Amazon: The Winning Strategy Without the Best AI Model

Amazon and the Diversity of AI Models

Amazon's financial results released on Thursday highlighted an emerging trend in the tech sector: it is not necessary to own the best artificial intelligence models to succeed. Amazon reported a significant 37% increase in revenue from its Amazon Web Services (AWS) division. This growth is attributed to the use of a variety of AI models from renowned providers like Anthropic and OpenAI. In contrast, Amazon's efforts to develop its own AI models appear disorganized, according to recent reports.

During a conference call with analysts, Andy Jassy, Amazon's CEO, downplayed the importance of owning a dominant AI model. He emphasized that customers prefer to use multiple leading models, which has accelerated the rapid development of Bedrock, Amazon's platform that provides access to these models. Jassy stated, "AWS and Amazon can have an extremely successful business without its own cutting-edge model." He added, "There will not be one model to rule the world."

A New Approach to AI

Jassy asserted that AWS and Amazon can thrive without possessing the most advanced AI model. According to him, there will not be a single model that dominates the sector. This perspective marks a shift from the traditional approach, where companies heavily invested to create the most powerful models. Since the rise of ChatGPT in 2022, competition has intensified around model performance.

However, the dynamic has shifted towards optimizing model efficiency, often referred to as "intelligence per dollar." This new orientation could be more advantageous for Amazon than for some of its competitors.

Amazon's Strategy Regarding AI Models

Historically, Amazon has not focused its efforts on developing its own AI models. Although the company created internal models under the Nova brand, these have failed to compete with market leaders. Recently, Amazon has adjusted its strategy, scaling back the Nova range and redirecting its resources towards new AI model initiatives.

Jassy's statements indicate that Amazon is not solely relying on its own models but aims to make AWS the preferred platform for accessing the best AI systems, whether developed by Amazon or others.

Bedrock: The Cornerstone of Amazon's AI Strategy

Rather than concentrating solely on model performance, Amazon has bet on Bedrock as the central element of its AI strategy. Jassy stated last year that AWS was building Bedrock to become the largest inference engine in the world, with the ambition to compete with EC2, AWS's flagship cloud computing service.

At the recent conference, Jassy explained that companies developing AI applications want access to a variety of models, as each system can excel in specific tasks at different times. He said, "If you are a company building significant AI applications, you want to ensure that you have the ability to use all available models." This flexibility gives Bedrock a significant competitive advantage.

Amazon's Ongoing Commitment to AI Development

Despite this diversified approach, Jassy reaffirmed Amazon's commitment to developing its own AI models. Having a cutting-edge model would allow Amazon to better control costs for its applications and AWS customers, while also providing the opportunity to prioritize essential features.

For example, Alexa has reduced its reliance on costly models from Anthropic by adopting Amazon's internal models, which has helped lower costs.

Jassy anticipates that in the coming years, there will be "at least half a dozen models that will be comparatively good against each other." He stated, "They will all be on Bedrock, and one of them will be ours."