(The Weekend Insight) - The Startup Market Around India’s Data Centers

In today’s deep dive, we will look at the startup market forming around India’s data centers. While attention usually goes to operators, hyperscalers and capacity announcements, another set of companies is working on cooling, backup power, fire safety, fiber, security, water management and maintenance. As India builds more data centers, these support businesses could become an important infrastructure market.

India already has around 1.6 GW of operational data-centre capacity, making it the second-largest market in Asia-Pacific by operational capacity. Another 3.1 GW is in the development pipeline. Global capacity could almost double to around 200 GW by 2030.

Most of the attention will go to operators such as NTT, STT GDC, Nxtra, CtrlS, Yotta, AdaniConneX and Sify. They announce campuses, buy land and sign deals with cloud companies.

The supplier market receives far less attention.

That may change as AI facilities become harder to build and operate. Data centres are moving from relatively predictable server rooms towards dense industrial campuses that consume large amounts of electricity and produce enormous heat. The International Energy Agency expects global data-centre electricity consumption to more than double to around 945 TWh by 2030.

Every increase in computing demand eventually becomes a physical requirement somewhere else. More processors need more electricity. More electricity creates more heat. More heat needs additional cooling. Higher rack density requires stronger power distribution, better sensors, new fire systems and faster fiber links.

The data-centre economy therefore extends far beyond servers.

A data centre is closer to a factory than an office building

The easiest way to misunderstand a data centre is to think of it as a building filled with computers.

The computers are only the productive equipment. The rest of the facility exists to make sure that equipment receives uninterrupted electricity, remains within a narrow temperature range, stays connected and survives equipment failures.

A large data centre contains substations, transformers, switchgear, uninterruptible power supplies, battery banks, diesel generators, chillers, pumps, cooling towers, fire systems, security checkpoints, cable rooms and thousands of sensors. Teams monitor these systems every hour of the day.

Much of this infrastructure is duplicated. If one system fails, another must take over. The customer is paying for computing capacity, but the operator is spending heavily on redundancy.

This creates a different kind of supplier market.

A food-delivery company can experiment with a new software provider and replace it if the product fails. A data-centre operator cannot casually test an unproven UPS, cooling unit or fire-suppression system inside a live facility. Failure could affect hundreds of customers at the same time.

Suppliers must pass technical evaluations, receive certifications, complete pilots and prove that spare parts and service engineers will remain available for years. Approved-vendor status can take a long time to obtain.

Once a supplier enters, however, removal is also difficult. The operator becomes dependent on the company’s equipment, software, spare parts and technicians.

This is why companies such as Schneider Electric, Vertiv, Eaton and Munters have built large data-centre businesses globally. They sell physical equipment, but their strength comes from the installed base, maintenance contracts, replacement demand and customer trust. Schneider now covers power, cooling, security and management software within the same data-centre offering. Vertiv supplies UPS systems, cooling and integrated infrastructure, while Munters has been expanding manufacturing capacity for data-centre cooling equipment.

The Indian startup opportunity follows the same logic. The better businesses will not rely entirely on fresh construction. They will earn from equipment already deployed.

Cooling is becoming the first major bottleneck

Traditional data centres were mainly cooled by circulating cold air around server racks. Operators separated hot and cold aisles, installed computer-room air-conditioning systems and used chillers to remove heat from the building.

That model becomes less effective as rack density increases.

AI servers place several power-hungry processors inside the same rack. The heat is concentrated in a smaller area, and moving larger volumes of air through that space eventually becomes inefficient. ASHRAE’s latest AI data-centre guidance discusses liquid-cooling systems for racks operating at 50 kW to more than 100 kW.

Liquid cooling moves water or another fluid close to the processor. A cold plate absorbs heat directly from the chip. The heated fluid then passes through a cooling distribution unit, pumps and heat exchangers before returning to the server.

This changes the supplier list.

The facility now needs cold plates, manifolds, pumps, connectors, cooling distribution units, leak sensors, filtration, specialised fluids and technicians who understand both computing equipment and industrial cooling.

KühlTherm is one Indian company working on direct-to-chip liquid cooling. Its system uses a patented Mesh Cold Plate designed around the heat profile of CPUs and GPUs. The company is also working around cooling distribution and high-density rack requirements.

Uravu Labs is approaching the problem from another direction. Its technology uses reject heat from the data centre to extract water from atmospheric moisture. The company is pitching a combination of heat reuse, water generation and lower cooling costs.

The cooling market will not belong only to companies manufacturing complete systems. Many startups can enter through smaller components.

One company may specialise in leak detection. Another may monitor coolant quality. A third may help retrofit old facilities for liquid-cooled racks. Other businesses can inspect pipes, clean heat exchangers, recycle cooling fluids or predict pump failures.

Retrofitting may become particularly important.

A facility designed for ordinary cloud workloads cannot automatically accommodate dense GPU racks. It may lack adequate cooling pipes, water capacity, power distribution or floor strength. Upgrading the building while customers continue using it is a difficult engineering job.

The opportunity therefore includes installation, inspection and maintenance alongside manufacturing.

Cooling also shows why this market should not be analysed like software. A startup may have excellent cold-plate technology, but it still needs manufacturing quality, warranties, inventory and field technicians. One leak inside an expensive AI rack can damage the company’s reputation before it has built a meaningful installed base.

Technical performance gets the startup into the room. Reliable execution keeps it there.

Backup power is turning into an energy business

A data centre cannot wait for the electricity grid to recover after an outage.

UPS systems and batteries take over immediately. Generators start soon afterwards and can continue supplying the facility for a longer period. Operators maintain several layers of backup because even a brief power interruption can affect customers.

Diesel generators remain common because they are proven and can run for long periods when fuel is available. The problem is that they are noisy, polluting and expensive to operate. As data-centre campuses grow, the size of the required backup fleet also increases.

Indian startups are entering through battery systems and alternative generation.

Cygni Energy manufactures lithium-ion storage products and lists data-centre UPS systems as one of its business segments. The company has built a 4.8 GWh manufacturing facility and also works across commercial storage and large battery systems.

Offgrid Energy Labs is developing ZincGel, a zinc-bromide battery designed to store electricity for six to sixteen hours. The company says the system uses a water-based electrolyte and is non-flammable by design. Data centres are one of its stated applications.

Other Indian companies are exploring hydrogen generators and fuel cells. These technologies are still at an earlier stage, but the customer problem is clear. Operators want to reduce their dependence on diesel without compromising backup duration.

Batteries may eventually perform more than one job.

Today, a UPS battery largely sits idle and waits for an emergency. In the future, the same storage system could reduce peak electricity demand, store renewable power and provide services to the grid. Eaton, for example, markets grid-interactive UPS systems that allow data centres to use battery capacity for energy management while preserving backup requirements.

This could create several business models.

A battery company can sell equipment. It can lease batteries to the operator. It can charge a monthly fee for backup capacity. It can monitor battery health and guarantee availability. It can also replace degraded modules and recover material from old packs.

The strongest company may separate the technology from the financing.

Venture capital can fund battery development, software and customer acquisition. Long-term debt or infrastructure capital is more suitable for battery assets deployed at customer sites. Using expensive equity to own every battery will eventually weaken the balance sheet.

The same lesson is already visible in EV battery-swapping businesses. Technology and network operations may earn equity returns. Large pools of standardised batteries need lower-cost capital.

Maintenance software has to understand physical equipment

A data centre may have thousands of maintenance activities scheduled every year.

Generators must be tested. Fire cylinders need inspection. Cooling units require servicing. Batteries must be checked for degradation. Pumps, fans and motors produce signals that may indicate failure.

Much of this work is still managed through separate systems, spreadsheets, vendor reports and manual checklists.

Data-centre infrastructure management software usually focuses on racks, power, cooling capacity and IT equipment. Building-management systems control physical machinery. Maintenance software manages technicians and work orders. Vendor records and audit documents may sit somewhere else.

The startup opportunity is to connect these layers without forcing the operator to replace every existing system.

Facilio is one of the clearest Indian-origin examples. Its data-centre product manages preventive maintenance, vendor service levels, compliance records and energy monitoring across assets such as HVAC systems, generators, UPS units and fire equipment. The platform is designed to work alongside existing building-management and data-centre infrastructure systems.

Companies such as UptimeAI, Infinite Uptime and Detect Technologies have built predictive-maintenance products for industrial equipment. Their original customers are often factories, power plants and process industries, but the underlying technology can also be applied to pumps, cooling equipment, motors and generators inside data centres.

The valuable product is not another dashboard.

Operators already have dashboards. They need software that identifies a problem, explains the likely cause, creates a work order and records whether the repair was completed.

A temperature alert without context creates more work. A useful system would show that one cooling unit is drawing more electricity, producing less cooling and following the same pattern that preceded an earlier compressor failure.

That requires equipment data, maintenance history and repeated deployments. A young startup will rarely have enough data at the beginning. It needs customers to improve the product, while customers want evidence before allowing the product near critical equipment.

This makes partnerships important. A predictive-maintenance company may enter through an equipment manufacturer, facility-management contractor or insurance provider rather than selling directly to the operator.

The recurring revenue can be attractive, but only after the startup completes the slow work of integration.

Fire safety remains an incumbent-controlled market

Fire protection should be an obvious startup category. Data centres contain high electrical loads, batteries, cables, diesel fuel and equipment that must remain operational throughout the year.

Yet very few Indian startups specialise in data-centre fire suppression.

Most facilities still depend on established fire-engineering companies, multinational equipment suppliers and local installation contractors. Operators prefer systems with a long safety record because the cost of an unproven product is too high.

NFire is one of the few Indian companies introducing a connected layer. Its systems combine wireless smoke, heat and flame sensors with cloud monitoring and remote alerts. The company also markets predictive risk detection and integration with suppression systems for data-centre environments.

This is more likely to be the startup entry point.

A new company does not need to invent a fire-suppression gas immediately. It can monitor the health of existing systems, test whether cylinders remain pressurised, detect overheating cables or identify early thermal runaway inside battery rooms.

Battery fires will require particular attention. A conventional smoke detector may react only after the event has progressed. Sensors that monitor temperature, gases, electrical behaviour and battery-management data could identify risk earlier.

There is also a compliance problem.

Fire systems undergo inspections, tests and component replacements. Records may remain scattered across contractors and facility teams. A software company could create a live compliance record for every detector, cylinder, pipe and inspection.

The challenge is liability. A startup that claims to predict fires will be judged harshly when it misses one. Certification, insurance and careful product claims will matter as much as the technology.

This category may produce fewer venture-backed companies, but it can support profitable engineering businesses with monitoring and maintenance revenue.

Security extends from the gate to the fiber cable

Data-centre security begins before anyone reaches a server.

The outer perimeter needs cameras, intrusion detection and vehicle controls. Building entry may involve identity checks, biometrics and mantraps. Inside the facility, access becomes more restricted at every stage. A customer may have a private cage, while individual racks may require separate authorisation.

Smart-i Systems provides rack-level access control and environmental monitoring. Its rack units track temperature, humidity, smoke and water leakage and can send remote alerts.

Spintly’s wireless access-control systems and Staqu’s video-analytics products could also be used across data-centre campuses, although data centres are part of a wider customer market for both companies.

The security requirement does not stop at physical access.

Data moves between data centres through fiber networks. Operators must protect customer records, network traffic and administrative access. QNu Labs is developing quantum-security products for data-centre and telecom networks, including quantum key distribution and secure data-centre interconnection.

India’s regulatory environment strengthens this demand. CERT-In’s cybersecurity directions explicitly cover data centres, cloud providers and virtual private server providers. The Digital Personal Data Protection Rules, notified in 2025 with phased implementation, will also increase pressure on companies to maintain appropriate security and incident-management controls.

The more interesting product may combine physical and digital events.

Suppose a maintenance engineer opens a rack, changes a network component and a security alert appears shortly afterwards. Most facilities will record these events in separate systems. A combined platform could connect access logs, camera footage, maintenance tickets and network activity.

That is harder to build than a standalone access-control product, but it creates a deeper position inside the customer’s operations.

Fiber companies are building the roads between data centres

A data centre without reliable fiber is simply a building containing expensive machines.

Facilities need several network routes because a single cable cut should not isolate customers. They connect to telecom carriers, internet exchanges, cloud platforms and other data centres.

Lightstorm has built one of the more relevant Indian businesses in this layer. Its fiber network was connecting around 60 Indian data centres across roughly 24,000 route kilometres by 2024. Its Polarin platform allows customers to provision data-centre and cloud connectivity through a Network-as-a-Service model.

CloudExtel, Astrome and iBUS Networks address other parts of the connectivity market. CloudExtel builds shared fiber infrastructure, while Astrome develops wireless systems that can provide high-capacity links where physical fiber is difficult to deploy.

Fiber businesses can earn recurring revenue because customers pay for capacity over long periods. But they are also capital intensive. Routes must be built before every strand is fully occupied.

Inside the data centre, another problem appears.

Thousands of cables connect racks, switches, patch panels and meet-me rooms. Technicians must know where every cable starts, where it ends and whether it is still in use. Poor records lead to abandoned cables, incorrect connections and slower repairs.

Lightstorm has also described using remote robotics to automate physical network connections within data centres. This offers a glimpse of where fiber management may go.

There is still room for Indian startups to build cable-mapping software, automated tracing, optical-loss monitoring and digital records of every physical connection.

The customer will not buy these products because the interface looks better. It will buy them if they reduce installation time, prevent mistakes and allow the same team to manage more connections.

The rack is becoming a measurable business unit

Operators often sell capacity through racks, cages and power commitments.

Yet many facilities still lack a complete live view of each rack. They may know how much power was allocated but not whether the rack has enough cooling headroom, whether the rear temperature is rising or whether unused equipment is consuming electricity.

Rack-monitoring systems combine smart power-distribution units, temperature sensors, humidity sensors, door controls, smoke detectors and water-leak sensors.

Smart-i Systems already offers this type of hardware in India. Companies such as AuroEDGE package monitoring with micro data centres, while software providers can collect information from multiple brands of sensors and equipment.

The larger opportunity is to connect operational information with commercial information.

A rack-intelligence platform could show available power, cooling capacity, customer ownership, maintenance history, door access, equipment age and billing status. The operator could then see which racks are underused, which ones are approaching thermal limits and where additional capacity can be sold.

Global suppliers are already combining these categories. Schneider offers integrated row systems with racks, power, cooling and management software. Vertiv packages power, cooling, monitoring and racks for edge and micro data centres.

Indian startups will need to decide whether to compete by manufacturing the rack, supplying the sensors or owning the software layer above different hardware brands.

Manufacturing alone may become a price-driven business. Software alone may struggle without access to reliable sensor data. Combining hardware, software and installation creates a stronger product, but it also creates inventory and service costs.

Water could decide which campuses receive permission to grow

Data centres use water mainly through cooling systems. The exact requirement depends on the cooling design, climate, workload and source of electricity.

In a water-stressed Indian city, however, even an efficient facility can face opposition. Residents may not accept large industrial campuses drawing potable water while nearby areas depend on tankers.

This creates demand for water accounting, leak detection, wastewater reuse and alternative cooling.

DigitalPaani provides software for operating water and wastewater-treatment plants. Its system guides plant staff, tracks performance and helps facilities increase water recovery and reuse.

FluxGen uses sensors and analytics to identify leaks, abnormal consumption and inefficient water use. Similar products from Kritsnam, WEGoT and SmartTerra can help operators measure water across cooling towers, treatment plants and individual facility zones.

Indra Water works on decentralised wastewater treatment, while Uravu’s heat-reuse approach attempts to combine cooling and atmospheric water generation.

These companies are mostly adjacent suppliers today. Data centres are not always their largest customer segment. But the product fit is strong.

Water management could become a condition for approval rather than a voluntary sustainability purchase. Operators may need to show where water comes from, how much is reused and how consumption changes during summer.

That creates a more defensible customer need than a generic environmental dashboard.

A startup could monitor every water inlet and outlet, calculate Water Usage Effectiveness, identify leaks and prepare reports for regulators, customers and local communities. It could also operate the treatment plant and charge for each litre recovered.

The difficult part will be proving savings. Water systems vary greatly between facilities. The company must install accurate meters, maintain them and separate genuine efficiency gains from changes in workload or weather.

Once proven, however, the service can become recurring and difficult to remove.

The phrase “picks-and-shovels” can be misleading

It is tempting to place every data-centre supplier inside one attractive market. That would be a mistake.

A cooling manufacturer, fiber operator, maintenance-software company and fire contractor have very different economics.

The cooling company needs factories, testing and warranties. The fiber operator needs debt and long-term capacity contracts. The software company needs integrations and recurring subscriptions. The fire contractor may earn through projects and annual maintenance.

Some of these can become venture-scale technology companies. Others are better suited to private equity, infrastructure funds, industrial capital or profitable family ownership.

The venture-friendly categories are usually those where the company can sell the same intellectual property across many facilities. Predictive-maintenance software, network automation, security products, cooling controls and novel battery chemistry fit this description.

Hardware businesses can also attract venture capital when they own difficult technology and have export potential. KühlTherm’s cold plates, Offgrid’s battery chemistry and QNu’s quantum-security products fall closer to this category.

Standard equipment distribution is different. Importing a sensor, adding a dashboard and installing it at customer sites does not automatically create a technology company.

Investors should examine five things.

Does the company own the product? Can it earn revenue after installation? Does each new deployment make the product or service better? Can it enter another facility without redesigning everything? Can it finance growth without constantly raising equity for inventory and customer projects?

These questions separate a repeatable infrastructure company from a contractor wearing startup language.

India’s supplier opportunity will depend on service, not announcements

India will continue importing several important data-centre components. Global suppliers have decades of reliability data, approved products and service networks.

Indian companies do not need to replace all of them.

They can begin with local conditions that international products handle poorly. Indian facilities operate in high heat, humidity, dust, unstable grid conditions and regions with water stress. Imported equipment designed for another climate may require changes.

Startups can also move faster in retrofits, local servicing and software integration. A multinational may supply the chiller. An Indian company can monitor it, predict failures, optimise its water use and provide technicians within a few hours.

Over time, the company can move deeper into components and manufacturing.

The strongest businesses will probably combine three capabilities: proprietary technology, field execution and recurring service.

Technology without service will fail during deployment. Service without technology will remain a contractor business. Hardware without repeat revenue will struggle with working capital.

Data-centre operators will not select suppliers because they use AI in their pitch deck. They will ask whether the equipment has failed before, how quickly a technician can reach the site, whether spare parts are available and who takes responsibility when something goes wrong.

That makes this a difficult startup market. It also makes successful suppliers unusually sticky.

India’s data-centre boom may begin with land, servers and cloud contracts. The longer business will sit in cooling those servers, powering them, connecting them and keeping them alive year after year.

Construction creates capacity. Maintenance turns that capacity into reliable infrastructure.