Amazon is trying to do two things at once in artificial intelligence: deepen its relationship with Anthropic and prove that its own AI team can build a stronger model in-house. That puts Olympus, Amazon's next flagship language model, at the center of a high-stakes test for the company.
The situation is unusual because Anthropic is both a major Amazon partner and a benchmark Amazon wants to beat. Amazon has invested up to $4 billion in the AI startup, but its internal AI division reportedly views Anthropic's Claude models as direct competition.
A partner that also raises the bar
Amazon completed the second phase of a deal announced in September, adding $2.75 billion to its investment in Anthropic. The total commitment now stands at up to $4 billion, described in the source as the largest amount Amazon has ever invested in another firm.
The Anthropic deal matters because Amazon is trying to avoid losing ground to Microsoft in the cloud race. Microsoft has been attracting many AI customers through its exclusive deal with OpenAI, and Amazon needs strong AI offerings tied to its own cloud business.
That makes Anthropic strategically valuable. Its models can help Amazon serve customers who want advanced generative AI capabilities. But the same relationship also creates pressure inside Amazon: if Claude is the strongest option available to teams and customers, Amazon's own model work risks looking secondary.
Olympus becomes the internal test
Amazon's in-house answer is Olympus, a large-scale language model that is currently in training. The model has been described as having "hundreds of billions of parameters," and the first reports about Olympus appeared in November 2023.
The internal target is ambitious. Amazon's AI division, led by SVP Rohit Prasad, is reportedly aiming to outperform Anthropic's latest Claude models by the middle of the year. That goal turns Claude into more than a partner technology; it becomes the measure against which Olympus will be judged.
Olympus is also expected to become a broad platform inside Amazon. When it launches later this year, it is expected to be integrated into almost all areas of Amazon and made available to other companies through AWS. The model is also said to be Amazon CEO Andy Jassy's pet project.
Why Claude is creating internal pressure
The challenge for Amazon is that its own language models have not yet produced the same level of confidence internally. According to the source, an Amazon insider told Alex Heath for The Verge that Anthropic's models are significantly better and more powerful than anything Amazon has built in-house so far.
That gap appears to be affecting internal teams. Amazon employees are reportedly frustrated by the long wait for Olympus and are considering using Anthropic's models instead. For teams that need to use large language models rather than train them, Claude may already be the practical choice.
"Claude competes with us and makes us nervous," says an Amazon insider. "For any team not training LLMs and only using them, Claude is a clear winner for now."
That quote captures the core tension. Amazon needs Anthropic because Claude strengthens its AI position, especially for cloud customers. At the same time, the stronger Claude looks, the more pressure Olympus faces before it has even launched.
Where Amazon already uses generative AI
Amazon is not waiting for Olympus to use generative AI across its business. The company is already applying the technology in several visible and internal-facing areas.
- Customer review summaries
- The business chatbot Q
- Tools that support retailers in creating product pages
These uses show why a flagship model matters to Amazon. A stronger in-house model could support many parts of the company and become a product for AWS customers. But until Olympus is ready, Amazon has to balance its own ambitions against the immediate usefulness of Anthropic's models.
The stakes for AWS
The broader business question is how Amazon positions itself in AI for cloud customers. AWS is central to the plan, because Olympus is expected to be offered to other companies through that platform. If Olympus performs well, Amazon can argue that it has both outside partnerships and its own competitive model.
If Claude remains clearly ahead, the picture becomes more complicated. Amazon would still benefit from its Anthropic investment, but its internal AI work would face sharper comparisons. The company would be backing a leading AI startup while trying to convince employees and customers that its own flagship model deserves attention.
For now, Olympus is the project that could change that balance. Amazon has a major partner in Anthropic, an internal model in training, and a clear target: beat Claude by the middle of the year.