Amazon now expects to spend about $220 billion on capital projects in 2026, up from roughly $200 billion, and CEO Andy Jassy said most of the increase comes from higher memory costs rather than more construction. Even at that level, he said on the company’s July 30 second-quarter call, AWS won’t have enough capacity to meet demand.
“We will still not have enough capacity to meet all the demand we have in 2026,” Jassy said, according to Data Center Knowledge. “I believe this dynamic will also be true in 2027, too. In fact, the demand we already have for 2028 is striking.”
AWS growth and backlog
AWS revenue rose 36.7% year over year to $42.2 billion, its fastest growth in 18 quarters, and operating income reached $16.6 billion. Amazon said AWS is running at a $169 billion annualized revenue rate, its backlog has reached $496 billion, and it remains on pace to double its power capacity by the end of 2027 compared with 2025. The company also listed “resource and supply volatility, including for memory chips” among its business risks.
Jassy said Anthropic and OpenAI have made multi-year, multi-gigawatt commitments to Amazon’s Trainium chips, and that Amazon is exploring selling Trainium outside AWS. “I expect there’s a real chance we’ll do that in the future,” he said.
Sid Nag, CEO of Tekonyx, said the increase “wasn’t driven by building more data centers, but by the rising cost of equipping them, particularly high-bandwidth memory.”
The depreciation question
Jassy split the investment into two parts. Data centers carry “30 plus year useful lives,” while servers “currently have a useful life of at least five to six years,” he said, adding that AWS has a track record of “finding ways to extend the useful life of this equipment,” as Yahoo Finance reported. Data centers need capital about two years before they bring in revenue, and servers typically break even in under three years, he said.
Amazon has moved in the other direction before. Effective January 2025, it cut the useful life of a subset of servers and networking gear from six years to five and retired some equipment early, citing “an increased pace of technology development, particularly in the area of artificial intelligence and machine learning.” The change took roughly $700 million off 2025 operating income, according to Yahoo Finance.
Yahoo Finance also cited internal planning documents, reported by Business Insider, showing some facilities on Amazon’s Indiana campus planning to put Trainium 3 servers in place of Trainium 2 systems. Trainium 2 reached general availability in December 2024, so the oldest units are about 20 months old.
Separately, Amazon has committed up to $50 billion to build AI and supercomputing capacity for U.S. government customers. The project, set to break ground in 2026, will add nearly 1.3 gigawatts across AWS Top Secret, AWS Secret and AWS GovCloud (US) Regions, Amazon said.
WebProNews is an iEntry Publication