At this pace, the cloud market has never been more strategic for IT departments, and competition between hyperscalers has never been more intense.
In this context, AWS remains the leader with around 30% market share, ahead of Microsoft Azure (25%) and Google Cloud (13%). But AWS's leadership is no longer a given: Azure is catching up, Google is accelerating, and generative AI is redistributing the cards. In Q1 2026, AWS generated $37.6 billion in revenue, up 28% year-over-year — its strongest growth in three years, driven precisely by AI.
Beyond the numbers, why is AWS still a strategic partner for businesses looking to transform? And what changed in 2026?
AWS doesn't just provide cloud infrastructure. For nearly twenty years, Amazon has set the standards for innovation in big data, AI, and now accelerated computing.
Maybe you've heard of S3, Hadoop, or Redshift? These technologies have structured the modern data ecosystem. SageMaker, launched in 2017, democratized machine learning in business. Bedrock, released in 2023, made AWS the go-to platform for operating foundation models in a secure and governed manner. And more recently, the Trainium and Inferentia chips, designed by the Annapurna Labs teams at AWS, are redefining the economics of training and inference at scale.
Where other hyperscalers rely on the brand effect or on vertical integration with their historical ecosystem, AWS has always bet on the builder. An open, modular platform, on which you build what you need, at your own pace, with unparalleled functional depth.
Starting a cloud project with AWS means first and foremost carrying out an accurate audit of what already exists. What applications are in place? Who uses them and for what purpose? What are the current infrastructure costs, bottlenecks, regulatory constraints? Once this diagnosis is made, it becomes possible to define a coherent migration trajectory.
This process should not be seen as a simple technical migration. It's an opportunity to rethink how a business uses its digital assets. AWS helps businesses rearchitect their applications to take full advantage of cloud-native services, ensuring a secure and non-disruptive transition.
Several approaches are possible depending on the level of maturity:
- **Lift and shift** to migrate quickly without rewriting, useful for freeing up a datacenter or rationalizing a legacy fleet
- **Replatforming** to benefit from managed services (RDS, ElastiCache, EKS) without rewriting everything
- **Refactoring** to rebuild in cloud-native, using Lambda, microservices, serverless, event management
- **Re-architecting** for organizations that want to integrate AI and data capabilities as soon as they migrate
It is this flexibility that is the strength of AWS in the face of CIOs: there is not a single path, there is the one that corresponds to the business context.
AWS made the historic choice to focus on B2B. The tools and services offered are designed to integrate with existing systems, whether they are chatbots, CRM, intranets, ERP or data platforms. Unlike other cloud players who are positioning themselves on more mainstream solutions, AWS offers a robust, flexible, and high-performing platform for custom projects.
Take Netflix, Airbnb, or Capital One as an example: these companies have chosen AWS to manage their data processing and AI needs at a very large scale. Why? Because they know that with AWS, they not only benefit from a powerful infrastructure but also from expertise in security, reliability, and compliance that few players can match.
In France, groups such as Société Générale, Engie, Veolia or Decathlon have also undertaken major projects on AWS, often in an assumed multi-cloud logic.
The AWS ecosystem has evolved very rapidly over the past 18 months. Four structural trends deserve to be closely monitored.
Amazon Bedrock has established itself as the reference layer for exploiting foundation models in business, with governance, security, and costs under control. The platform now offers a catalog of models that is unique on the market: Claude d'Anthropic, now trained and served on Trainium, Llama from Meta, Mistral, Titan, and since April 2026 the GPT-5.4 and soon GPT-5.5 models from OpenAI, and since April 2026, the GPT-5.4 and soon GPT-5.5 models from OpenAI, on the same infrastructure.
It is a major change. For the first time, businesses can access OpenAI models without using Azure, and combine Claude, GPT, and Llama in the same architecture, with unified governance and granular cost control (tag billing arrived at the end of April 2026).
With Bedrock AgentCore, AWS is pushing an industrialized approach to AI agents. The service offers a managed harness, a CLI for deploying agents with the governance of an infrastructure as code, and native integration with everyday tools. Bedrock Managed Agents, launched in preview in April 2026 and in particular supported by OpenAI, make it possible to build agents that maintain context on long tasks, at the production scale.
For CIOs, this is the time when generative AI is moving from POC to operational production.
The other major movement of 2026 is the AWS silicon strategy. The Trainium 2 and Trainium 3 chips (Trainium 4 expected in 2027) allow AWS to offer significantly lower training and inference costs than NVIDIA GPUs for a variety of uses. In April 2026, Anthropic committed more than $100 billion in investments in AWS over ten years, with up to 5 GW of Trainium capacity to train and serve Claude. OpenAI, as part of its new agreement with AWS, plans 2 GW of Trainium capacity for its workloads.
For businesses, the interest is clear: as these strategic agreements come to fruition, the costs of inference of foundation models on AWS should continue to fall, making generative AI economically viable on use cases that were not viable yesterday.
Announced at the end of April 2026, Amazon Quick is an AI assistant for work, with a desktop application, Free and Plus plans, and integrations that are expanding rapidly. It's AWS's answer to Microsoft Copilot, with a more open, multi-application approach. For CIOs who are hesitant to lock themselves into the Microsoft 365 ecosystem, this is an alternative to watch out for.
Three practical lessons for decision-makers.
Multicloud is becoming the norm, plus the exception. With OpenAI and Anthropic now available on the three major hyperscalers (AWS, Azure, GCP), with OneLake on the Microsoft side that interoperates natively with S3, and with open standards such as Iceberg or OpenTelemetry that are becoming established, the argument of technical lock-in is losing its strength. The real issue is multi-cloud governance, not the choice of a single vendor.
AI is moving the center of gravity of cloud costs. Cloud budgets have historically been dominated by classical compute, storage, and database. In 2026, the AI bill (Bedrock inference, training, RAG, agents) becomes a significant line. Andy Jassy reported that customer spend on Bedrock jumped 170% between Q4 2025 and Q1 2026. CIOs need to incorporate this dimension into their FinOps.
AWS skills remain rare and expensive in France. Demand continues to greatly exceed supply on the profiles Solutions Architects AWS, DevOps Engineers AWS, Data Engineers Bedrock, and now AI Engineers specialized agents. The remuneration ranges for confirmed profiles in France are typically between 60 and 90K € for a confirmed Solutions Architect or DevOps, 80 to 120 K € for seniors and leads, and beyond for principals with recognized platform expertise.
AWS is no longer the undisputed leader it was five years ago. Azure has caught up a significant portion of its enterprise cloud gap, Google Cloud has strengthened its position on AI and analytics, and the competition on silicon with NVIDIA, AMD, and the other hyperscalers' proprietary chips is fierce.
But AWS remains, in 2026, the broadest cloud platform, the most functionally profound, and the best positioned to support businesses on their AI trajectory. The combination Bedrock + Trainium + AgentCore is, to date, one of the strongest proposals on the market to industrialize generative AI in production.
For an IT department that is now setting its cloud strategy at 3-5 years, AWS remains a choice that cannot be seriously ruled out. Not by default, but because the ecosystem, operational maturity, and innovation trajectory continue to put AWS ahead.
At Nacimut, we support companies in recruiting their Cloud, Data and AI experts, in France and internationally. To discuss your needs or the AWS profile market, write to us.
