Top cloud providers to outspend Ireland’s GDP on AI in 2026

Steven

Top cloud providers to outspend Ireland’s GDP on AI in 2026

Summary

Market researcher TrendForce estimates the world’s eight largest cloud providers (Google, Amazon, Meta, Microsoft, Oracle, Tencent, Alibaba and Baidu) will invest around $710 billion in 2026 on datacentres, servers and AI infrastructure — a 61% increase year-on-year and more than Ireland’s entire GDP last year.

The spending targets high-performance servers packed with GPUs from Nvidia and AMD, while many providers are also deploying custom ASICs for efficiency and performance. Google is notable for favouring its TPUs (expected to be in about 78% of AI servers it ships), while Amazon, Meta, Microsoft and others remain GPU-heavy. This surge is intensifying shortages and price rises in memory markets, prompting vendors like SK hynix and Sandisk to push a new memory tier called high-bandwidth flash (HBF) to sit between HBM and SSDs and ease scaling and costs for large AI workloads.

Key Points

  • Eight hyperscalers are forecast to spend approximately $710bn on servers and infrastructure in 2026 — roughly 61% more than 2025.
  • The top four cloud providers (Google, Amazon, Meta, Microsoft) account for the vast majority (~$635bn) of that capex.
  • Spending is focused on datacentres and high-performance servers largely using GPUs, but with growing deployment of ASICs (e.g. Google’s TPUs).
  • Google expects TPUs to make up about 78% of AI servers it receives in 2026; Amazon is expected to use ~60% GPU servers with Trainium3 ramping later in the year.
  • Soaring demand for AI kit is exacerbating a memory shortage and upward pressure on prices for HBM and server-grade DRAM.
  • SK hynix and Sandisk are proposing high-bandwidth flash (HBF) as a mid-tier memory layer to offer much higher capacity than HBM at lower cost and reduce TCO for inference workloads.
  • HBF aims to bridge the gap between ultra-fast HBM and high-capacity SSDs, potentially improving inference scale without frequent SSD access.

Context and relevance

This story matters because hyperscaler capex drives component supply chains, pricing and technology direction across the entire cloud and server market. Rapid, concentrated investment by a few giant providers amplifies shortages (memory, HDDs, GPUs), alters vendor roadmaps towards ASICs and bespoke designs, and influences where and how enterprise workloads run. For anyone buying cloud capacity, renting GPUs, procuring servers, or tracking AI infrastructure costs and sustainability, these shifts will affect availability, contract pricing and long-term architecture choices.

Why should I read this?

Short version: the biggest cloud players are about to spend an eye-watering amount on AI kit, and that will ripple into prices, supply and the tech everyone else depends on. If you’re involved in cloud strategy, purchasing, capacity planning or AI deployment — this directly affects your budgets and timelines. We’ve done the heavy lifting so you can see the knock-on effects fast.

Author tone

Punchy — the story highlights the scale and consequences of hyperscaler spending and why it’s shaping the AI infrastructure market right now.

Source

Source: https://go.theregister.com/feed/www.theregister.com/2026/02/26/trendforce_cloud_ai_spend/