What began as infrastructure to store data is now evolving into a system that generates intelligence at scale—from tokens and code to images, voice, and real-time decisions. The ChatGPT moment did not just accelerate adoption; it fundamentally redefined the role of data centres. Today, they are transitioning into what industry leaders are calling “AI factories.”
For CIOs, this shift is not theoretical—it is deeply operational and immediate.
In this conversation, Manish Kumar, Executive Vice President – Secure Power & Data Centers, Schneider Electric breaks down what this shift to AI factories really means: where the gaps lie, why energy is becoming the new bottleneck, how skills are being redefined, and how enterprises can rethink data centres not just as infrastructure—but as core engines of competitive advantage.
Some edited excerpts:
Can you take us through some of the key trends that you are seeing in the data centre industry?
If you think about data centres, the ChatGPT moment marked a clear inflection point. Data centres are no longer just about storing data. Their mission has fundamentally changed—they are now generating data and tokens. In that sense, they have become factories that generate voice, images, text, and more.
Data centres are evolving into AI factories. That shift has already begun; now it is about the pace at which it scales. The impact of AI factories is horizontal across industries. Whether it is life sciences accelerating research, nurses becoming more productive through better access to information, or programmers writing code faster—the potential of AI across sectors is enormous and, in many cases, underestimated.
If we look back, even with mobile phones, we did not anticipate in the early 2000s how deeply they would reshape the economy. AI is likely to follow a similar trajectory.
At present, we are seeing massive capacity being built globally. The United States has led this for the last two to three years, but now every country recognises the need to invest. In India, the installed base is roughly 1 gigawatt today, with plans to reach 8 gigawatts by 2030—an extremely significant increase.
India has a unique advantage. It has a large, digitally native population. It generates massive and diverse datasets across languages and use cases, making it an ideal environment for training and testing AI models. Additionally, India has demonstrated its ability to leapfrog technology adoption, as seen with UPI and digital infrastructure.
This is why hyperscalers like Microsoft, Google, and Amazon Web Services have announced significant investments. We have demand, the right conditions, and strong capital inflows. What was built over the last 15 years is effectively being doubled in the next three years.
If you compare the AI infrastructure demand with current infrastructure in India, where do you see the gap and what needs to be done?
Globally and in India, the biggest constraint today is power infrastructure. Three years ago, compute was the bottleneck, but that constraint has largely been addressed. Now, power and energy availability have become the primary challenges.
The second major gap is skills. AI factories require new capabilities—not just to build them at scale, but also to operate them efficiently. These are highly complex environments. For example, earlier, data centres were cooled using traditional air conditioning, which was widely understood. Today, we are moving toward liquid cooling, where cooling systems go down to the chip level, removing heat directly and transferring it to chillers.
These factories run 24/7 with no downtime. This raises new questions around maintenance and operations. We need purpose-built digital operating systems designed specifically for managing such environments. Alongside this, there is a need to invest in both digital infrastructure and the associated skill sets to ensure reliability, efficiency, and scalability.
Can you give us a breakdown of the kinds of skill sets that will be needed over the next few years?
If you look back at the early 2000s, there was a massive shift where engineers from various disciplines were retrained to become software engineers. Companies like Infosys and Tata Consultancy Services built large training ecosystems and worked closely with universities to align curricula with industry needs.
A similar transformation is likely to happen again.
Going forward, there will be a strong need for expertise in electrical architecture, cooling technologies, and precision engineering. As Jensen Huang has pointed out, there will be increased demand for electricians and plumbers—reflecting the importance of core engineering skills.
We will need capabilities in managing power systems, grids, transmission, and distribution networks. AI factories will place significant demand on electrical infrastructure and also feed back into the grid. Managing this dynamic, along with integrating storage systems, will be critical.
In addition to engineering skills, digital, operational, and system-level capabilities will also be essential. Overall, a combination of electrical, mechanical, power systems, cooling, and digital expertise will define the workforce of the future.
Do you see a reversal in the trend from IT back to core engineering fields?
I would not describe it as a reversal, but foundational engineering skills are becoming significantly more important again. This creates an opportunity, especially for a country like India, which has historically been more focused on software and application development.
At the same time, software will continue to play a critical role. With the rise of agentic AI, there is significant potential to improve efficiency across industries.
For instance, in buildings, systems are often programmed in a one-dimensional way, which can lead to inefficiencies such as overcooling. With additional sensors and AI-driven reasoning, systems can dynamically optimise performance.
Similarly, in healthcare—a sector that is often overstretched—AI can assist by automating information capture and analysis. Nurses and doctors can interact with systems more naturally, enabling faster decision-making and improving accessibility and affordability of care. The broader point is that both engineering and software capabilities will evolve together, creating new opportunities across sectors.
Where do you see opportunities for agentic AI in the data centre industry?
We are already deploying agentic AI in multiple use cases within data centres. AI factories consist of mechanical, electrical, and cooling systems that generate vast amounts of data. We bring all of this into a unified data layer and then apply agentic AI.
Traditionally, operators had to deal with an overwhelming number of alarms and data points, making it difficult to prioritise actions. Agentic AI simplifies this by identifying what truly matters and where human intervention is required. Another key use case is lifecycle maintenance. These assets are expected to run continuously for 20–25 years. Using AI, we can analyse system data to predict potential failures—similar to monitoring health indicators—and take preventive action. This helps avoid downtime, reduces the need for on-site intervention, and enables remote operations.
Do you see this as a service opportunity?
Yes, and we are already delivering such services. For example, a large MNC has globally connected facilities that are monitored and managed through a central command centre in India. These systems provide real-time visibility into energy usage, renewable integration, storage, and operational alerts. This level of digital infrastructure management is becoming a baseline requirement. Organisations that do not adopt such models risk becoming less competitive.
Do you see this becoming a large opportunity similar to IT services?
We are already managing infrastructure remotely through operating centres in India. These centres oversee data centres and buildings, ensuring uptime, reliability, and efficiency. Through digital platforms, we continuously optimise energy consumption, asset life, and operational performance. While the opportunity is significant, it may not reach the scale of the IT services industry, which spans multiple sectors. However, the model itself is already proven and expanding.
What are some of the key trends you are most optimistic about?
One major trend is the shift toward modular construction. There is a race to build infrastructure quickly, and skilled labour is a constraint. The approach now is to break down AI factories into modular components. Power systems, including distribution and UPS, are pre-integrated into containerised units. Similarly, compute and cooling systems are packaged into deployable modules.
These are manufactured, tested, and validated in factories before being shipped and installed on-site. This reduces risk, accelerates deployment, and enables scalability. In some cases, dedicated manufacturing facilities are being set up to support large deployment pipelines.
Compared to traditional on-site construction, prefab models can reduce timelines by months. They also significantly lower execution risk, while still allowing for customisation based on client requirements.
What advice would you give at a policy level for India?
Energy and power infrastructure will be critical to competitiveness. Rather than focusing only on large, centralised data centres, there is an opportunity to develop distributed and edge data centres. This approach can reduce latency, distribute load, and bring compute closer to where data is generated and consumed. States should prioritise energy availability, infrastructure development, and data centre ecosystems. This will attract investment and drive economic growth. Even smaller states can leverage this as a strategic opportunity to accelerate development.