Amazon's AI Chief Says Company's Models Lag Behind OpenAI, Anthropic

Amazon’s AI Chief Says Company’s Models Lag Behind OpenAI, Anthropic Amazon is openly conceding that it trails rivals in cutting‑edge AI models even as it prepares an aggressive push into the AI hardware market, setting up a year of catch‑up on two critical fronts.

In mid-June, Amazon AI chief Peter DeSantis acknowledged that the company’s models “haven’t been at the very frontier,” lagging systems from OpenAI and Anthropic, while expressing hope Amazon can close the gap within a year. The comments come as the company continues to build its own frontier models and simultaneously pours tens of billions of dollars into Anthropic, whose Claude models compete directly in the generative AI space.

A day later, another piece of the strategy became clearer: Amazon Web Services (AWS) is moving to commercialize its in‑house AI chips more broadly. DeSantis told Bloomberg that AWS is in talks to sell its Trainium accelerators to other data centers, a shift later detailed in reporting on Amazon’s evolving chip business. The company is positioning this as one of the most serious challenges yet to Nvidia’s dominance in AI compute.

Internally, top leadership frames the chip push as a massive opportunity. CEO Andy Jassy has argued that if Amazon’s chips business were standalone and sold to AWS and external customers “as other leading chips companies do,” it could already support an annual run rate of around $50 billion — and that “it’s quite possible we’ll sell racks of them to third parties in the future.”

Seen together, the strategy reflects two parallel bets. On the software side, Amazon aims to rapidly improve its own models while leveraging its investment in Anthropic. On the infrastructure side, it hopes that selling AI chips — even to potential model competitors — will entrench AWS as a foundational supplier in the AI economy, narrowing the gap in capabilities while expanding its role in the underlying stack.

Continue reading https://foxvector.com/stories/019ef2f1-e27c-1f38-7396-005478d1435e