A technician installs a temperature sensor on an industrial tank. The device communicates in real-time with a dashboard hosted in the cloud. Three years earlier, the same operation required a daily manual reading and an Excel sheet.
This shift can be seen in dozens of professions. It has accelerated since the beginning of 2024 with the rise of generative artificial intelligence, connected objects, and new regulatory constraints that are game-changers for both companies and individuals.
Energy Cost of Data Centers: The Hidden Bill of AI
When deploying a generative AI model, the first limitation is not software-related. It’s the electricity bill. According to the International Energy Agency (IEA), data centers consumed about 415 TWh of electricity in 2024, accounting for nearly 1.5% of global consumption. The IEA forecasts a notable increase in this demand by 2030, driven by the training and use of AI models.
On the ground, this translates into local tensions. A large data center can consume several million gallons of water per day for cooling, with concentrated effects on the territory that hosts it. This water footprint is rarely discussed when touting the merits of a chatbot or an image generator.
To follow technological innovations on C Nouveau, one must also look at what they consume. Innovation is not only measured by model performance but also by its long-term sustainability.
European AI Act: What the Regulation Changes Practically
The European Union adopted the AI Act in 2024, and its effects are already being felt in the technical choices of companies. The text classifies AI systems according to their risk level and imposes progressive obligations on both providers and users.
In practice, a company using an automatic CV sorting tool (classified as “high risk”) must now document the model’s operation, ensure decision traceability, and allow for human oversight. For a recruitment SME, this means reviewing supplier contracts, training teams, and sometimes changing tools.
Feedback varies on this point: some organizations see the AI Act as a barrier to adoption, while others view it as a welcome clarification that reduces legal risk. What is certain is that the era of deploying AI without a legal framework is over in Europe.
Concrete Obligations for Companies
- Document the datasets used to train or refine an AI model, with source traceability
- Implement a human oversight mechanism for any system classified as high risk (recruitment, credit, health)
- Inform end-users when they interact with an AI system, including for chatbots and generated content
- Conduct a compliance assessment before market launch for the most sensitive systems
Connected Robots and Sensors: Automation Beyond the Factory
Robotics is often associated with automotive assembly lines. In 2024, robots are leaving the factory. In logistics, fleets of autonomous mobile robots sort packages in warehouses where temperature, humidity, and flows are controlled by IoT sensors connected to cloud solutions.
Embedded intelligence in sensors reduces the need for bandwidth. Instead of sending all raw data to a remote server, the sensor filters, analyzes, and only transmits alerts or summaries. This is known as edge computing, and it changes how digital infrastructure is sized.
In healthcare, wearable devices continuously measure physiological parameters and transmit data to the treating physician. The benefit is not only in comfort for the patient: it’s the early detection of anomalies that justifies the investment.
Three Sectors Where IoT Transforms Operations
Precision agriculture uses ground sensors and drones to adjust irrigation plot by plot. Predictive maintenance in industry relies on vibrations and temperatures analyzed in real-time to anticipate failures. And in construction, smart meters and connected thermostats help reduce energy consumption without daily manual intervention.
Generative AI in Business: What Works and What Gets Stuck
There has been much talk about ChatGPT, Midjourney, and code generation tools. On the ground, the adoption of generative AI in companies follows a recurring pattern: initial enthusiasm, testing phase, then confrontation with limitations.
The most mature use cases involve drafting standard emails, summarizing long documents, and assisting in software development. For these tasks, time savings are measurable, and the risks of error remain manageable with human proofreading.
Where it gets stuck is with sensitive data. Feeding an AI model with internal data poses a confidentiality issue that many companies had not anticipated. On-premise solutions exist, but their cost and complexity hinder medium-sized organizations.
- Assisted writing and document summarization: rapid adoption, visible return on investment within weeks
- Code generation: useful for speeding up prototyping, but human review remains necessary for the quality of the deliverable
- Customer data analysis: promising, but constrained by GDPR and the quality of available datasets
Technological innovation in 2024 is not just a race for features. Between the energy pressure of data centers, the regulatory framework of the AI Act, and the operational realities of automation, each technical advancement comes with constraints that cannot be ignored. The companies that succeed are those that integrate these parameters from the choice phase, not after deployment.
