Amazon closed a $25 billion corporate bond sale on July 7, one of the largest debt offerings in the company’s history, with proceeds earmarked for AI data centres, chips, and the supporting infrastructure that underpins both AWS and the company’s growing suite of AI-powered services. The company has simultaneously confirmed it will not issue additional debt for the rest of 2026.
What Happened
Amazon priced an eight-tranche bond offering on July 7 with maturities ranging from three to forty years. The longest tranche, due in 2066, was priced approximately 1.45 percentage points above equivalent US Treasury yields. The deal was managed by Barclays, Goldman Sachs, JPMorgan, and Morgan Stanley.
Investor demand for the offering was exceptionally strong. Orders peaked at roughly $62 billion before the banks tightened spreads, which pared the final order book to approximately $41 billion — still 1.6 times the size of the deal itself. Amazon said the proceeds will go toward general corporate purposes, with a clear emphasis on capital expenditures related to AI infrastructure. The company has told its underwriters it does not plan to return to the debt markets in 2026.
Why It Matters
Amazon has projected its total capital expenditure will reach $200 billion in 2026, up sharply from $131 billion in 2025. The overwhelming majority of that spending is concentrated in data centres, custom AI chips, and the networking infrastructure required to serve AI workloads at hyperscale. The $25 billion bond sale is a signal that Amazon is willing to take on debt to close any timing gap between cashflow generation and its investment programme.
The scale of the deal underscores how seriously Big Tech is treating AI infrastructure as a generational bet. Microsoft has committed $2.5 billion to its Frontier Company to embed AI engineers inside large enterprises, but Amazon’s bond positions it as one of the largest single infrastructure investors in the world — not just in AI, but across any industry.
Critics have raised questions about the sustainability of AI infrastructure spending at this scale. Google’s AI buildout drove a 37% surge in electricity consumption in 2025, and Amazon’s planned data centre expansion will similarly put pressure on power grids and energy markets. Whether the revenues from AI services can eventually justify the capital deployed remains the defining question for all three hyperscalers.
Background and Context
AWS is Amazon’s most profitable business segment, consistently generating the operating income that funds investment across the rest of the company. The cloud unit has been in an AI arms race with Microsoft Azure and Google Cloud since the generative AI boom began in 2023, and all three hyperscalers have made multi-hundred-billion-dollar capital commitments to remain competitive.
Amazon has also been building out its own AI chip portfolio. Trainium and Inferentia, its custom silicon for training and inference respectively, are central to the strategy of offering competitive AI compute at lower cost than Nvidia GPU instances. Even AI startups like DeepSeek are now designing custom inference chips to reduce dependence on commercial silicon — a dynamic that suggests the entire industry is moving toward proprietary hardware wherever scale allows.
What Comes Next
Amazon is expected to announce several new AWS AI services at its re:Invent conference later this year, with the infrastructure built or expanded through this bond’s proceeds forming the backbone of those offerings. The company is also expanding its Amazon Bedrock platform, which gives enterprise customers access to foundation models from multiple providers including Anthropic, and has been investing in physical data centre locations across Europe, the Middle East, and Southeast Asia.
For investors, the bond’s 40-year tranche is the most striking signal of all: Amazon is effectively telling markets that it expects AI-driven cloud computing to remain a dominant force in its business for the next four decades. How quickly the revenues catch up to that ambition will be closely watched in every quarterly earnings cycle between now and then.
