Amazon is launching the largest capital expenditure program in its history, aiming to reclaim momentum in artificial intelligence and defend its leadership in cloud computing through massive investments. Facing intense competition from Microsoft and Google, the tech giant is expanding data centers, developing chips, and building AI models to regain ground.
The company's CEO announced last week that capital expenditures will climb to $200 billion this year, exceeding the spending levels of Google and Microsoft. The vast majority of this investment is allocated to computing infrastructure, with roughly three-quarters of the budget directed to its cloud division, AWS. In a strategic consolidation, the CEO has merged chip, AI model, and advanced research teams under a unified leadership structure to streamline its AI strategy. This move accompanies cost-cutting measures, including the elimination of approximately 30,000 corporate roles.
Investors have expressed unease over the scale of this gamble. Concerns that the massive outlay will be slow to generate returns have contributed to a more than 20% decline in Amazon's stock price from its peak last November. Despite this market punishment, the CEO remains committed to the strategy, citing the company's extensive experience in interpreting AWS demand signals and converting capacity into capital returns, and expressing strong confidence in the investment.
Employees within AWS indicate that the aggressive moves reflect internal anxiety that Amazon failed to fully capitalize on its first-mover advantage in cloud computing. Specifically, after OpenAI launched ChatGPT in 2022, Amazon was slower than its rivals to secure major contracts with leading AI providers.
**Largest-Ever Capital Expenditure Plan** According to public filings, the planned $200 billion capital expenditure will be predominantly used for AWS expansion. In comparison, the combined expenditures of Microsoft, Google, and Oracle for this year are projected to be near $4 trillion. The CEO stated that Amazon plans to significantly increase data center capacity this year. The company added nearly 4 gigawatts of capacity in 2025, an amount equivalent to the annual energy consumption of over 3.2 million US households, with plans to double capacity by 2027.
To support this expansion, the CEO reorganized the corporate structure in December, consolidating chip, model, and advanced research teams under a single leadership framework. The CEO emphasized that the company will continue to operate like the "world's largest startup," maintaining a customer-centric, lean, and agile approach.
**Competitive Pressures on AWS** Although AWS generated nearly $130 billion in sales last year, contributing over 60% of Amazon's total profit, and remains the world's largest cloud service provider, analysts predict that Microsoft's cloud business could surpass AWS within the next three years as demand for AI-driven cloud services surges.
According to more than a dozen current and former senior employees, AWS is losing ground in the competition for enterprise AI contracts. A former senior AWS employee admitted, "We were completely unprepared for how quickly things would evolve."
This lag is evident in key deals. Microsoft, an early investor in OpenAI, secured an exclusive cloud computing contract with the ChatGPT maker. Amazon only signed a $38 billion cloud agreement with the startup last year, after OpenAI permitted corporate restructuring. However, this deal pales in comparison to OpenAI's $250 billion contract with Microsoft and a $300 billion agreement with Oracle. Furthermore, Amazon's $8 billion investment in Anthropic and construction of data centers for it occurred after Google had already backed the startup.
**Chip Strategy and Computing Power Push** To reduce reliance on Nvidia's products and improve profit margins, Amazon is aggressively promoting its custom-designed chips. The company claims that its Graviton chips for general cloud computing and Trainium chips for AI training have the potential to generate over $10 billion in annual revenue. In December, Amazon launched the latest generation of its Trainium chip, promising significant performance improvements.
Sources indicate that Amazon is negotiating to participate in OpenAI's latest multi-billion dollar funding round, partly to ensure the ChatGPT maker adopts its semiconductor products. Competition remains fierce, however. Google successfully secured a deal worth tens of billions of dollars for Anthropic to purchase one million of its Tensor Processing Units (TPUs).
An analyst from a technology consulting firm expressed skepticism about whether leading AI startups would adopt Amazon's chips. He noted that while Amazon emphasizes cost-effectiveness, some users prioritize peak performance, even if Amazon's solutions are cheaper to run than Nvidia's.
**In-House Model Dubbed "Amazon Basics"** In model development, Amazon is also investing in advancing its AI model, codenamed "Nova," positioning it as a lower-cost alternative to competitor models. However, independent benchmark tests show that Nova's performance lags behind state-of-the-art models from OpenAI, Google, Meta, and Anthropic.
According to sources, some AWS employees have privately nicknamed Nova "Amazon Basics," a term typically used for the company's line of generic, low-cost household products, which has irritated executives. Although the company encourages employees to use its own AI tools and has set a target for 80% of developers to use AI for coding at least once a week, several company engineers stated they prefer using Anthropic's Claude model over Nova for coding tasks. One AWS engineer remarked bluntly, "I didn't even know we had a model."
The pressure to regain footing in the AI race weighs heavily on employees. Some worry that Amazon could slide into what its founder once described as "Day 2" – a state of business stagnation followed by a "painful decline." A senior AWS engineer commented, "The culture has changed, and the world around us has changed. We have to prove our value."