Circling the wagons: Amazon seeks to invest $10bn in OpenAI

Joins Oracle, Nvidia, AMD and Broadcom in infrastructure deals with OpenAI

Amazon is engaged in talks to invest more than $10bn in OpenAI, becoming the latest infrastructure provider to seek tie-ins with the world’s most visible GenAI company.

The talks were first reported by The Information and have since been independently confirmed by other media outlets, although the parties concerned are yet to comment in public.

A positive outcome would see Amazon join Oracle, Nvidia, AMD and Broadcom as major investors in OpenAI, the quid quo pro likely being that the ChatGPT maker uses Amazon’s cloud services and/or AI chips.

Microsoft, an early investor in OpenAI with an estimated $13 billion stake, is reportedly loosening its ties as it focuses on its in-house AI developments.

Amazon has also invested $8 billion in OpenAI rival Anthropic, creator of the Claude LLM series.

It’s not one way traffic. Last month, OpenAI agreed a long-term deal to buy $38 billion worth of datacentre capacity from AWS.

In total, OpenAI has committed to around $1.5 trillion in infrastructure deals with big US providers, despite recording annual losses of several billion dollars, with profitability not expected until 2029 at the earliest. The Wild West of datacentre-related activity is the only growth area in an otherwise moribund US economy.

The circularity of these investments, the lack of ready data centre capacity and unforeseeable developments in China - where tech startups like DeepSeek are releasing competitive open models for free – are worrying investors and intensifying fears of an unsustainable bubble.

And it’s not just investors. IBM CEO Arvind Krishna has warned that there’s “no way” that the current investment frenzy in AI infrastructure in pursuit of artificial general intelligence (AGI) will pay off.

Last week Oracle’s shares took a hit, blamed on its AI-related expenses and weaker-than-anticipated economic outlook, and earlier this month OpenAI announced “code red” due to Google’s Gemini 3 model beating the latest GPT models on some metrics.

Google, meanwhile, has said it must double its AI serving capacity every six months to keep up with soaring demand for AI services. Although its strategy seems to be more focussed on deploying its own hardware to mitigate supply chain bottlenecks, the question remains as to where the capacity – including energy, water and skills – is going to come from to support that kind of exponential growth.