Hyperscale vendors' 2026 capex totals approximately $725 billion, a 77% year-over-year surge.

Amazon, Microsoft, Alphabet (Google), and Meta are committing a combined approximately $730 billion to AI infrastructure in 2026 — up 78 percent from approximately $410 billion in 2025, marking a buildout scale with no real precedent in enterprise technology spending. Goldman Sachs projects the combined number nearing $1.14 trillion in 2027.

This is not defensive spending or R&D hedging. Every dollar is tied to a specific bet: that AI compute will be the scarce resource determining who wins cloud, enterprise software, and consumer products for the next decade.

**The Figures by Company**

Amazon is guiding to approximately $220 billion in 2026 capex, up roughly 72 percent from 2025. Microsoft has guided to approximately $175 billion, up approximately 48 percent. Alphabet is guiding to $195–205 billion, representing approximately 120 percent growth. Meta is committing $130–145 billion, up approximately 90 percent year-over-year. These 2026 figures reflect the latest full-year guidance as of September 2026, all reaffirmed or raised on Q2 2026 earnings calls. Microsoft's 2026 figure reflects a useful-life accounting change rather than smaller absolute build. Amazon and Microsoft figures represent total company capex.

**What Each Company Is Actually Buying**

Microsoft's $175 billion commitment funds GPU clusters powering Azure OpenAI Service, running GPT-4, o3, and next-generation models at enterprise scale. Microsoft's AI business crossed a $37 billion annualized revenue run-rate, up 123 percent year-over-year. The company is expanding globally with 100+ new facilities across the US, Europe, Middle East, and Asia, while simultaneously investing in custom silicon research through MAIA chips to reduce Nvidia dependency and lower per-token inference costs. Substantial capacity is dedicated to Copilot infrastructure, as every Microsoft 365 user generating AI requests requires dedicated inference capacity.

Alphabet is deploying $195–205 billion toward TPU v5 and v6 clusters — Google's custom AI accelerators now powering Gemini and Google Cloud AI services. Google Cloud revenue hit $24.8 billion in Q2 2026, up 82 percent year-over-year and accelerating from 63 percent growth in Q1. Search AI integration remains a major capital requirement as Google must serve AI Overviews at query scale without margin collapse. Substantial investment supports Waymo compute for autonomous vehicle training, which requires continuous large-scale simulation workloads, and DeepMind research infrastructure for AlphaFold successors and scientific AI programs.

Meta is committing $130–145 billion primarily to MTIA (Meta Training and Inference Accelerator), proprietary chip technology intended to eventually replace Nvidia GPUs for ranking and recommendations. Each Llama generation required clusters of 100,000+ H100s; Llama 5 will require more. The company is building infrastructure to serve its Meta AI assistant to 3 billion+ users across WhatsApp, Instagram, and Facebook, while simultaneously developing Reality Labs compute for VR and AR environments requiring real-time AI rendering at device edge.

Amazon's $220 billion in capex funds AWS data center expansion to serve Bedrock (enterprise AI), SageMaker, and direct GPU rental through P5 instances. AWS revenue hit $42.2 billion in Q2 2026, up 36.7 percent year-over-year — the segment's fastest growth in 18 quarters. Amazon is developing Trainium2 and Inferentia chips to handle significant internal ML workloads, rebuilding Alexa on a foundation model stack requiring substantial inference capacity, and maintaining infrastructure to support AWS's role as Anthropic's primary training cloud with dedicated capacity commitments.

**Where the Money Actually Goes**

Most coverage frames hyperscaler capex as simply "buying Nvidia GPUs." That is accurate but incomplete. A typical AI infrastructure build breaks down across five categories:

GPU clusters account for 40–50 percent of spending. Nvidia H100, H200, and Blackwell GB200 NVL systems cost $30,000–$40,000 per H100 and $60,000–$70,000 per GB200, such that even modest cluster sizes cost billions.

Data center construction represents 25–35 percent of expenditure. New builds in Virginia, Texas, Iowa, and internationally are underway, with Google expanding in 12 countries and Meta breaking ground on a 2GW+ facility in Louisiana. Lead times for purpose-built AI data centers now run 18–24 months.

Networking infrastructure consumes 10–15 percent of budgets. InfiniBand and high-speed Ethernet are required to connect GPU clusters at scale. Nvidia's networking division, alongside Arista Networks and Juniper, are major beneficiaries.

Power and cooling account for 10–15 percent. A single modern GPU cluster consuming 100MW+ requires purpose-built power delivery and cooling infrastructure. Companies are signing direct power purchase agreements with utilities and nuclear operators — Microsoft signed a deal to restart Three Mile Island.

Custom silicon development represents 5–10 percent of spend. Google's TPUs, Amazon's Trainium2, Meta's MTIA, and Microsoft's MAIA reduce Nvidia dependence over time. These are multi-year bets costing $2–5 billion per generation but pay off at hyperscaler scale.

**Why the Race Keeps Escalating**

Three structural forces prevent any company from pulling back unilaterally. First, training cost scaling: each frontier model generation requires 10–100x more compute than the last. Skipping a training cycle means falling behind on capabilities now core to product differentiation.

Second, inference demand explosion: deployed AI products like Copilot, Gemini, and Meta AI generate billions of queries per day. Inference at this scale requires more compute than training. Under-provisioning means slower responses and higher per-query costs.

Third, the custom silicon race: every company paying Nvidia $30,000–$40,000 per H100 is motivated to develop proprietary chips. But building silicon takes 3–5 years and billions in R&D — so capex now buys strategic independence later.

**Nvidia's Position**

Nvidia's data center revenue grew from $15 billion in fiscal 2023 to $115 billion in fiscal 2025 — largely on the back of hyperscaler capex. The most recent print for fiscal Q2 2027 (the quarter ended July 26, 2026) shows data center revenue of $89.0 billion — up 117 percent year-over-year — with the ramp of Blackwell Ultra systems benefiting from the four major hyperscalers accounting for an estimated 40–50 percent of total GPU purchases.

But every dollar Microsoft spends on MAIA, Meta spends on MTIA, Google spends on TPUs, and Amazon spends on Trainium is a dollar that will eventually stop flowing to Nvidia. The custom silicon programs are 3–5 year bets, not current-quarter disruptions. For now, Nvidia captures the capex cycle. The question is whether its moat survives the transition to proprietary silicon at scale.

**Second-Order Effects**

When four companies simultaneously ramp capex by roughly 50–120 percent in a single year, ripple effects extend well beyond the AI industry. The US electric grid was not designed for 100MW+ data center clusters. Dominion Energy has 30GW of data center capacity in the queue in northern Virginia alone — more than the state's entire current generating capacity. Big tech is now signing deals with nuclear power providers to secure supply.

Data center REITs including Equinix and Digital Realty have seen demand outpace supply in every major market. Land adjacent to power substations in rural Virginia, Iowa, and Texas has appreciated 300–500 percent since 2023.

Hyperscaler data center builds are competing with semiconductor fab construction (TSMC Arizona, Intel Ohio) and clean energy projects for specialized construction labor, pushing wages and project timelines across multiple industrial sectors.