Amazon is accelerating investments in AI as demand surges, with its cloud unit reporting rapid growth and committing significant capital to expand capacity, CEO Andy Jassy said in his 2025 letter to shareholders.
It plans to spend approximately $200 billion in capital expenditure in 2026, much of it tied to existing customer commitments, including a reported deal with OpenAI. Jassy said these investments will impact short-term free cash flow but generate returns over time.
Amazon Web Services (AWS) is seeing strong traction in AI, with its AI-related revenue run rate exceeding $15 billion in the first quarter of 2026, compared to $58 million three years after AWS’s initial launch. Customers are choosing AWS for its range of AI tools, including model development, inference, and agent-building capabilities, as well as proximity to data and existing applications, the CEO said.
However, Jassy also noted that AWS’s growth is constrained by capacity despite expanding infrastructure. The company added 3.9 gigawatts of power capacity in 2025 and plans to double capacity by 2027.
AWS reported a $142 billion revenue run rate in Q4 2025, with 24% year-on-year growth, but “still have capacity constraints that yield unserved demand,” he said.
Amazon is also expanding its custom chip business, including Graviton CPUs and Trainium AI chips, to reduce costs and improve performance.
The chips division has reached an annual revenue run rate of over $20 billion, growing at triple-digit rates, and could reach $50 billion if operated as a standalone business, Jassy said.
“Our second version of our custom AI silicon (Trainium2) … has largely sold out,” he said, adding that newer versions are already heavily reserved ahead of availability. He said custom chips could save tens of billions of capex dollars per year and improve operating margins.
AI adoption is outpacing previous technologies, Jassy said, citing the rapid rise of OpenAI’s ChatGPT, which reached 100 million users within two months of its November 2022 launch. He added that both OpenAI and Anthropic are approaching annual revenue run rates of $30 billion.
Jassy also highlighted internal changes driven by AI, including a redesign of Amazon Bedrock’s inference engine and the overhaul of Alexa into “Alexa+.”
He said the Bedrock team, managing the cloud computing service, rebuilt a core system in 76 days using AI-assisted coding tools, while Alexa’s upgrade has led to increased engagement, including higher usage, purchases, and smart home interactions.
ALSO READ: SiMa.ai to Expand Physical AI Systems with Micron Investment
