Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Amazon has announced a management shake-up, including the departure of its artificial intelligence chief, as the company prioritises catching up to rivals on their advanced chips and large language models.
Chief executive Andy Jassy told employees the company had reached an “inflection point” with several technologies that required a renewed focus to “maximise potential for customers and Amazon over the long term”.
“I’m excited about what this team will build and how these foundational technologies will help shape Amazon’s future,” Jassy said.
As part of the changes, Peter DeSantis, who previously headed up the company’s data centre engineering teams, will oversee a new group leading Amazon’s AI model development, chipmaking unit and quantum computing research, according to an employee memo published on Wednesday.
The move will be followed by several other management changes, including the departure of Rohit Prasad, head of Amazon’s large language model effort, at the end of the year.
Prasad will be replaced by Pieter Abbeel, a former OpenAI researcher and founder of robotics start-up Covariant.
Although the $2.4tn group’s web services division is the leading provider of rented data centre processing and data storage, it lags rivals such as Google and Microsoft on the adoption of its own chips and models.
The company’s shares are up nearly 1 per cent in the year to date, compared with Google parent company Alphabet’s 56 per cent leap and Microsoft’s 14 per cent increase.
Amazon is in talks to invest more than $10bn in OpenAI as part of a deal that is expected to increase the adoption of its Trainium chips. The company released its most recent line of chips during a summit held in Las Vegas earlier this month, heralding its data-processing capabilities relative to leading chipmaker Nvidia’s Blackwell series of high-performance chips.
Yet the customer base for Trainium is narrow, with the chips used mostly for the cloud giant’s own work and the development of the Amazon-backed Anthropic’s AI models.
The Seattle-based group has positioned its Nova AI models to be more competitive on cost than rivals. However, Google and Microsoft have also invested heavily in their in-house model development, reflecting a desire to compete more aggressively on performance with start-ups such as Anthropic and OpenAI.
Abbeel, who joined the company last year alongside Covariant’s co-founder Peter Chen, would continue to work on robotics, the company said. David Luan, head of the company’s AGI lab in San Francisco, will also remain in his post.
Read the full article here