Amazon is negotiating to move approximately $8 billion worth of high-end Nvidia processors off its balance sheet, according to the Financial Times. The deal would represent a significant shift in how the company manages its artificial intelligence infrastructure costs.
Amazon.com Inc. is in talks to transfer a substantial cache of top-tier Nvidia chips from its own books, marking a potential restructuring of the company's AI hardware strategy.
The $8 billion inventory of chips reflects Amazon's massive capital investment in AI infrastructure over the past year. As cloud providers race to build out generative AI capabilities, costs for specialized processors have skyrocketed, with companies competing aggressively for limited Nvidia GPU supply.
Moving the chips off its balance sheet could provide Amazon with accounting and financial flexibility. Such arrangements typically involve third-party financing or partnership structures that allow companies to access hardware without carrying the full asset burden.
Amazon Web Services has been expanding its AI offerings to compete with Microsoft and Google, both of which have made substantial investments in large language models and related infrastructure. AWS announced its own AI chips in recent years but continues to rely heavily on Nvidia's H100 and other advanced processors to meet customer demand.
The move comes as Amazon faces pressure to demonstrate profitability and efficient capital deployment. While the company has benefited from cloud computing demand, investors have scrutinized its spending on emerging technologies including AI.
Details of the potential deal structure remain unclear. Nvidia declined to comment. Amazon has not publicly responded to the Financial Times report.
The development underscores the immense financial commitment required to compete in the AI infrastructure space. Major cloud providers are spending tens of billions annually on data center buildouts and specialized chips, with costs expected to remain elevated as demand for AI computing power continues to grow.
This arrangement, if completed, would be one of several recent efforts by tech giants to optimize their hardware spending and balance sheet management amid the AI investment boom.
Most data centers are withholding information about their water and electricity consumption, according to a report. The lack of disclosure raises concerns about the environmental impact of the rapidly expanding data infrastructure industry.
Kevin O'Leary's ambitious Wonder Valley project aims to build a 40,000-acre AI data center in Utah requiring nine gigawatts of power—more than double the state's average energy consumption. The Verge's investigation reveals significant challenges to the "Stratos" proposal.
Private credit firms financing the AI infrastructure buildout face dangerous concentration risk similar to the SaaS sector collapse, according to Carlyle Group. The warning highlights how rapid capital flows into trending sectors can mask underlying vulnerabilities.
SoftBank Group's equity investors are prioritizing potential AI returns over concerns about the Japanese conglomerate's rising borrowing costs. The shift reflects growing confidence in the company's artificial intelligence strategy.