Amazon is redesigning part of a massive AI data center campus in rural Indiana into a sprawling cluster of powerful computers to build its next frontier AI models. Internal planning documents describe an effort to consolidate multiple data centers and deploy thousands of Trainium-powered AI servers.
According to people familiar with the matter, the effort is part of a broader initiative called “AGI Pivot” supporting the company’s AGI organization and its future in-house AI models.
The documents suggest Amazon’s frontier-model ambitions remain intact despite recent job cuts in its AGI organization and the wind-down of the previous Nova model series.
They are making better use of infrastructure they have already paid for and operate by linking existing data centers into a larger, more efficient AI cluster.
CEO Andy Jassy commented that AI infrastructure should generate attractive long-term returns because data centers remain productive for decades while servers and networking equipment can be refreshed over time.
The planning documents suggest Amazon is racing to expand computing capacity for its next frontier AI model push. One update describes an “emergent request” to deploy more than 6,000 Trainium-powered AI servers, speeding up launch schedules by several weeks.
The accelerated timeline was intended to prepare Amazon’s next AI model for this year’s re:Invent conference, typically held in early December, according to this document.
Amazon is creating a “larger, more efficient” AI system to maximize performance by consolidating computing across the Indiana campus into what it calls an AGI SuperCluster.
Related: Meta’s AI agents turned unruly in tests
To achieve this, Amazon is redesigning networking, storage, and fiber-optic infrastructure so multiple data centers function as one.
Some facilities are planning to replace older Trainium 2 systems with newer Trainium 3 servers.
Amazon is significantly expanding its AI investment, having recently raised its projected 2026 capital expenditures to $220 billion from $200 billion as demand for AI computing continues to outstrip available capacity and prices for some components rise.
As the company pushes forward with its AGI Pivot, it’s likely that Amazon’s custom chip business, including Trainium AI chips and Graviton processors, will play a key role in supporting the development of its next frontier AI models.
The planning documents offer a glimpse of what Amazon’s next frontier-model effort will require, including expanding data storage to support multimodal AI training, which requires processing large numbers of high-resolution images and repeatedly saving the model’s progress.
Despite last month’s layoffs in the AGI organization, one person familiar with the effort said the push has “not slowed down”.
Amazon’s efforts to improve its AI infrastructure are part of a larger trend in the industry, as companies like Amazon continue to invest heavily in AI research and development, driving demand for more powerful and efficient computing systems.
They are driving demand for more powerful and efficient computing systems.
Leave a Reply