Amazon Projects $230B AI Spending in 2027, Cuts Data Center Costs

On July 10, Amazon Web Services announced Loom, a new open-source platform designed to simplify how enterprises build and deploy AI agents while enforcing strong security controls and governance frameworks. Loom integrates with existing AWS services—Amazon Bedrock AgentCore and AWS Strands Agents—to offer lifecycle management at scale. The platform enforces required resource tagging, supports role-based access controls, and provides pre-validated blueprints for deployments. Agents can be launched using configuration-driven methods, with credentials securely managed through AWS Secrets Manager. The system supports both low-code Python agents and no-code deployments, implements OAuth2 identity propagation, and includes human-in-the-loop approvals for sensitive actions. Loom also connects to the AWS Agent Registry (currently in preview) to manage agent records and conduct governance reviews. The platform is available immediately via AWS Labs on GitHub and is aimed at platform engineering teams building applications with fully managed AWS services.

Three days earlier, on July 7, Bank of America Securities published updated AI spending estimates for Amazon that underline the scale of its infrastructure commitment. The brokerage projects Amazon will spend $159 billion（約25兆円） on AI infrastructure in 2026—unchanged from its prior forecast. For 2027, however, BofA raised its estimate to $230 billion（約37兆円）, up from a previous projection of $196 billion（約31兆円）. Across both years, Amazon is expected to invest $389 billion（約62兆円） in AI capacity expansion. The increase signals growing confidence that Amazon will maintain aggressive buildout momentum to secure compute capacity in a market where demand is rapidly outpacing supply.

BofA Securities also disclosed comparative cost data that reveals Amazon's structural advantage in the hyperscaler competition. The brokerage estimates it costs between $25 billion（約4兆円） and $45 billion（約7.2兆円） to build 1 GW of data center capacity, depending on the operator's efficiency and existing infrastructure. Amazon's unit cost is estimated at the low end—$25 billion（約4兆円） per GW—making it the lowest-cost builder among the major hyperscalers. By contrast, Google's estimated cost is $37 billion（約5.9兆円） per GW and Meta's is $45 billion（約7.2兆円） per GW. This cost advantage is significant: on the same dollar of capital, Amazon can deploy more capacity than competitors, potentially giving it a speed and efficiency edge as the AI infrastructure race intensifies. The lower unit costs likely reflect Amazon's long experience operating massive data centers for AWS, its established supply chain relationships, and optimized deployment processes. With analysts projecting more than 32% upside to Amazon stock at current valuations and 353 hedge funds holding positions in the company, the combination of scale and cost efficiency appears to be a key driver of institutional confidence in the stock's AI-driven growth potential.

Amazon is making a massive bet on AI infrastructure at a moment when data center capacity has become the central constraint in the hyperscaler race. Bank of America Securities' July 7 update reflects Amazon's aggressive expansion plans: the raised 2027 estimate to $230 billion（約37兆円） (from $196 billion（約31兆円）) signals BofA's confidence that Amazon will continue to invest heavily despite the already-enormous $159 billion（約25兆円） 2026 projection. Combined, the $389 billion（約62兆円） two-year commitment positions Amazon as one of the largest spenders in the buildout.

The real advantage lies not in the absolute scale of spending but in the unit economics. BofA's analysis shows Amazon can construct data center capacity at $25 billion（約4兆円） per GW—a meaningful gap versus Google's $37 billion（約5.9兆円） and Meta's $45 billion（約7.2兆円）. This cost structure is critical because data centers are long-lived capital assets; lower construction costs compound into superior returns over time. The platform's structural efficiency appears to stem from existing infrastructure, operational expertise, or sourcing advantages. In a market where the limiting factor is the speed and cost of capacity deployment, Amazon's edge may translate into faster scaling and better returns on invested capital—a point that likely underpins analyst projections of 32%+ upside.