Amazon’s AI Gambling: Can its new Nova models compete with OpenAI, Google and Anthropic?

However, the question on many Amazon watchers’ minds was: Why? After all, Amazon has invested $8 billion in Anthropic, including another $4 billion announced just last week. As part of the agreement, Anthropic has also committed to using Amazon’s AWS cloud and Amazon’s custom AI computer chip called Trainium to train and run its models.

Amazon wants to be in control of its own AI destiny

But Amazon clearly has no intention of relying solely on external partners for its AI strategy. It does not want to be beholden to Anthropic or any other third party. The company wants to significantly reduce the cost of its AI offerings for cloud customers – a cost-effective strategy that Amazon’s cloud computing service AWS has long pursued. This would be more difficult if only Anthropic’s models were used. Finally, the company says its customers wanted features like video that Anthropic doesn’t currently offer.

The Nova releases are part of a larger, perhaps master plan to pave its own path to AI dominance – which includes building what is said to be the world’s largest AI supercomputer called Project Rainier, which will include hundreds of thousands of Amazon’s Trainium chips in unison work.

In a keynote speech yesterday at re:Invent, Amazon CEO Andy Jassy’s cited three lessons on how Amazon develops its AI strategy. One, he said, is that as companies scale generative AI applications, “computing costs really matter” — so a cheaper, standardized model is becoming increasingly important.

Next, Jassy said that AI success doesn’t simply come from a powerful model, but from a model that addresses issues like latency – the time it takes to get the information or results you need from a model – as well as user experience and infrastructure problems. He also emphasized that Amazon wanted to offer both internal and external customers a variety of models to choose from.

“We have a lot of our internal developers using Claude from (Anthropic),” he said. “But they also use llama models. They also use Mistral models, and they also use some of our own models, and they also use home-developed models.”

This was surprising at first, he said, but added that “we learn the same lesson over and over again, which is that there will never be a single tool to rule the world.”

This appears to be the core of Amazon’s AI master plan: Ben Thompson, a business and technology analyst and author of Stratechery, wrote yesterday that Jassy’s comments on AWS’ AI strategy “look very similar to AWS strategy in general.” Just as AWS offers a wide choice of processing or databases, AWS will offer a choice of AI models on its Bedrock service. And that includes Amazon’s own Nova models, which just so happen to be probably the cheapest option for third-party developers. Amazon is betting that AI will become a commodity, he said: “AWS is betting that AI will be so important that it will end up being nothing special at all, which is exactly Amazon’s sweet spot.”

Amazon needs to find a balance between cost and performance

However, some argue that lower-cost models are not the key to building reliable AI applications. Instead, they claim that performance should be prioritized over cost when developing effective, high-quality solutions. The question is whether Amazon’s Nova models, which Amazon says are “as good or better” than competing AI software in many but not all benchmark tests, are viewed by developers as good enough to support them to encourage people to switch.

This could be the balance Amazon is aiming for, but it’s unclear whether they’ve found the right path. I spoke yesterday with Rohit Prasad, Amazon’s SVP and chief scientist for AGI (artificial general intelligence), who told me that the name Nova is purposeful – it signals a new and very different generation of AI models of “exceptional quality.”

When I asked why Amazon didn’t approach Anthropic to develop new models for it, Prasad pointed to Amazon’s own “urgent” internal customer needs like video creation – something that he said, to the best of his knowledge, Anthropic doesn’t offer. “We have our Prime Video team that wants to recap seasons, and they can do that with the model’s video understanding capability,” he said. “Amazon ads needed models that could generate videos.”

Prasad did not comment on Amazon’s long-term roadmap regarding its AI models, but said there will be “further paradigm shifts,” including more powerful models. In the meantime, he said that the Nova models are available to all internal Amazon teams, including the team working on a new generative digital assistant Amazon’s Alexa (which, as I reported back in June, is a long, less than successful one slog).

Amazon wants to offer its customers the choice between models from different providers, he emphasized. The question is, will the same strategy that worked so well for Amazon’s AWS in the past – offering low costs, product choice and flexibility – pay off again in this new era of AI?

How far are you from the world’s most powerful business leaders? Discover who made our brand new list of the 100 most powerful people in business. Plus, learn more about the metrics we used to create it.