Amazon executives highlighted strong and growing customer demand for AI, noting the company is accelerating spending to keep pace with customer demand.
On its July 30 earnings call, Amazon said it is ramping up its capital expenditure for 2026 from $200 billion to $220 billion, with the majority directed toward AI-related investments. Even at that level, executives acknowledged capacity will remain constrained. “We will still not have enough capacity to meet all the demand we have in 2026,” said CEO Andrew Jassy, adding that the imbalance is likely to persist into 2027 and that “the demand we already have for 2028 is striking.”
At RBN, we’ve been tracking how hyperscalers are rapidly constructing more data centers and a key challenge is securing electricity to power these facilities. A central question for the tech and energy sectors is how that load will be served and the answer will likely involve a mix of natural gas, coal, renewables, and nuclear generation. See: Go Speed Race Go.
Amazon didn't disclose the size of its data center footprint. It reported a $600 million unrealized gain from changes in the fair value of long-term energy contracts used to secure electricity for existing and future data center operations. These contracts help lock in power for its data centers.
Executives also outlined the different investment timelines across AI infrastructure. Capital for data centers is typically deployed about two years before facilities become operational, and then they can generate revenue for 30-plus years without requiring another comparable upfront investment. In contrast, servers and networking equipment are purchased just months before use. The servers reach break-even in under three years and can last for up to six years.
