Amazon Web Services (AWS) hosted its re:Invent event in Las Vegas, and we saw some exciting announcements, so lets break them all down.
First, the company unveiled its next generation of AI-Chips that they claim, are 4-times faster and more capable for AI training. Christened ‘Trainium2’, the next generation chips are clearly meant for training purposes has been designed to be 2x more power efficient in addition to 4x better performance. Amazon believes that these chips will allow users to train their AI models faster and cheaper as the power consumption is exponentially reduced, leading to a smaller bill. Anthropic, a competitor to OpenAI, who has been backed by Amazon, has already announced its intention to start training its AI models using Trainium2.
Next up, we saw the Graviton4, a chip that Amazon says, is for general use. These next-gen processors are based on the ARM architecture, but consume less energy compared to both Intel and AMD chips, which in itself, is an impressive feat. Now, if you were to use a pre-loaded AI Model on the Graviton4, you can expect to see a 30% increase in general performance. According to Amazon, the Graviton4 will allow AWS customers to “process larger amounts of data, scale their workloads, improve time-to-results and lower their total cost of ownership.”