAGR, while Apple allocated $20 billion toward on-device and stealth AI, and Tesla invested $15 billion in specialized autonomous AI.

Infrastructure — data centers, GPUs, and custom silicon — dominates every year, representing approximately 80% of all tracked AI spending. However, 2025 saw a historic surge in external deals. Google acquired Wiz for $32 billion; Meta signed $41 billion in compute contracts with Nebius and CoreWeave combined; and Amazon accumulated a $50 billion position in OpenAI. The 15 largest AI deals between 2020 and 2025 reveal the scale of these commitments. Amazon's cumulative $50 billion stake in OpenAI ranks as the single largest tracked acquisition-equivalent deal, followed by Google's $32 billion Wiz acquisition and Meta's $27 billion five-year Nebius compute contract. Microsoft's $19.7 billion acquisition of Nuance Communications, Meta's $14.3 billion stake in Scale AI, and Meta's $14.2 billion six-year CoreWeave contract round out the top tier, alongside historical stakes and structured deals with OpenAI, Anthropic, Mellanox, Mandiant, and others totaling over $1 billion each.

Stargate, announced in early 2025, represents the single largest private infrastructure investment in AI history. This joint venture between OpenAI, SoftBank, and Oracle will deploy $500 billion to build an AI data center network across the United States over 4–5 years, operating independently of hyperscaler capex budgets. SoftBank provides the majority of financing, Oracle supplies cloud infrastructure, and OpenAI serves as the primary compute consumer, with initial campuses planned for Texas.

Industry consensus by mid-2026 has shifted toward viewing AI capex as a permanent fixture in big tech cost structures rather than a temporary cycle. Proponents point to concrete returns: Microsoft Azure is growing at approximately 40%, Meta's AI ad targeting is boosting revenue, and Google Cloud expanded 82% in Q2 2026. Skeptics counter that $725 billion in annual spending still outpaces demonstrable AI revenue, assumptions around GPU depreciation face scrutiny, and the arms race shows no signs of deceleration. Regardless, Nvidia remains the primary beneficiary, with its H200, B200, and GB200 GPUs powering most AI training and inference across every major hyperscaler, capturing roughly 40% of hyperscaler capex. Secondary winners include TSMC for fabrication, Arista for networking, Vertiv and Eaton for power and cooling infrastructure, and emerging AI cloud providers like CoreWeave and Nebius scaling rapidly to meet demand.