Amazon’s AI Capacity Shortage Drives Customers to Competitors Like Google

Amazon’s Summer Struggles in AI Demand

This summer, Amazon’s cloud computing division, Amazon Web Services (AWS), faced unexpected challenges as it struggled to keep pace with surging demand for artificial intelligence (AI) services. Internal documents reveal a landscape marked by missed revenue opportunities for its flagship AI product, Bedrock—a service that was intended to position AWS at the forefront of the competitive AI arena.

The Heart of Amazon’s AI Initiative: Bedrock

At the core of Amazon’s AI endeavors lies the Bedrock service, designed to empower developers by providing access to advanced AI models, including partners like Anthropic’s Claude and Meta’s Llama. Despite this strategic focus, the summer months revealed significant limitations as Bedrock encountered “critical capacity constraints.” These constraints were so severe that some AWS customers were pushed to seek alternatives, notably Google Cloud, raising urgent concerns about AWS’s retention capabilities amid growing competition.

The Financial Fallout

The consequences of these capacity limitations have been stark. Reports suggested that Amazon could have lost or experienced delays amounting to tens of millions of dollars in potential revenue. For instance, Epic Games faced hurdles with a $10 million Fortnite project due to insufficient quota on Bedrock—quota limits that determine how much intelligence a customer can harness through AWS’s AI services. Similarly, oil trading giant Vitol hovered at the brink of transferring projects away from AWS, anticipating a potential $3.5 million revenue drop.

Other customers like Atlassian and GovTech Singapore also found themselves waiting on quota increases, leading to a standstill in over $52.6 million in projected sales during the summer. This situation highlights not only the financial risks of AWS’s capacity issues but also the competitive pressure sparked by rival services eager to capture these clients.

Leadership’s Response

Despite the looming challenges, Amazon’s CEO, Andy Jassy, has emphasized ramping up cloud infrastructure as a remedy for these struggles. As AWS aims to increase its capacity, Jassy shared in an October earnings call that extensive upgrades had resulted in an additional 3.8 gigawatts of power—more than any of its competitors achieved, and plans to double capacity again by 2027 are underway. However, he acknowledged that a significant portion of Bedrock’s current workloads are primarily from a select number of large clients, prompting a potential need to include more mid-sized companies in future growth plans.

Customer Feedback and the Shift to Competitors

The urgency for improvement has driven AWS to focus on proactive customer feedback, as internal documents indicated a wide array of industries affected by slow approval processes for capacity increases, delays, and overall service limitations. Companies across finance, gaming, and technology sectors—including HelloFresh, Zalando, and Ryanair—experienced disruptions prompting them to consider alternatives like GCP, OpenAI, and Anthropic.

Additionally, concerns around latency and lack of essential features compounded customer frustrations, with many migrating to competitor platforms. Companies like Figma and Thomson Reuters shared experiences where they opted for Google Cloud primarily due to faster service capabilities or superior compliance options not yet available through Bedrock. This trend signals a critical need for AWS to address its service issues to maintain its standing in the rapidly evolving AI market.

Rising Competition: Google’s Challenge

Google’s advancements are especially notable, as internal AWS reports suggested that Bedrock was losing traction to Google’s Gemini models, which not only offer superior quota limits but also outperformed Bedrock in various benchmarks. Customers have reported significant cost savings and enhanced efficiency by switching from Bedrock to Gemini Flash, making it a substantial concern for AWS leadership.

The internal document also lamented a visible lack of a cohesive product vision for the AI inference domain, essential for positioning AWS favorably against emerging competitors. Without clear strategic direction, AWS risks losing out on a lucrative segment of the booming AI industry.

The Broader Implications

For Amazon, these capacity challenges reflect both the intense demand for AI capabilities and the perplexing nature of maintaining competitiveness amidst dizzying technological advances. As various sectors lean heavily into AI integration, the stakes have become higher for AWS, which is positioned not just as a service provider but as an essential player in shaping the AI landscape.

Furthermore, investors and stakeholders have grown wary, considering the substantial financial implications and potential market bubble fueled by massive AI expenditures. Indeed, Amazon’s financial forecasts indicate plans for a staggering $125 billion investment into capital expenditures in the upcoming years, even as AWS celebrated a notable revenue increase—$33 billion last quarter—marking its fastest growth cycle since 2022.

The Road Ahead

As AWS navigates these turbulent waters, the path toward future improvements centers around swift capacity expansion, addressing key customer feedback, and refining their vision for AI service delivery. The company faces ongoing challenges, but the drive for continuous improvement remains a fundamental aspect of its operational philosophy, as voiced by Amazon’s teams, emphasizing that embracing criticism is essential for long-term success and sustainable growth in the competitive landscape of AI technology.

Amazon CEO Andy Jassy

Noah Berger/Noah Berger