Amazon delivered fourth-quarter 2025 financial results on February 7, 2026, reporting net sales up 14% year-over-year to $213.4 billion and Amazon Web Services revenue growth of 24%, yet shares plunged 5.6% to $210.32 as investors reacted to the company’s guidance for capital expenditures projected to exceed $200 billion in 2026 — a figure that analysts described as representing over 50% growth from already elevated 2025 spending levels and the highest single-year capex commitment in Amazon’s history.
AWS revenue reached $32.1 billion for the quarter, driven by surging enterprise adoption of AI services, growth in its Bedrock AI development platform, and expanding infrastructure contracts with AI-native companies seeking cloud scale. The company disclosed that AI workloads are growing faster than any other category on AWS, with enterprise customers increasing AI-related cloud spending by 80% year-over-year in Q4.
The $200 billion capital expenditure projection for 2026 encompasses data center construction across North America, Europe, and Asia-Pacific; procurement of tens of thousands of Nvidia and custom Amazon Trainium and Inferentia AI chips; networking infrastructure upgrades supporting AI model training and inference at scale; and power infrastructure investments including renewable energy contracts required to meet the electricity demands of AI-optimised compute clusters.
CEO Andy Jassy framed the investment as essential for maintaining AWS’s competitive position against Microsoft Azure, Google Cloud, and emerging AI-native infrastructure providers. He emphasised that historical patterns demonstrate cloud infrastructure investment reliably generates long-term returns as enterprise adoption expands — drawing a parallel to AWS’s data center buildout in the mid-2010s that initially alarmed investors but ultimately created one of the most profitable businesses in technology history.
The market’s sceptical reaction reflects genuine uncertainty about whether AI infrastructure spending at this scale generates proportional revenue growth. Unlike the 2010s cloud buildout where demand for on-demand computing was clear and expanding linearly, AI infrastructure investment is partly speculative — bet on capabilities and use cases whose commercial value remains unproven at enterprise scale.
Beyond infrastructure, Amazon announced expansion of its Amazon Q enterprise AI assistant, which provides AI capabilities embedded directly into business workflows across coding (Amazon Q Developer), business intelligence (Amazon Q Business), and contact centre automation. Amazon Q Developer has gained traction among software development teams, with the company claiming significant reductions in code review time and bug detection rates for enterprise customers.
The investment in custom silicon continues accelerating. Amazon’s Trainium 2 chips, purpose-designed for AI model training, are now deployed in AWS regions globally. Inferentia chips optimised for cost-efficient inference serve customers deploying AI models at scale. The custom silicon strategy aims to reduce dependency on Nvidia GPUs while delivering better price-performance for specific AI workloads — a competitive differentiation unavailable to cloud rivals without comparable semiconductor programmes.
