Amazon CEO predicts 6 to 8 AI models will be ‘good enough’ for almost everything

Amazon CEO Andy Jassy foresees the AI model landscape becoming populated with 6 to 8 models that are roughly equivalent in capability, rather than a single standout winner, according to insights shared during Amazon's Q3 2026 earnings call. Jassy emphasized that businesses will need to select among these models, rather than committing to just one.

He cited the rapid shifts in tech dominance over the past two decades as evidence, pointing to examples like databases, analytics, and TensorFlow being supplanted by PyTorch. Jassy expects a similar dynamic in the AI model realm, with different models excelling in different areas such as reasoning, coding, visual tasks, speed, and cost per token.

This multi-model approach allows customers to leverage the best tool for each specific task, rather than relying on a single provider. Amazon Bedrock, Amazon's managed service for generative AI, currently offers access to over 100 foundation models from various providers including Anthropic, OpenAI, Meta, and others through a single API.

This API-centric approach enables developers to seamlessly switch between models without the need to rewrite their integrations, although it's important to note that switching models may not always be straightforward due to variations in context limits, tool capabilities, and supported features. Security is another key consideration, with AWS ensuring that Bedrock does not expose customer prompts or responses to model providers, nor does it use this data to train the underlying models.

Jassy highlighted these protections as an additional reason enterprises choose Bedrock. The company's strategy is to provide a neutral access layer that monetizes the ongoing shifts in AI model dominance, similar to how Google's Vertex AI and Microsoft's Azure AI Foundry approach the market. Based on Jassy's observations, Bedrock is expected to grow into a service as significant as EC2, AWS's largest business.

Early data supports this prediction, with Fortune 500 companies spending nearly seven times more on Bedrock year-over-year, with token volume increasing nearly 17-fold, and being ranked as the fastest-growing multibillion-dollar product in AWS history. This growth is further bolstered by the fact that most Bedrock traffic currently runs on Amazon's own Trainium chips, which helps keep prices low while protecting margins.