Sources: Amazon's AGI team is aiming to outperform Anthropic's latest Claude models by the middle of 2024 using the company's forthcoming LLM, codenamed Olympus
This week, Amazon completed the second phase of a deal announced last September, when it committed to investing up to $4 billion in OpenAI rival Anthropic.
Context & Ripple Effects
Amazon’s Olympus effort sits alongside, rather than in place of, its Anthropic partnership: the company had committed up to $4 billion for a minority stake in Anthropic in its initial Anthropic investment, and this report follows completion of the deal’s second phase.
That pairing matters because Amazon is simultaneously funding an external model supplier and trying to establish an internally developed model as a performance competitor. It makes model capability a strategic input Amazon does not intend to source from only one provider.
First-order effects
- Amazon’s AGI team is tasked with benchmarking the forthcoming Olympus LLM against Anthropic’s latest Claude models, creating an internal performance target despite Amazon’s financial relationship with Anthropic.
- Anthropic gains additional capital through the completed investment phase, while facing a better-funded prospective rival inside one of its major backers’ ecosystem.
Second-order effects
- Amazon can use a credible in-house alternative to strengthen its negotiating position with Anthropic over model access, deployment terms, and the role Claude plays in Amazon products.
- The partnership becomes more complex: teams building Amazon services may have to weigh a proprietary model against Claude rather than treating Anthropic as the default external frontier-model path.
Third-order effects
- If major cloud platforms continue to invest in model startups while developing competing models, the market is likely to organize around multi-model portfolios rather than exclusive supplier relationships.
- Competition will increasingly turn on the ability to pair model development with distribution, cloud infrastructure, and product integration—not only on standalone benchmark leadership.
The trend: This is one data point in AI industrialization, where cloud platforms hedge dependence on frontier-model partners by building proprietary models of their own.