Amazon's AI Woes: A Capacity Crunch and Performance Issues Drive Customers to Competitors
Amazon's AI division is facing a critical challenge. This summer, Amazon Web Services (AWS) struggled to meet the soaring demand for AI services, causing a capacity crunch that pushed some customers to seek alternatives. The issue was so severe that it led to a loss of real revenue for AWS's flagship AI product, Bedrock.
Bedrock, the heart of Amazon's AI strategy, allows developers to access powerful models like Anthropic's Claude and Meta's Llama. However, the service hit a major roadblock when it encountered "critical capacity constraints" during the summer, as revealed by an internal document from July. This issue forced some customers to turn to rival cloud services, with Google's cloud service being a notable beneficiary.
The capacity shortages resulted in significant financial consequences. Epic Games, for instance, relocated a $10 million Fortnite project to Google Cloud due to AWS's inability to provide sufficient quota for Bedrock. The document also highlighted the potential loss of revenue from other clients, such as oil trader Vitol, who considered moving projects away from AWS due to prolonged quota approvals. These delays affected projected sales, with companies like Atlassian and GovTech Singapore awaiting quota increases, totaling at least $52.6 million in potential revenue.
The document painted a dire picture, stating that Bedrock was "experiencing critical capacity constraints that are threatening customer adoption and potentially causing substantial revenue loss across multiple industries." This situation underscores the delicate balance between high demand and the ability to meet it, as customers may turn to competitors if their needs are not satisfied.
Amazon CEO Andy Jassy has emphasized the need to bolster cloud infrastructure, especially AI chips and data center power. However, the company's progress in resolving these issues remains unclear. Former and current employees confirmed that the capacity crunch remained a top concern through September.
An Amazon spokesperson acknowledged Bedrock's rapid growth and assured that AWS is working to meet the demand. They emphasized Amazon's culture of customer feedback, stating, "At Amazon, we're vocally self-critical... This internal candor is a feature of our culture... We're grateful for all customer feedback because it helps us make Bedrock even better."
But here's where it gets controversial: While Amazon is addressing the issue, the broader AI market is evolving rapidly. The document warned that AWS lacked a cohesive product vision for AI inference, allowing competitors like Databricks, FireworksAI, and Nvidia's Dynamo to gain ground. Without a clear strategy, AWS could miss out on a massive AI opportunity.
As investors worry about the tech industry's massive AI spending and a potential bubble, Amazon's capacity issues are a double-edged sword. They indicate strong customer demand but also pressure big tech companies to spend heavily, potentially exacerbating the AI bubble.
Amazon's planned capital expenditures of $125 billion this year and more in 2026 reflect its commitment to AI. AWS's revenue growth of 20% year-over-year to $33 billion last quarter further highlights the importance of this market.
And it's not just about capacity: Performance issues also played a role in customers' decisions to leave Bedrock. Latency and missing features were significant concerns. Customers using Anthropic's Claude models through Bedrock chose Anthropic's platform or Google Cloud due to ongoing capacity, latency, and feature parity issues. Companies like Figma, Intercom, and Wealthsimple migrated their workloads for these reasons.
Google's Gemini models posed a growing threat to Bedrock, offering larger quota limits and better performance. The internal report showed that Gemini outperformed Bedrock in multiple benchmarks, and Gemini Flash provided comparable quality at a lower cost. This led some startups to switch, with TainAI saving $85,000 a day by moving 40% of its Claude workloads to Gemini Flash.
As Amazon prepares to reveal more about Bedrock and its cloud strategy at the re:Invent conference in December, the company must address these challenges to maintain its competitive edge in the rapidly evolving AI market. Will Amazon's efforts be enough to keep customers satisfied and prevent further losses to rivals? The coming months will be crucial in determining the future of AWS's AI ambitions.
