India is preparing to close out 2026 with roughly 2 GW of operational data center capacity, after adding around 130 MW during the first half of the year and with more than 200 MW expected in the second half. The problem is that much of the existing infrastructure was designed for workloads that predate today’s artificial intelligence boom, while new AI systems multiply the power that must be concentrated in each rack.
Key facts about India’s data centers facing the AI boom in 30 seconds
- India could reach 2 GW of operational data center capacity by the end of 2026.
- The country added about 130 MW in the first half of the year and expects more than 200 MW more through December.
- Investment commitments reached $38 billion in the first half of the year, and $173 billion since 2021.
- More than 15% of commitments accumulated since 2021 are linked to AI demand.
- The challenge is shifting from building capacity to preparing power, cooling, and distribution for high-density racks.
The 2 GW figure reflects the rapid growth of India’s digital infrastructure, but it also masks a deeper transformation. Data centers that for years were designed around enterprise workloads, cloud services, and digital content now have to adapt to servers equipped with AI accelerators that concentrate far more power in a much smaller footprint.
A CBRE report published in September puts India’s operational capacity at approximately 1.75 GW at the close of the first half of 2026, after adding about 130 MW over the first six months. The consultancy expects more than 200 MW to come online during the second half of the year, which would bring the market close to 2 GW by the end of December.
The growth is accompanied by a huge wave of investment. Commitments accumulated since 2021 reached approximately $173 billion by the end of June, including about $38 billion announced during the first half of 2026 alone. Hyperscaler and colocation operator projects account for close to 62% of accumulated commitments.
Artificial intelligence is already behind a significant share of that money. More than 15% of the commitments recorded from 2021 through the first half of 2026 are linked to AI-related demand.
The problem is no longer just how many megawatts India can add
India is building capacity at a fast pace, but the expansion of AI is changing how those megawatts need to be used.
A traditional data center can spread its load across many racks at relatively moderate densities. Systems built to train or run AI models, by contrast, concentrate large numbers of accelerators into just a few cabinets. That changes power delivery, cooling, physical distribution, and redundancy requirements.
In India, the difference is already visible. The sector is moving from racks of around 100 kW toward configurations that could potentially reach 1 MW per rack in certain AI-oriented facilities, according to statements made by Vertiv in February. That evolution doesn’t mean every rack in the country will reach those figures, but it does show how much the range of densities that new facilities must support is expanding.
An analysis by the Alliance for an Energy Efficient Economy points to an even bigger gap between conventional infrastructure and AI workloads: traditional racks can run at average densities of 4 to 8 kW, while certain AI workloads exceed 120 kW per rack. That jump forces a rethink of cooling, especially because conventional air-cooling systems lose their ability to remove such large amounts of heat.
The consequence is that a data center with available physical space doesn’t necessarily have the capacity to install just any AI server. It may have enough floor space, network connections, and total electrical capacity, but lack the power distribution or cooling needed to concentrate 100 kW or more in a single cabinet.
That turns retrofitting existing facilities into one of the key technical challenges of this new phase.
Mumbai still dominates, but India is starting to diversify
Mumbai remains the country’s leading data center hub. The city accounted for 94% of the new supply added during the first half of 2026, according to data published by CBRE.
Mumbai’s weight has a lot to do with telecommunications infrastructure. The city has international connections, submarine cable landing stations, and a developed financial and business ecosystem.
However, growth is starting to spread to other markets. Chennai, Delhi-NCR, Hyderabad, Bengaluru, Kolkata, and Visakhapatnam are all part of the expansion planned by operators.
CBRE notes that more than 80% of India’s operational capacity remains concentrated in Mumbai, Chennai, and Delhi-NCR. The next growth phase should increase the weight of secondary markets, especially where power availability, connectivity, and suitable conditions exist for building large facilities.
Site selection itself is changing. Access to power and the ability to secure renewable electricity carry increasing weight alongside network connectivity. In its analysis of the Asian market, CBRE identifies power availability, renewables, and expansion capacity as increasingly decisive factors in locating new data centers.
For AI, this matters especially because electrical demand doesn’t simply grow in step with built square footage. Server density determines how much power must reach each area of the building and how much heat has to be removed.
43% of supply is already Tier IV
The technological shift also shows up in the availability and redundancy level of facilities.
According to CBRE, Tier IV data centers accounted for 43% of total supply in the first half of 2026, up from 21% in 2024. The growth shows how operators are increasing the sophistication of their facilities as demand for mission-critical services rises.
Project structure is also changing. More than 85% of the commitments made during the first half of the year belonged to portfolios with a presence across multiple cities.
Diversification has a clear advantage: it reduces dependence on a single market and allows loads to be spread across different facilities. But it also complicates power management, networking, cooling, and system operations.
Domestic investment is also carrying significant weight. Indian investors accounted for 38% of capital accumulated since 2021, ahead of the United States at 27%. In terms of new supply, domestic developers contributed 48% during the first half of 2026, compared with 28% from U.S. operators and 24% from European ones.
India is no longer just a market where foreign companies build capacity. Local operators and developers are taking on a significant role in expanding the infrastructure.
AI is also changing cooling
The power concentrated in AI racks has an immediate physical consequence: nearly all the electricity consumed by the servers ends up turning into heat that must be removed from the building.
With conventional workloads, air cooling can handle most of that task. Once density rises substantially, thermal and airflow capacity limits appear. That’s why new AI-ready data centers are adopting liquid cooling systems, heat exchangers, and circuits capable of capturing heat directly from the components.
Not every facility needs the same system. The right solution depends on power per rack, accelerator type, server design, and data center architecture. But rising density is shifting liquid cooling from a technology tied to certain high-performance facilities into a capability operators need to consider for any new AI project.
In India, this retrofitting is especially relevant because many existing facilities were designed before today’s accelerators reached these densities.
The problem doesn’t necessarily mean replacing an entire data center. Some facilities can add liquid cooling circuits, change their power distribution, or set aside certain rooms for high-density workloads. But the capacity to adapt has limits, and it could determine which facilities can host the next generations of AI systems.
India wants to turn growth into long-term infrastructure
India’s market has already entered a new category within Asia-Pacific. CBRE elevated India in 2026 to the status of a Leading Data Centre Market, alongside Japan, Australia, and mainland China. The consultancy ties this shift to the scale the market has reached, the growth in investment, and the expansion of infrastructure.
The country also ranks among the regional markets best positioned to benefit from AI growth. The combination of population, digitalization, cloud services, and growing hyperscaler activity creates structural demand for capacity.
But the next stage will be more demanding than the last. Building 2 GW of data centers does not guarantee having 2 GW suited to the densest AI workloads.
The difference will lie in how that power is distributed and what portion of the infrastructure can support racks drawing tens or hundreds of kilowatts. It will also be necessary to secure sufficient network capacity, cooling, physical space, connectivity, and electricity supply with continuity guarantees.
The issue matters especially because investment commitments are already running ahead of demand. CBRE estimates that the operators it surveyed plan to keep expanding capacity aggressively, with 92% of survey respondents expecting to add more than 100 MW over the next 24 months.
India is therefore building an ever-larger data center market at the same time that it changes the type of hardware it will need to house. The 2 GW projected for the end of 2026 will be a benchmark of scale, but the real test will come as a growing share of that capacity has to power high-density AI clusters.
The country can keep adding megawatts, but its infrastructure will have to learn to turn them into usable power exactly where the accelerators need it. That gap between installed capacity and capacity genuinely ready for AI will be one of the factors shaping the next phase of the Indian market. The same density pressures are already reshaping rack design elsewhere in the industry, including in high-performance networking gear such as Arista’s 1.6 Tbps Ethernet platforms built for rack-scale AI networks.
Frequently asked questions
How much data center capacity will India have by the end of 2026?
CBRE expects India to reach about 2 GW of operational data center capacity by the end of 2026. The country had approximately 1.75 GW at the close of the first half of the year.
How much has India invested in data centers?
Investment commitments accumulated since 2021 reached about $173 billion through the first half of 2026. Roughly $38 billion was announced during the first six months of 2026 alone.
Why does AI require new data centers?
AI accelerators concentrate far more power in certain racks than traditional workloads do. That requires reinforcing power delivery, distribution, cooling, and other parts of the infrastructure.
Which city has the most data centers in India?
Mumbai remains the leading market. It accounted for 94% of the new supply added during the first half of 2026, although expansion is also reaching Chennai, Delhi-NCR, Hyderabad, Bengaluru, Kolkata, and Visakhapatnam.
