This week, Amazon completed the second phase of the deal announced in September last year, committing to invest up to $4 billion in Anthropic, a competitor to OpenAI. The additional $2.75 billion investment is Amazon's largest single investment in another company, once again highlighting the importance of large language models for big tech companies.

This is understandable as Amazon needs to provide a model through AWS to compete with its cloud computing rival Microsoft (using OpenAI technology), and Anthropic is the best existing alternative. If we go back to the era when large tech companies could make large acquisitions without being hindered by regulatory agencies, I believe Amazon would definitely try to acquire Anthropic directly. However, Amazon is currently only passively investing billions of dollars and explicitly stating that it only holds a minority stake and does not have a seat on the board. For Amazon, it is clearly advantageous that Anthropic has also agreed to invest $4 billion in AWS in the coming years.

This situation is obviously similar to Microsoft's sponsorship of OpenAI's growing computational needs. However, Amazon's relationship with Anthropic is far from harmonious on the surface. In fact, another part of Amazon is trying to directly compete with Anthropic's model. The Amazon AGI team, led by Senior Vice President Rohit Prasad, has the ambitious goal of surpassing Anthropic's latest Claude model by mid-year. Their upcoming flagship model, internally codenamed Olympus, is currently in training and is of considerable scale with billions of parameters.