Amazon in talks to invest US$10bn in OpenAI

The high-stakes battle for dominance in artificial intelligence has entered a new and more expensive phase. Amazon, the e-commerce titan that also operates the world’s largest cloud computing business, is in preliminary discussions to invest at least US$10bn in OpenAI. This potential infusion of capital, which could value the creator of ChatGPT at more than $500bn, signals a shift in the strategic alliances that define the “Cloud Wars”. While OpenAI has long been synonymous with its primary backer, Microsoft, the emergence of Amazon as a major suitor suggests that the startup is diversifying its dependencies to fund its voracious appetite for computing power.

For Amazon, the deal is less about philanthropic support for research and more about a calculated push into the semiconductor market. Central to the negotiations is the proposed adoption of Amazon’s in-house “Trainium” chips by OpenAI. While Nvidia’s graphics processors remain the undisputed industry standard, tech giants are increasingly seeking cost-efficient alternatives to break that hegemony. If OpenAI, the industry’s bellwether, shifts a portion of its massive training workloads to Amazon’s proprietary hardware, it would provide a powerful validation of Amazon’s semiconductor unit and differentiate its cloud offering from rivals.

The urgency of these talks is underscored by the eye-watering costs of staying at the frontier of AI development. OpenAI’s chief executive, Sam Altman, has recently declared a “code red” to fend off competition from Google’s Gemini, yet the financial requirements to maintain this lead are staggering. The company is projected to burn through US$143bn between 2024 and 2029. With a projected spending commitment on “compute”—the chips and servers that power its models—reaching US$1.4tn over the next eight years, OpenAI’s reported US$13bn in annual revenue appears almost incidental. This fiscal reality has forced the company to restructure into a for-profit corporation and consider an initial public offering that could cost it US$1tn.

This bilateral maneuvering is taking place against a backdrop of a “global construction frenzy” in digital infrastructure. In 2025, investment in data centers worldwide hit a record US$61bn, a surge driven by the insatiable demand for the real estate and energy required by generative AI. The scale of this build-out has birthed unusual financial arrangements; hyperscalers like Meta and Alphabet are tapping debt markets at historic levels, while private equity firms are increasingly viewing data centers as a stable, long-term asset class. In one landmark transaction this year, a consortium including Nvidia and Microsoft participated in a US$40bn acquisition of Aligned Data Centers, the largest deal of its kind on record.

However, the speed and scale of this capital deployment have sparked concerns among more cautious observers that investment levels may be running ahead of sustainable demand. Critics point to “circular” deals—where AI companies and chipmakers invest in one another to create the appearance of growth—as a potential sign of market froth. Furthermore, the rise of “neoclouds” like CoreWeave, which rely on narrow customer bases and massive debt to lease GPUs, adds a layer of volatility to the ecosystem. As electricity demand for these facilities is projected to double by 2030, reaching levels comparable to the total consumption of Japan, the industry is entering uncharted territory. Whether these billions in investment will yield a commensurate revolution in productivity remains the US$1.4 trillion question.