The AI disruptions of recent years have been just the prologue to what awaits us in 2026: the full integration of AI into data center processes and constructs.
It’s a moment we’ve been preparing for since OpenAI’s ChatGPT brought artificial intelligence into the mainstream in late 2022, sending shockwaves across every field, from academia and mental health care to all types and sizes of businesses.
A truly profound transformation will begin to take place in 2026, as AI becomes increasingly ingrained in all aspects of life and the focus shifts from large language models (LLMs) to AI inference. In some ways, 2026 will be the year the rubber really hits the road when it comes to AI.
AI is rewiring functions and industries
According to the latest McKinsey report State of AI Survey78% of organizations use AI in at least one business function; that’s an increase from 72% at the start of 2024 and 55% the year before. Although the majority of adoption is in sales and marketing, AI is growing rapidly in manufacturing, healthcare, finance and, most importantly, data centers:
- Manufacturers using AI to support demand forecasting improved their accuracy by an average of 30 percentage points.
- Hospitals are using predictive AI for billing automation, appointment scheduling, and proactive identification of high-risk outpatients.
- Financial institutions are leveraging AI for fraud detection, payment optimization and risk management.
- Data centers are increasingly using AI-powered cooling systems and predictive analytics to minimize overheating, reduce energy waste, and improve network efficiency through a better balance between electricity supply and demand.
As adoption increases, AI will not only support business functions; it will transform industries. For example, AI agents operating with little or no supervision will become essential to operations, relying on multiple models and requiring vast computing capacity within AI factories.
The rise of AI factories
An AI factory is a data center that not only stores data but also produces intelligence. Indeed, we are moving beyond model training towards inference. This is where the ROI is realized, and these environments become essential.
Additionally, inference workloads are increasingly varied, from chatbot prompts to real-time analytics in healthcare, retail, and autonomous systems. Although they generally require less power per server than training, inference workloads are increasingly varied and ubiquitous.
They now range from simple chatbot prompts to complex, real-time analytics in healthcare, retail, and other industries using autonomous systems and agent agents. Depending on deployment and workload, inference environments can vary from less than 20 kW for compressed or optimized models up to kW per rack for more advanced agentic use cases.
To keep pace, operators will adopt next-generation GPUs such as the NVIDIA Rubin CPX, expected in late 2026. And combined with NVIDIA Vera CPUs and Rubin GPUs in the NVIDIA Vera Rubin NVL144 CPX platform, this system delivers 8 exaflops of AI compute and 7.5x more AI performance than the NVIDIA GB300 NVL72.
Robotics is becoming very advanced
AI-based robotics will boom in 2026. Beyond long-standing applications such as radiation detection or bomb disposal, AI will extend automation to drones, firefighting systems, search and rescue tools, medical robotics and even passenger transportation.
Again, these technologies require immense processing and network capacity as they rely heavily on high definition video as input.
Additionally, we will see data centers increasingly deploy robotics for security monitoring, server installation, maintenance, cable organization, disk replacement, and optimization of liquid cooling systems.
Digital twins take center stage
In 2026, we will see the rise of digital twins as processing power continues to evolve in AI data centers and advanced platforms are developed, such as NVIDIA’s Omniverse and Cosmos. Data center operators will use digital twins to achieve greater efficiency and accelerate development by designing and simulating highly complex physical objects, systems and processes.
Take for example the power system of a data center itself. Sophisticated ETAP modeling technology can create a virtual replica of a data center’s electrical infrastructure through integration with NVIDIA Omniverse.
Liquid cooling is becoming more widespread
As we well know, traditional cooling cannot support next-generation computing density. By 2026, rack densities are expected to reach 240 kW per rack, rising to 1 MW per rack by 2028, with research exploring the feasibility of 1.5 MW per rack capacity.
This makes liquid cooling inevitable, moving from niche to mainstream as high-density AI clusters continue to dominate.
Sustainability remains essential
Electricity supply will remain a major challenge in 2024. Operators will rely on diverse energy mixes, including natural gas turbines with carbon capture, HVO-powered backup generators, wind, solar, geothermal and battery storage.
According to the International Energy Agency (IEA), Renewable energy currently provides 27% of the electricity consumed by data centers and is expected to meet almost half of the incremental demand growth by 2030.
Expect 2026 to be a critical year, where the impact of AI will shift from being a disruptive force to becoming a fundamental part of business and technology. As AI reshapes every layer of digital infrastructure, tomorrow’s data centers will not only support the technology; they will enable intelligence itself.
By Canninah Dladla, President of the Anglophone Africa Cluster at Schneider Electric