Hyperscale Data centres: Scaling for the AI and Cloud Era
Hyperscale data centres are large facilities that support cloud and AI workloads at scale. They are designed to handle growing demand with infrastructure that can expand as requirements increase.
These facilities are still often discussed in terms of capacity and expansion, but they do not fully explain how these facilities scale today.
What has changed is not only the volume of demand, but also how that demand appears and what it requires from the underlying data centre infrastructure.
Changes are now visible in how facilities are designed, how hyperscale expansion strategies are planned, and why certain locations continue to attract more capacity than others.
Scaling Now depends on Decisions Made Much Earlier
Earlier designs assumed gradual growth, with capacity added over time. That assumption doesn’t hold in the same way now.
Now, in environments supporting AI infrastructure, demand often arrives at scale and runs for longer periods, at densities that existing systems were not originally designed to support.
At the start, the difference is not always visible.
A data centre may go live and operate as expected, with no clear indication of its limits. It becomes apparent later, once workloads increase and systems begin operating closer to capacity.
Power capacity that appeared sufficient starts to come under strain, and cooling systems are required to run more consistently.
Once a facility is operational, making changes to the underlying infrastructure becomes more difficult.
For this reason, hyperscale expansion strategies are evolving.
Expansion is no longer only about adding capacity. It also depends on whether the existing setup can support additional demand without introducing constraints.
Facilities, like the data centres at STT GDC India, handle increasing workload density and sustained demand more effectively and often have flexibility built into their initial design, including additional power headroom, scalable cooling systems, and infrastructure that can be expanded without major changes.
How Mumbai Fits into India’s Hyperscale Expansion
Capacity is growing across multiple cities in India, including Bengaluru, Chennai, and Hyderabad, where data centre development has been expanding steadily.
Data centres in Mumbai play a different role within this landscape, attracting strong interest from hyperscale operators due to their connectivity and the type of demand they support.
A large number of international subsea cables land in Mumbai, making it a key point for data entering and leaving the country. This supports lower latency for global workloads and enables applications that depend on stable connectivity.
Demand is also supported by the presence of financial institutions, many of whom are actively adopting AI infrastructure for use cases such as fraud detection, risk modelling, and real-time operations.
As a result, the need for reliable and scalable infrastructure is already established.
STT GDC India has expanded its presence in Mumbai over time, including facilities in Navi Mumbai designed for higher-density workloads. The Prabhadevi site is located close to cable landing infrastructure, aligning with connectivity requirements identified early in the planning process.
Many of these decisions were made before current demand levels became fully visible.
Why Capacity Alone Doesn’t Explain Scaling
Once a facility operates under sustained load, differences in performance become clearer.
Capacity figures, such as megawatt availability, provide a useful reference point. They offer a way to compare scale across facilities. At the same time, they do not fully reflect how a data centre performs when systems are running continuously.
Workloads driven by AI infrastructure do not follow cyclical patterns. Systems are required to operate consistently, often close to their limits.
Over time, there is sustained pressure on power and cooling systems.
Cooling is often one of the first areas where the effect becomes visible. Many operators are moving toward liquid cooling to support higher-density workloads. Outcomes depend not only on the technologies used, but also on how early these requirements were considered.
A similar approach can be seen in how some operators are building.
STT GDC India has been developing liquid cooling capabilities alongside expanding capacity across key markets. These are typically decisions made well before demand becomes visible.
What Scaling Really Looks Like Now
When infrastructure design, workload patterns, and location are considered together, a consistent pattern becomes visible. More than capacity, scaling also depends on when key decisions are made.
Power, cooling, connectivity, and location all matter, but timing often determines how well they hold up as demand increases.
Facilities that scale smoothly are not always expanding the fastest. In many cases, they are the ones where earlier decisions continue to support current requirements. In markets where demand is increasing quickly, this becomes easier to notice.
Some facilities take on additional workloads with minimal adjustment. Others need more time to adapt, often because the underlying infrastructure was not designed for that level of demand.
Scaling, in this context, depends on whether systems are prepared to support expansion without becoming a constraint.