In response to Google's challenge, Amazon AWS urgently restructures its AI cloud services

On Thursday, June 12, according to media reports, in the face of customer loss and the encroachment of competitors, Amazon Web Services (AWS) is preparing to comprehensively upgrade its AI cloud platform "Bedrock."

The reason is that in building AI agents, the user experience of AWS is not as flexible and convenient as that of Google Cloud and Microsoft Azure, leading some customers to turn to other platforms. If AWS continues to lose market share, the entire Amazon will be affected.

AI agents are not ordinary chatbots, but intelligent assistants that can "get things done," such as automatically organizing bills and fixing website issues. Due to the high computational power and token consumption required for such applications, along with broad profit margins, they have become a key area of competition among major cloud vendors.

As early as March this year, AWS announced the establishment of a new department dedicated to developing "AI agents," demonstrating its determination to ramp up its layout.

Why does AWS feel immense pressure?

Amazon CEO Matt Garman stated that AI agents have the potential to become the next billion-dollar business for AWS. That said, the reality is not so smooth.

Some customers have expressed dissatisfaction, noting that developing AI applications is easier and more user-friendly on Google Cloud and Microsoft Azure, especially in building AI agents using open-source software.

For example, Google Cloud's Vertex AI platform is more open, allowing customers to call any AI model across platforms, such as Microsoft's OpenAI model and AI models from other clouds like AWS, as long as security and network settings are properly configured.

Meanwhile, Microsoft is also comprehensively upgrading its cloud products to seize the AI agent transformation wave. It has launched low-cost models on Azure, adopted open-source protocols, and introduced customizable development tools to help customers create their own intelligent agents.

In contrast, AWS's Bedrock currently lacks this flexibility.

Firstly, the model selection is limited, and it still cannot be compatible with OpenAI models. Many customers hope to run code on AWS while calling models from Microsoft, Google, etc., but the current process is cumbersome, requiring customers to use multiple cloud management platforms to operate each vendor's models and services separately.

Secondly, Bedrock has set more development restrictions, allowing customers to only use AWS's own tools and specific AI models, making it more difficult for customers to build AI intelligent agents and reducing flexibility. Some customers have stated that building AI applications and agents on Bedrock is not easy compared to competing cloud platforms.

Naturally, customers feel less inclined to use it, with some even bypassing AWS to use open-source tools from startups like Hugging Face and PydanticAI to develop agents themselves.

Additionally, some customers have encountered difficulties accessing AI company Anthropic's models through AWS. A CEO of a software company using AWS stated that they had to abandon obtaining models through AWS and instead directly use Anthropic's API to access the same models.

Bedrock is undergoing a "major transformation"

In the face of customer loss and the encroachment of competitors, AWS is preparing to comprehensively upgrade Bedrock The new Bedrock will be more open and flexible, allowing businesses to more easily access a wider variety of AI models and development tools to better build applications like AI agents.

However, it is still unclear whether this upgrade will allow AWS users to use OpenAI's models. Models like GPT-4o can only run on Microsoft Cloud.

In addition, AWS is also promoting the development of AI agents through other means. In mid-May, AWS released a set of open-source development tools called Strands Agents, enabling developers to build agents using text prompts. Amazon's CEO has publicly expressed support for this project.

The upgrade plan for Bedrock has not been publicly disclosed, and media reports indicate that Swami Sivasubramanian, head of AWS's AI agent division and a veteran since 2019, is in charge of the Bedrock upgrade project. Amazon may announce this news at its annual customer summit in New York in July.

The Sense of Crisis Behind: Amazon Cannot Afford to Lose

The Bedrock upgrade is part of the fierce competition among cloud service providers. AI agent applications "consume computing power and use tokens," with significant profit margins, making it a battleground for cloud vendors.

The core goal of AWS's Bedrock upgrade is to prevent enterprise customers from shifting to AI competitors like Microsoft and Google or third-party APIs.

AWS's senior management is racing against time to ensure that its cloud computing leadership is not shaken, as more and more traditional enterprises begin to embrace AI.

The outcome of this AI competition is crucial because, with its leading position in the cloud market, AWS has become Amazon's main source of profit. If AWS continues to lose market share, the entire Amazon will be affected.

Although AWS remains the leader in the cloud market, Microsoft and Google, with their earlier and faster AI deployments, have revenue growth rates that exceed AWS by 10 to 15 percentage points. In the face of intensified competition, AWS must hold its ground. However, in terms of scale, Microsoft and Google are still significantly smaller than AWS