Amazon senior VP Peter DeSantis concedes the company’s AI models have not reached the frontier of demanding workloads, but expresses confidence Amazon will be competitive among top AI providers within the next year.
- Amazon’s in-house AI models have not matched the capabilities of OpenAI’s GPT and Anthropic’s Claude.
- The company has invested $33 billion in Anthropic, gaining both equity and cloud revenue opportunities.
- Amazon plans to use custom Trainium chips and proprietary data from retail and logistics to accelerate AI development.
What happened
Peter DeSantis, Amazon’s senior vice president overseeing AI models, chips, and quantum computing, admitted that Amazon’s AI models have not been at the leading edge for the most demanding workloads. Despite launching its own model, Nova2, in December, Amazon acknowledges it has yet to match the performance of models like OpenAI’s GPT-5.5 and Anthropic’s Claude.
Alongside this candid admission, Amazon disclosed its significant $33 billion investment into Anthropic, a leading AI startup. This deal includes Anthropic’s access to Amazon’s custom Trainium chips and commits Anthropic to spend over $100 billion on AWS over the decade, creating dual revenue streams and strategic positioning for Amazon in the AI space.
Why it matters
Amazon’s announcement sheds light on the competitive dynamics in AI where even tech giants face challenges in leading frontier model development. By openly acknowledging its position, Amazon signals a realistic but ambitious commitment to closing the gap with OpenAI and Anthropic within a year.
The company’s approach combines massive financial investments in AI startups with leveraging its proprietary assets such as custom silicon technology and vast proprietary data from Amazon’s retail and logistics operations. This strategy could uniquely position Amazon to excel given its cloud infrastructure scale and specialized computing resources.
What to watch next
The market will closely observe Amazon’s ability to translate its investments and technology into competitive AI models able to meet the highest enterprise and research use cases. The upcoming release of Trainium3 chips, boasting four times the performance of current versions, will be a critical enabler for this goal.
Additionally, Amazon’s further developments around Project Prometheus and integrating its in-house AI capabilities with the Anthropic partnership will be key indicators of its long-term AI ambitions and whether it can close the innovation gap against OpenAI, Anthropic, and Google.