Amazon Boosts AI Infrastructure Spending to $220B

Amazon is dramatically increasing its projected cash capital expenditures for AI infrastructure, now targeting approximately $220 billion by 2026. This represents a significant jump from its earlier estimate of $200 billion, signaling an intensified focus on scaling its AI capabilities.

This heightened investment stems directly from the escalating costs of crucial memory chips and persistent, robust customer demand for AI compute capacity. Industry expectations suggest these capacity constraints will continue through 2027 and extend well into 2028.

Key Investment Figures

• Projected 2026 AI infrastructure cash capital expenditures: $220 billion (up from $200 billion).
• Amazon Web Services (AWS) reported 36.7% year-over-year revenue growth, its fastest in 18 quarters.
• AWS revenue reached $42.2 billion in the latest reporting period.

CEO Andy Jassy affirmed that the primary challenge for data center operators is not attracting customers, but rather the critical task of providing sufficient infrastructure capacity. This underscores a pivotal phase in AI infrastructure deployment where supply must catch up with overwhelming demand.

Industry expert Sid Nag of Tekonyx highlighted that the current focus is on deploying physical infrastructure, identifying compute, power, networking, and data center construction as significant bottlenecks. Nag further explained that the cost increase is less about building new facilities and more about equipping existing ones with high-bandwidth memory, making silicon and memory economics central to future AI investments.

Strategic Long-Term Vision

• AWS backlog has reached $496 billion, indicating robust future demand.
• Amazon plans to double its power capacity by the end of 2027 to support growth.
• Data centers are projected to generate revenue for over 30 years.
• Servers and networking equipment achieve break-even in under three years, contributing substantial free cash flow.

Jassy elaborated on the long-term rationale, noting the enduring revenue generation of data centers and the rapid return on investment for equipment. He anticipates that AWS could eventually evolve into a trillion-dollar annual revenue business, driven by this strategic investment.

Beyond specialized AI, these workloads also significantly boost demand for conventional cloud infrastructure, with services like reinforcement learning and vector databases relying on traditional compute and storage. AWS’s custom AI silicon, Trainium, is seeing increased adoption from major players like Anthropic and OpenAI, with Amazon even considering external sales due to growing customer interest.

Amazon’s aggressive capital allocation reflects a strategic imperative to dominate the foundational AI compute layer, securing its competitive edge in a market where capacity, not demand, dictates growth. The company is positioning itself for sustained, long-term revenue generation from these foundational investments.