Amazon Web Services

AI Infrastructure for Everyone

Founded: 2006 (AWS); Amazon founded 1994 | HQ: Seattle, WA | Key People: Andy Jassy (CEO, Amazon), Matt Garman (CEO, AWS) | Cloud Market Share: ~31% | AWS Revenue: $142 billion run rate (Q4 2025) | Key AI Products: Amazon Bedrock, Amazon SageMaker, Trainium/Inferentia chips, Amazon Q, Amazon Quick AI, Alexa for Shopping (white-label), Amazon Bee wearable, Amazon GenAI Creators' Fund

The Origin Story

Amazon Web Services launched in 2006 as a radical experiment: selling Amazon's internal infrastructure as a service to external developers. By the time the AI era arrived, AWS had become the world's largest cloud provider, generating over $100 billion in annual revenue with a 31% market share. But when OpenAI's ChatGPT ignited the generative AI revolution in late 2022, AWS found itself in an unfamiliar position—playing catch-up. Microsoft's exclusive partnership with OpenAI gave Azure a perceived AI advantage, and Google's internal DeepMind capabilities gave Google Cloud a strong narrative. Amazon's response was characteristically pragmatic: rather than building a single frontier model, AWS would build the infrastructure layer on which all AI models run, positioning itself as the Switzerland of AI computing. The strategy centered on Amazon Bedrock, launched in April 2023, which provides managed access to models from multiple providers—including Anthropic, Mistral, Cohere, and Amazon's own Titan models—through a single API.

Key Milestones

The September 2023 investment of $1.25 billion in Anthropic, expanded to $4 billion by year-end and ultimately to $8 billion total, was Amazon's signature AI move. The investment was structured as convertible notes that have appreciated spectacularly: Amazon's Anthropic stake grew from $8 billion invested to $60.6 billion in value by Q4 2025, generating over $12 billion in unrealized gains. Anthropic selected AWS as its primary cloud provider, using Amazon's custom Trainium and Inferentia chips for training and inference. Amazon Bedrock became the centerpiece of AWS's AI strategy, offering enterprises a model-agnostic platform that avoids vendor lock-in to any single AI provider. By early 2026, AWS AI revenue was running above $15 billion annually—nearly 260 times larger than AWS revenue at the same stage of its cloud lifecycle. Analysts project Anthropic's contribution to AWS revenue will grow from $3.9 billion in 2025 to $25 billion by 2027. Amazon's custom silicon strategy matured significantly. The annual revenue run rate for Amazon's chip business—including Graviton (general purpose), Trainium (AI training), Inferentia (AI inference), and Nitro (networking)—exceeded $20 billion, growing triple-digit percentages year-over-year. Trainium2, announced in late 2024, offered cost savings that Amazon claims provide several hundred basis points of operating margin advantage versus relying solely on NVIDIA GPUs. Amazon committed $200 billion in AI capital expenditure for 2026, including $50 billion earmarked for expanding AI infrastructure for U.S. federal agencies. AWS revenue growth reaccelerated to 24% year-over-year in Q4 2025, driven partly by AI workload demand. The company's Anthropic partnership has become a core competitive differentiator, with Claude models available as first-class citizens on Bedrock.

The "What's Next with AWS" event on April 28, 2026, marked a significant expansion. Amazon Quick AI, a new AI assistant, was launched to compete directly with Microsoft Copilot and Google Gemini for enterprise productivity. More strategically, AWS announced that OpenAI models — including GPT-5.5, GPT-5.4, and Codex — are now available on Bedrock, alongside new Bedrock Managed Agents powered by OpenAI. The move dramatically broadens Bedrock's model catalog and signals AWS's determination to be the single pane of glass for all frontier AI models, regardless of provider. Amazon Connect expanded into four agentic AI solutions for customer service, and Bedrock AgentCore received new optimization capabilities in May 2026 for more efficient agent deployment.

Amazon also invested $50 billion in OpenAI as part of a broader funding round, a move that gives AWS a financial stake in the frontier model ecosystem alongside its existing Anthropic position. The dual investment strategy — backing both Anthropic and OpenAI — ensures AWS has privileged relationships with the two most important AI labs simultaneously.

In late May 2026, Amazon expanded its AI footprint on multiple fronts. The company began selling access to the technology behind its Alexa for Shopping assistant, enabling other retailers like Kate Spade to build their own AI-powered shopping chatbots with personalized recommendations. Amazon also launched the "Bee" wearable—an AI-powered health and activity device that has drawn both interest and privacy concerns from early reviewers. On the content side, Amazon MGM Studios greenlit three AI-animated series (Cupcake & Friends, Love, Diana Music Hunters, Punky Duck) through its GenAI Creators' Fund, though animator Jorge Gutierrez exited the Punky Duck project amid online backlash over AI-generated content. Internally, Amazon shuttered an AI usage leaderboard after employees began assigning AI agents to carry out needless tasks to climb the rankings—a phenomenon dubbed "toxenmaxxing" that reportedly increased costs without generating value. Executive Dave Treadwell cautioned staff: "Don't use AI just for the sake of using AI."

Current Position

AWS's AI strategy is fundamentally about infrastructure and choice. Rather than competing with frontier model labs, AWS provides the compute, storage, and managed services that make AI deployment possible at scale. The Bedrock platform's model-agnostic approach — now encompassing Anthropic, OpenAI, Mistral, and others — appeals to enterprises wary of single-vendor dependency, while custom silicon offers cost advantages for high-volume inference workloads. Amazon Quick AI gives AWS its own branded AI assistant for the first time, competing with Microsoft Copilot and Google Gemini. Amazon Connect's expansion into agentic AI solutions signals that AWS is moving beyond infrastructure into AI-native applications. AWS remains the cloud market leader by a significant margin, and its AI revenue growth rate suggests it is successfully converting that leadership into AI-specific revenue.

What Leaders Should Know

If your organization runs on AWS, the path to AI adoption runs through Bedrock—and that is a defensible choice. The multi-model approach now includes OpenAI's frontier models alongside Claude and others, providing unprecedented optionality in a fast-moving market. Amazon's custom Trainium chips offer 30 to 40% cost savings over NVIDIA GPUs for compatible workloads, which matters at scale. The Anthropic partnership gives AWS customers privileged access to Claude, while the new OpenAI relationship adds GPT-5.5 and Codex to the same platform. Amazon Quick AI provides a ready-made assistant for organizations that want to deploy AI internally without building from scratch. Leaders should note that AWS's dual investment in both Anthropic and OpenAI creates a unique strategic position — AWS has financial stakes in both of the leading frontier labs, reducing the risk of any single lab's trajectory. Negotiate long-term committed spend agreements now, as AWS is pricing aggressively to capture AI workloads.