THE POINT / WHAT CHANGES FOR YOU

From news to practice.

This selection is a map of the sector: it distinguishes who develops models, who offers infrastructures, and who integrates AI into services. It is not an absolute ranking.

Who is building the AI world

Artificial intelligence has no single centre. Some develop the models, others bring them into applications, and some build the infrastructure needed to run them. These ten companies help understand the transformation from complementary perspectives. The desk's selection considers research activity, product offerings, and ecosystem role: it is not a ranking by turnover, number of users, or absolute quality of models. The order of presentation does not indicate a ranking.

1. OpenAI — ChatGPT and models for building applications

OpenAI develops artificial intelligence models and products such as ChatGPT. Its role spans two levels: the tools with which people and businesses interact every day and the technologies that developers integrate into their own applications. It is a company to follow to understand how research on models becomes an accessible service and how ways of working with digital assistants are changing.

Visit the official OpenAI website ↗

2. Google — DeepMind research, from Gemini to science

Google unites, with DeepMind, research ranging from the Gemini models to scientific applications. Its ecosystem includes image, video, and music generation, robotics, and projects like AlphaFold, dedicated to predicting protein structures. Its presence shows how AI has become a set of different technologies: conversation, perception, content creation, and research tools.

Visit the official Google website ↗

3. Microsoft — Artificial intelligence within work

Microsoft brings AI into work tools and platforms used by organisations. Its offering connects assistants, application development, and business services. The point to observe is integration into processes: how intelligent functions enter daily activities, which data they use, and which controls remain with people. It is a useful case to observe the transition from individual trial to enterprise adoption.

Visit the official Microsoft website ↗

4. NVIDIA — The chips and software that make AI work

NVIDIA operates at the infrastructure level of artificial intelligence, with GPUs, models, and software tools. Its platform includes components to build agents and Physical AI applications capable of interacting with the physical world. Following it helps understand that behind an assistant there is not just an algorithm: computation, systems, networks, and tools are needed to execute the models.

Visit the official NVIDIA website ↗

5. Amazon — AWS and model access in the cloud

Amazon participates in the transformation through AWS, its cloud services business. Amazon Bedrock offers a platform to develop generative applications and agents with models from various providers. The perspective is that of infrastructure and enterprise adoption: choosing a model, connecting it to data, and managing the application during use. The cloud thus becomes one of the places where research is turned into services.

Visit the official Amazon website ↗

6. Meta — Models, developer tools, and multimodal content

Meta complements its products with an ecosystem of models and tools for developers. The official portal presents the Muse family and solutions for code, agents, images, voice, and visual analysis. The variety of projects makes the company an observatory on the evolution of multimodal AI. For developers, it is important to distinguish the possibilities offered by the APIs from those of models with available weights, always checking the terms of use.

Visit the official Meta website ↗

7. Anthropic — Claude and research on reliability

Anthropic is the company developing Claude and conducting research on safety, interpretability, and control of AI systems. The assistant is usable through conversational interfaces and APIs. Its presence in the sector brings to the fore a concrete question: how to make models useful and verifiable for the tasks assigned to them. The producer's evaluations are a starting point to compare with evidence in one's own work.

Visit the official Anthropic website ↗

8. Alibaba — Qwen and the Alibaba Cloud platform

Alibaba is present in AI through Alibaba Cloud and the Qwen family of models. The platform links infrastructures, development, customisation, and application deployment. Model Studio offers access to models and tools to build generative services. It is a company to follow to observe the contribution of the Chinese ecosystem and the relationship between model availability, computing capacity, and services intended for businesses.

Visit the official Alibaba website ↗

9. DeepSeek — Models, research, and API access

DeepSeek develops artificial intelligence models and offers a conversational service and an API platform. The official site provides references to model families and technical documentation. Its presence broadens the international comparison among approaches, products, and access methods. To assess its use, it is necessary to verify the version used, the available documentation, and service conditions, avoiding judgements based solely on name recognition.

Visit the official DeepSeek website ↗

10. Mistral AI — Open models and AI governance in organisations

Mistral AI offers models with available weights and tools to customise, deploy, and manage AI applications. Its offering emphasises control over infrastructure and data, with various execution modes. It is a key reference in the debate on technological sovereignty: where the model runs, who governs it, and how adaptable it can be. Accessible weights and open source software remain concepts to distinguish by reading the licences.

Visit the official Mistral AI website ↗

Three levels to understand the competition

Models, deployment, and infrastructure are often intertwined within the same group. To understand an announcement, it is useful to ask which problem it addresses, who it targets, and which results are verifiable. These cards offer a first map, not a purchase recommendation: products, availability, and terms of use can change. Links lead to the official pages used for presentations.

ChatGPT, Gemini, Claude: read the comparison between assistants ↗

Sources and documents

Document date or reference period: Official sites consulted on 23 October 2026.

Text prepared with AI assistance based on the indicated sources. Interpretive evaluations are distinct from data and statements from sources.

Report an update or correction →