India Data Centre Investment Map 2026: Big-Tech Commitments, Operators, Finance and State Incentives

In this guide

1. How to read an investment announcement
2. Selected India investment announcements and what they cover
3. The main types of capital entering the sector
4. From announced commitments to investable project capital
5. State policy targets and incentives are a separate ledger
6. Where capital goes beyond the building
7. How to build a non-double-counted investment map
8. What an investor, operator, vendor or lender should verify
9. India’s regional investment map: follow infrastructure, not just state headlines
10. From announcement to live asset: the investment lifecycle
11. Illustrative capital scenarios: useful for thinking, not forecasts
12. Investment signals for suppliers: what is actionable and what is noise?
13. Common errors in India data-centre investment coverage
14. Suggested editorial method for publishing a credible market map
15. Turn the public map into a decision brief
16. Related Gridvik guides
17. Frequently asked questions
18. Sources and further reading

How to read an investment announcement

“Investment” is one of the most overloaded words in the data-centre market. In one article it may refer to a company’s multi-year India plan; in another it may mean the cost of one campus; elsewhere it may describe a memorandum of understanding, a lending facility, a state government’s ambition, a renewable-energy portfolio or spending across the wider AI ecosystem. Adding these figures together without checking their scope can make the market look larger than it is and can mislead the reader about what has actually been built.

Before writing or using a number, capture at least eight fields:

1. Announcing entity: The company, joint venture, state government, lender or research firm that made the statement.
2. Scope: One facility, one state, a national plan, cloud infrastructure, all business units, or the wider AI ecosystem.
3. Amount type: Investment commitment, estimate, capex budget, debt facility, equity, acquisition value, policy incentive or economic impact estimate.
4. Period: The years over which the amount is expected to be spent or delivered.
5. Project status: Proposal, MoU, signed agreement, financing announced, under construction, commissioned, live or operating.
6. Capacity unit: IT load, utility load, planned campus capacity, phase capacity or a broader platform target.
7. Overlap: Whether the amount is included in a larger parent-company or sector total.
8. Evidence date and link: The exact original statement and the date it was checked.

A commitment is not the same as deployed capital

A commitment can describe a company’s intention to invest over several years. It may be subject to market demand, approvals, financing, customer uptake or project execution. It does not show how much has been spent as of today. A loan announcement is not always the same as a loan drawdown. A facility under construction is not necessarily fully funded. A state’s target is an ambition, not an audited pipeline. Those distinctions are essential when comparing markets or discussing returns.

For example, company announcements often describe a package of cloud, AI, skills, operations, jobs, energy or digital services. A multi-year company-wide pledge may include data-centre infrastructure, but it should not be relabelled as the cost of a single data-centre campus. Use the company’s wording and identify the component that directly relates to data-centre infrastructure, if that component is disclosed.

Capacity is not a capital amount

A gigawatt figure measures power capacity, not how much has been invested. It may refer to a campus plan at full build-out, a platform goal across cities, an initial phase, a lease commitment or a forecast. A project’s site and building may be sized in one way while its IT load, grid draw and power infrastructure are measured in another. Use the unit exactly as the source presents it and state whether the figure is operational or planned.

A project can have a “1 GW” long-term plan and only a fraction of that capacity in its first phase. If a press release says “gigawatt-scale,” do not turn it into a precise 1,000 MW operational number unless the source makes that exact statement and the status supports it. If a report estimates operational stock, do not add its planned pipeline to that number and call the result live capacity.

Why the source date matters

Data-centre markets change quickly. A 2024 platform ambition may have been revised by a later statement; a planned investment can be expanded or re-scoped; a state policy may be replaced by a new policy; a capacity estimate may refer to a different period. An investment map needs a last-checked date and should point to the original company or government page where possible. Secondary reporting can help interpret a deal, but the primary statement should be linked if available.

A credible research note should also preserve uncertainty. “Not publicly disclosed” is a valid data point. If the cost of a lease is not published, do not infer the amount from MW. If a state announces a capacity target, record the target as a target. If an MoU is reported but the binding terms are not public, do not describe it as a closed financing or construction contract.

Selected India investment announcements and what they cover

The table below is a selection of public announcements with different scopes. It is not a sum, market-size calculation or ranking. Amounts come from the named sources and should be refreshed before use in an investment decision.

A sector-wide commitment estimate is not a project-level cash total

CBRE’s India Alternate Sectors Outlook 2026, as summarized in BusinessToday’s coverage, describes approximately US$126 billion in cumulative investment commitments associated with India’s data-centre sector through 2025. Treat this as a source-specific aggregate estimate—not audited cash deployed, not a measure of one operator’s spending and not an amount to add to the individual announcements in the table above. The aggregate may cover overlapping programmes and different types of commitment. Preserve CBRE’s date and scope, and use its methodology before comparing the figure with another report. See the CBRE India market report.

Google: a multi-part AI hub, not a single building budget

Google’s official announcement describes an approximately US$15 billion commitment over five years, from 2026 to 2030, for an AI hub in Visakhapatnam. The scope includes gigawatt-scale data-centre operations, energy and an expanded fibre network. Google’s 28 April 2026 update says the hub broke ground and describes new transmission lines, clean energy generation and storage as part of the broader programme. That is useful evidence of movement from announcement to construction. It does not make all phases operational, disclose every facility’s capex or prove the amount spent to date.

When mapping the announcement, a careful entry would read: “Google AI hub, Visakhapatnam; US$15B company-stated five-year investment across the hub ecosystem; gigawatt-scale compute ambition; ground broken April 2026; phase capacity, facility-level capex and live MW to be verified.” This is longer than a headline but more useful to a lender or vendor.

Microsoft: infrastructure plus skills and operations

Microsoft announced US$17.5 billion over four years, 2026 to 2029, for India’s cloud and AI infrastructure, skilling and ongoing operations. Its announcement described an earlier US$3 billion investment plan in 2025. Microsoft’s India South Central cloud region in Hyderabad became operational in 2026. These are related but distinct facts: one describes a company commitment, another an earlier commitment, and another the launch of a cloud region. The region’s public launch does not reveal the precise IT load, project-level capex or revenue earned.

A spreadsheet should store the company statement and period, mark the scope as broader than a single data-centre campus, and record the region status separately. It should not add the US$3B and US$17.5B again if the user is calculating a later US$20.5B company commitment that already represents both.

AWS and Amazon: avoid overlap across parent and subsidiary announcements

AWS has described a US$12.7 billion plan to invest in India’s cloud infrastructure by 2030. In January 2025, it announced a US$8.3 billion plan for cloud infrastructure in Maharashtra, specifically associated with the Mumbai Region; AWS said this was part of the earlier US$12.7 billion plan. A map that lists both can show the state component and the national total, but it must flag the overlap.

Amazon’s broader India announcements have a different scope. In 2025, Amazon announced a US$35 billion investment across its businesses in India. In June 2026, it described an additional US$13 billion focused on AI and cloud infrastructure and a US$48 billion total across the 2026–2030 period. Those corporate totals span more than the cost of individual data-centre buildings. A researcher should not add Amazon’s broad commitment to AWS’s earlier cloud-infrastructure plan without confirming whether those plans overlap or have been superseded.

Digital Connexion: a project-specific development plan

Digital Connexion is a joint venture of Brookfield, Reliance Industries and Digital Realty. It announced a US$11 billion plan to develop a 1 GW AI-native campus on 400 acres in Visakhapatnam by 2030 under an MoU with Andhra Pradesh. This is more project-specific than a broad corporate India commitment, but it remains a plan with phases and delivery milestones to monitor. An MoU can support policy facilitation and site planning; it is not by itself proof that every financing, permit, customer or power agreement is complete.

Meta and Reliance: capacity and lease, cost not disclosed

Meta and Reliance announced an agreement for Reliance to build an AI-enabled facility in Jamnagar with a 168 MW initial capacity that Meta will lease, with options to scale. Meta described renewable power and desalinated seawater cooling, and said it would cover the full cost of energy and water supporting the facility. The announcement does not give an investment amount. A rigorous investment table should state “amount not disclosed,” record the lease and capacity separately, and wait for a later source before changing the operating status.

The main types of capital entering the sector

A map that only lists dollar announcements misses how projects are financed and where spending goes. A useful market view separates the capital source, the recipient, the use of funds and the evidence status.

1. Corporate investment and internal capex

A hyperscaler or operator may use its balance sheet to fund buildings, power systems, networks, servers, software, staff and operations. The company can allocate capital across multiple sites or business units. It may revise its plan as demand changes. A public announcement gives a signal of strategic intent, but a project-level model needs the actual phase budget, commissioning plan and customer commitments where available.

2. Joint-venture equity

A local infrastructure company, global operator, property group or infrastructure fund can combine capital and capabilities in a joint venture. One partner may contribute land or local project execution; another may contribute operator experience, capital, customer access or technology. The investment map should record the legal entity, owners, role of each partner, site, governance and source. A brand partnership is not always a 50:50 capital arrangement; do not assume equal stakes unless disclosed.

3. Construction and project debt

Lenders may provide construction facilities or project loans tied to a defined portfolio or asset. Debt can finance a phase before operating revenue begins. The borrower may draw the facility in stages after meeting conditions. A press release announcing a financing facility does not always say how much is drawn, what is secured or which project receives every dollar. Record the facility size, borrowers, collateral if public, project scope and close date separately.

4. Acquisition and portfolio transactions

An operator or investor may acquire operating capacity, land, a development platform or a project company. An acquisition can transfer ownership of existing assets and pipeline, but should not be counted as new construction capex unless additional spending is disclosed. A reported portfolio size may combine live, under-construction and planned capacity. Track the acquired asset status and transaction value separately from future buildout.

5. Customer-led build-to-suit and pre-leasing

A large customer may sign a lease or pre-lease capacity at a facility designed for its requirements. A pre-lease can improve a lender’s confidence, but the contract’s exact terms, tenant obligations, start date and termination rights matter. A build-to-suit agreement can tie the customer to a site while the developer carries construction risk. Record the capacity agreed, lease status and delivery conditions; do not infer the rental value if it is not public.

6. Power and renewable-energy capital

A campus may rely on new transmission, substations, renewable generation, storage or backup power. These investments can be owned by the data-centre company, a utility, a power producer or a separate project company. A renewable project’s nameplate capacity is not the same as the DC’s IT load; the contract and delivery point should be recorded. A broader renewable commitment may serve more than one customer or state.

7. State incentives and public support

Policy can reduce eligible project costs through subsidies, duty exemptions, power support, land measures or streamlined approvals. These are not the same as a private investment or an outlay already paid. Eligibility conditions, caps, approvals and reimbursement timelines matter. Record the policy, qualifying project size, location, application process, scheme period and whether an award has actually been approved.

8. Equipment and supply-chain investment

Data-centre construction creates spending on electrical and mechanical systems, servers, storage, networking, racks, cooling, security, fibre and controls. A project developer may buy equipment directly or through EPCs and system integrators. Supplier sales can be an important local economic opportunity, but the aggregate value of equipment is not necessarily disclosed in a public announcement. Do not manufacture an estimate from a campus MW without a credible cost model and assumptions.

9. Operations and service contracts

After a facility opens, recurring spend continues on power, cooling, network, security, maintenance, staff, software and replacement equipment. A project investment table should not stop at groundbreaking: operating costs can affect long-term viability and suppliers’ service revenue. Some companies may earn recurring fees even when no new campus is being built.

From announced commitments to investable project capital

A headline investment commitment can represent the top of a financing story, not the amount available for a particular project today. To understand how capital can become a facility, separate at least six concepts: a strategic commitment, a board-approved budget, a project-level cost estimate, a signed financing facility, a drawdown and a payment for completed work. Public announcements rarely disclose every stage, so the investment map should report only the stage it can verify.

These terms matter because a press release may describe the total programme while a lender evaluates one phase. A company can announce a multi-year commitment before selecting all sites; a project vehicle can secure debt before it draws the entire amount; and an operator can spend on servers separately from the real-estate developer’s building budget. Each number belongs in its own field.

A fictional example of how overlap occurs

Suppose a technology parent announces an India-wide investment programme. A subsidiary later announces a cloud-infrastructure commitment, and a state government publishes an MoU for a local campus that it describes as part of that same programme. The project company then announces construction finance. Those may be four useful records: parent programme, subsidiary plan, state project and loan facility. They are not automatically four independent investments. The state project can be a component of the subsidiary plan; the loan may finance part of the campus budget; and neither the commitment nor the loan necessarily equals cash already spent.

The map should link the records and identify their relationship. If the source does not clarify the overlap, mark it “possible overlap—unresolved” instead of assuming that the numbers are additive. This is especially important when comparisons are repeated in media, policy presentations or investor materials, where an overlap can quietly inflate the total.

Currency, dates and price basis

Investment amounts can be denominated in US dollars or rupees, with exchange rates and dates that differ. A project’s estimated cost may be announced in one year and revised later. If converting currency, state the exchange-rate basis and date; do not combine a current conversion with historic nominal figures without a note. A nominal rupee commitment is not inflation-adjusted project spend. Where exact conversion is not needed, preserve the original currency to avoid false precision.

A project-level capital ledger should show an amount’s period, currency, type, scope, source and overlap flag. A separate “verified deployment” column should remain blank unless the source establishes actual spending or drawdown. This makes a market map useful to finance teams without suggesting that every announcement is already investable or funded.

State policy targets and incentives are a separate ledger

States compete to attract data-centre projects because facilities can support cloud, AI, digital services, employment and infrastructure. A state policy may offer an investment framework, incentives, land support, access to utilities, renewable requirements and a single-window process. These measures can improve a project’s business case, but they should be documented separately from private capex and live capacity.

Gujarat: 7.5 GW policy target, with conditions

The Viksit Gujarat Data Centre Policy 2026–29 sets a stated target of 7.5 GW of data-centre capacity and a public investment ambition of ₹6 lakh crore. NASSCOM’s policy analysis describes a 51% green-energy requirement and notes that certain incentives apply to new projects with a minimum approved IT load of 150 MW. The exact benefits, caps, eligibility and implementation guidance should be checked against the official policy document before a project uses them in its model.

A target of 7.5 GW is a policy objective, not 7.5 GW of operational data centres. A government may receive proposals or expressions of interest that exceed the target; that still does not mean each proposal is financed or approved. The researcher should separate the state’s ambition, investor proposals, formal approvals, active construction and live capacity.

Uttar Pradesh: a 2026 policy with a separate investment target

Invest UP’s 2026 policy page lists a target to mobilise more than ₹2 lakh crore in investment and sets out measures for data-centre parks and units. It describes support including interest and capital subsidies, dual-grid power support for eligible parks, land subsidy, stamp-duty and electricity-duty provisions, and renewable-energy adoption objectives. The source’s exact scheme rules, definitions, caps, notifications and application requirements matter. Older 2021/2022 policy provisions should not be mixed with the 2026 framework without checking which policy applies to the specific project.

How to compare incentives responsibly

A comparison table should include the recipient, minimum project size, eligible capex definition, maximum subsidy, duration, location restrictions, power and water requirements, application process, disbursement timing and source. A headline percentage is not the same as a project’s net savings. Incentives may be capped, reimbursed after commissioning, limited to specific equipment or subject to conditions. A company should run the base case both with and without the incentive so it knows whether the project still works if the benefit is delayed or unavailable.

Policy is only one part of site selection. Reliable power, transmission, water, fibre, land, customers, construction capability and talent can matter more than a nominal incentive. A state can announce a generous package but still require a project to solve its own node-level constraints. Conversely, a lower-incentive market may offer a mature ecosystem and faster execution.

Where capital goes beyond the building

A data-centre campus is the visible part of a broader infrastructure system. A complete investment map considers at least six connected layers.

The categories can have different owners and funding sources. A utility may invest in transmission; a power developer may own a renewable project; an operator may build the facility; a hyperscaler may buy servers; a network carrier may finance fibre; a local authority may provide land or policy support. Adding every associated commitment into one “data-centre capex” total can overstate direct spending and obscure the allocation of risk.

Capital expenditure versus economic impact

A company may publish an estimate of GDP contribution, job support or supply-chain activity related to an investment. That is an economic-impact estimate, not a project’s cash outlay and not a return to investors. It may include indirect effects across suppliers. A research article should label the metric exactly—e.g. “estimated GDP contribution” or “annual average jobs supported”—and identify who produced the estimate.

Similarly, a company’s total India investment may include employee compensation, technology operations, logistics, fulfilment, digital payments, cloud infrastructure and other business lines. If the reader’s question is “How much data-centre capex is being built?”, the broad corporate total cannot answer it without further segmentation.

How to build a non-double-counted investment map

A well-structured investment map is more useful than a large total. The recommended record contains:

Keep parent and subsidiary plans linked

If AWS announces a state component within an existing national plan, store both entries if the map needs state detail, but add an overlap relationship. If Amazon announces a broad investment and later an AI/cloud component, record the latter as a sub-scope or later refinement if the company says so. Do not sum the parent and subsidiary commitments automatically. A reviewer should be able to click from the sub-programme to the wider announcement and understand why the amounts are not additive.

Keep status fields separate from amount fields

A project’s status can change without changing the announcement amount. A financing announcement can occur before construction; a construction milestone may happen before the first operational MW; one phase can go live while later phases remain planned. The record should have separate fields for announced date, construction status, operational date, planned capacity and live capacity. Do not make one “status” field try to represent the whole campus if phases differ.

Include unknowns and conflicts

If a project’s expected live date differs across sources, show both dates with attribution and note which one is more recent or primary. If the investment amount is undisclosed, say so. If an operator does not publish IT capacity, leave it blank rather than deriving it from building size. A transparent map can be incomplete and still be valuable; a falsely precise map is difficult to trust.

A simple status taxonomy

• Announced: A company or government has publicly stated an intention or plan.
• MoU / proposal: A memorandum or proposal exists; the exact legal and financing status may remain unknown.
• Financing announced: A facility, equity round or debt plan is publicly described; drawdown and use may need confirmation.
• Under construction: Reliable evidence indicates work has started on the relevant phase.
• Commissioned / operational: A source confirms the capacity is in service; specify the phase and metric.
• Live / expanding: Some capacity is live while later phases continue; record both.
• Policy target: A government objective, not a project status.

These are editorial labels for clear communication. They do not replace legal definitions or project-specific diligence. The source link and the date checked should remain visible.

What an investor, operator, vendor or lender should verify

Investor checklist

• What exactly is being financed: land, shell, fit-out, power, servers, network or an operating portfolio?
• Is the proposed capacity IT load or total site power demand?
• How much is committed, drawn, spent and still unfunded?
• What is the phase schedule and how does it depend on utility work?
• Is there a signed customer commitment? What are the customer’s rights and obligations?
• Who owns the asset, who operates it and who bears cost overruns?
• Are debt, interest, currency and refinancing assumptions stress-tested?
• What are the water, environmental, community and permitting conditions?
• How does the plan compare with alternative sites and other assets?

Operator checklist

• What customer workloads and density are being designed for?
• Are power, network, cooling and water arrangements deliverable at the site?
• Are the first phases sized to contracted demand?
• Are procurement and construction packages on a realistic schedule?
• Are service-level commitments supported by redundancy and tested operations?
• How will the company staff, secure and maintain the facility?

Vendor checklist

• Is the project at a stage where this package can be procured?
• Who selects and approves the vendor?
• Is the supplier qualified, insured and locally serviceable?
• What scope, quantity, specification and acceptance test are confirmed?
• What are the payment milestones, retention and cancellation terms?
• What inventory, currency, warranty and working-capital risks exist?
• Which evidence supports the estimated order window?

Lender checklist

• What is the borrower’s credit and construction track record?
• What is the collateral and what conditions must be met before drawdown?
• Is the project supported by a lease or other contracted demand?
• Are grid connection, permits, land and construction contracts in place?
• Can the project service debt under lower occupancy or delayed operations?
• Are water, power, climate and community risks assessed?
• Is the source of each capital-cost assumption clear and current?

Policymaker and ecosystem checklist

• Does the target distinguish announced proposals from approved and operational projects?
• Are eligibility rules simple enough for investors to understand?
• Are incentives measurable, capped and tied to delivery milestones?
• Is power and water planning coordinated with the target capacity?
• Are local suppliers, workforce and community impacts included?
• Is progress reported by phase and status rather than by headline MoU totals?

India’s regional investment map: follow infrastructure, not just state headlines

Investment announcements are often attributed to a state, while the delivery system is local to a city, substation, fibre corridor, water source and project parcel. A useful national map should therefore combine a state-level policy view with a city- and node-level infrastructure view. A state target can indicate policy intent; it cannot tell a developer whether a particular site can obtain firm power on the required schedule.

Mumbai and the wider Maharashtra market

Mumbai is a major established data-centre market in industry reporting, with connectivity, financial-services demand and an existing operator ecosystem. Maharashtra also features in AWS’s publicly stated cloud-infrastructure plan. A map should distinguish the AWS Maharashtra commitment from the separate national programme and should not assume that the entire state-level amount corresponds to one Mumbai project. For a specific investment case, the researcher should verify the site, power node, phase, customer and equipment package rather than relying on the state headline.

Chennai

Chennai’s connectivity position and role in the country’s digital infrastructure make it an important market to track. A regional investment brief should consider carrier networks, subsea-cable routes and interconnection along with land and power. The existence of a network advantage does not establish a specific project’s construction schedule, and an announced campus should be recorded separately from commissioned capacity.

Hyderabad

Hyderabad is a key cloud and technology location. Microsoft’s India South Central cloud region has been announced as live, which is a useful distinction between a cloud service launch and a new investment commitment. Cloud-region status does not disclose the physical MW available, customer utilisation or total capex. A vendor researching Hyderabad should identify the relevant campus and procurement chain rather than assuming that every cloud-region announcement is an open tender.

Bengaluru and Pune

Bengaluru and Pune have technology, enterprise, engineering and connectivity ecosystems that can support data-centre demand and supplier activity. That does not make every technology investment a data-centre project. For these cities, separate local demand signals—operator releases, lease activity, planning filings, power access and procurement notices—from broader software-sector growth. A market map should show whether a project is an operating facility, a proposed campus, a cloud region, or a supplier/office investment.

Delhi NCR, Noida and Greater Noida

The National Capital Region combines enterprise demand and state-level policy activity across Delhi, Uttar Pradesh and Haryana. These jurisdictions should not be merged in a state incentive comparison. The UP 2026 policy applies according to its text and eligibility criteria; the location of a project must be confirmed before assuming it qualifies. A project in Noida or Greater Noida should be mapped to its actual parcel, utility authority, policy jurisdiction, operator and delivery phase.

Visakhapatnam and Andhra Pradesh

Visakhapatnam has attracted attention from Google’s AI hub plan and Digital Connexion’s 1 GW campus MoU. These are large announcements with different structures and status. Google reported groundbreaking for its broader AI hub in April 2026; Digital Connexion’s public 1 GW figure is a plan under an MoU, not operating capacity. A project database should hold separate records for the two programmes, their announced partners, land, power work, phase schedule and public status. It should not collapse them into a single “Vizag 2 GW live” figure.

Jamnagar and Gujarat

Meta’s announced 168 MW leased facility with Reliance is a project-specific capacity signal in Jamnagar, while Gujarat’s 2026–29 policy sets a wider state-level investment and capacity ambition. They should be tracked in separate categories. The policy does not establish that the Meta–Reliance project receives every listed incentive; that would require confirmation against the policy and project’s eligibility.

A practical city-by-city comparison

When comparing investment locations, use the same fields for every city:

A project’s competitive position depends on the combined system. A state may offer an attractive incentive while a site faces a long grid-connection timeline; another location may have better fibre and customer access but higher land costs. A serious investment memo should show the trade-offs rather than create a simplistic winner ranking.

From announcement to live asset: the investment lifecycle

A company press release is one point in a longer capital lifecycle. The exact sequence differs by project, but a generic data-centre development may pass through the following gates:

1. Market hypothesis: The developer identifies a customer segment, location and potential demand.
2. Site control: The company secures a site option, lease, purchase or development agreement, subject to conditions.
3. Feasibility: Technical and commercial teams test power, water, connectivity, land, zoning, climate and cost.
4. Customer validation: The operator tests demand through discussions, proposals, reservations or leases.
5. Permits and utility process: The project pursues approvals and grid connection, often with dependencies outside the developer’s direct control.
6. Capital plan: Equity, debt, customer commitments and any eligible support are structured.
7. Design and procurement: Engineering is developed, and packages are specified, tendered and awarded.
8. Construction and commissioning: The building and systems are completed, tested and accepted.
9. Operations and expansion: Capacity is made available, customers are served and later phases are triggered by demand and infrastructure.

A project can move forward on some tracks while waiting on others. A site may be acquired before it has firm power; a customer may sign a lease before every construction package is awarded; a campus may break ground while a later phase remains contingent. A status tracker should therefore record progress by workstream rather than infer that one milestone proves all the others.

The funding stack can change as risk falls

Early development may rely on sponsor equity because the project has land, approval and demand risk. As the site secures approvals and customer commitments, it may become eligible for construction debt. After commissioning and stabilised occupancy, the asset may access refinancing or a lower-cost capital source. These are general financing patterns, not a guarantee that a particular Indian project can obtain any financing instrument. Currency, interest rates, credit quality, power contracts, tenant terms and the regulatory environment all matter.

The labels in this table are illustrative. Actual financing terms and eligibility are transaction-specific and should be verified with the borrower, lender and counsel.

Illustrative capital scenarios: useful for thinking, not forecasts

The following examples are fictional and are included to show how different structures change the way an investment should be tracked. They are not estimates of project cost, promised returns or recommendations.

Scenario A: a regional operator with phased development

A regional operator controls a suitable site and plans to develop it in phases. It first funds feasibility work and site preparation with sponsor capital. It seeks an anchor tenant before ordering the major equipment package. A lender may assess the site, lease and construction plan before considering a facility; the first phase may be financed differently from the second. The project tracker should distinguish planned ultimate campus capacity from the first phase and mark the later phase as conditional on demand and utility availability.

The investment story should not report the entire ultimate capacity as “under construction.” It should record the first phase’s site status, contracted demand, power work, construction start and planned commissioning date, along with the separate longer-term plan. If no tenant commitment is public, the map should say that rather than imply pre-leasing.

Scenario B: a global cloud buyer leases capacity from a local developer

A global technology company wants capacity in India but does not own the local project entity. A domestic developer builds the facility and leases it to the customer. The developer’s investment may fund the campus; the customer may invest in its own servers and network separately. A press release might disclose MW and a lease but omit the construction budget, lease rate and customer fit-out cost.

The map should have linked records: one for the developer’s project and capital; another for the customer lease and planned capacity; and a third for the customer’s broader India cloud commitment if relevant. The records should not be added as if they were unrelated capex. If the customer funds equipment directly, the source and amount should be stated separately.

Scenario C: a renewable power project supports a data-centre campus

An energy developer builds renewable generation and storage under a power arrangement that can serve one or multiple customers. The data-centre operator may invest in campus systems and pay for contracted energy; the energy company funds generation assets; grid and transmission investment may have a separate owner. The projects may have different commissioning dates.

An investment map should describe the energy agreement, ownership, capacity and delivery mechanism only to the extent public. It should not add the renewable project's full nameplate capacity to data-centre IT load or call all of the power investment direct data-centre capex. It should also record whether the arrangement is contracted, proposed or simply a renewable target.

Scenario D: state incentives support a large campus proposal

A developer submits a proposal that appears to meet a state policy’s size threshold. The company includes potential incentives in an early model. The application is not yet approved, and the project still needs land, power, customers and financing. In the public ledger, the policy is a target or potential support mechanism; the project is a proposal. The incentive should not be shown as cash received or as secured financing.

If a later official award confirms eligibility, update the project’s record with the approved support, amount or formula, qualifying capex, conditions and dates. If the application is denied or the project changes size, update the model again. This scenario illustrates why policy data should be versioned rather than permanently attached to a project without a status field.

Investment signals for suppliers: what is actionable and what is noise?

Suppliers often want to know which investment announcements may produce near-term orders. The amount is only one signal. A large national commitment can be spread over years and many business units. A smaller project with a named EPC and public tender could be more actionable for a particular vendor.

Stronger evidence signals

• A named project and location with a confirmed owner or operator.
• A public procurement notice or vendor qualification process.
• A tender for a clearly defined package with a deadline.
• A planning or construction milestone that is confirmed by a primary source.
• A named EPC, consultant or procurement channel, if publicly disclosed.
• A customer requirement that matches the supplier’s product and service capability.
• A published commissioning phase or schedule backed by current project evidence.

Weaker signals that need validation

• A national investment headline with no site or phase.
• A state policy target, MoU or press-release capacity number.
• A social post suggesting a supplier is “approved” without a public record.
• An estimated tender date generated from construction timing.
• A project directory entry copied from an old report and not rechecked.
• A broad AI strategy that does not identify a physical facility or buyer.

The supplier can use weaker signals for research prioritisation, but should not forecast a purchase order from them. A sales note should label whether a lead is source-confirmed, buyer-confirmed or inferred. That practice improves qualification and protects the business from spending heavily on opportunities that are not ready to procure.

Questions that convert a headline into a qualified opportunity

• Which legal entity owns the project?
• Who is acting as developer, operator, customer and EPC?
• Is the site confirmed or one of several candidate sites?
• Which phase is funded and what is its target delivery date?
• Is the procurement package open, awarded or not yet defined?
• What qualification documents are required?
• Does the buyer have an approved budget for this package?
• What are the contract, payment, warranty and acceptance conditions?
• What source proves the opportunity is active today?

If a team cannot answer these questions, it may still have a market lead—but it should be recorded as an unqualified signal, not a committed opportunity.

Common errors in India data-centre investment coverage

Adding every headline into one total

The same corporate plan may be reported at parent, subsidiary, state and project level. A state MoU can describe a component of a national plan; a new announcement may update rather than add to an earlier amount. Keep overlap tags, and never add two amounts without checking the wording.

Treating a target as capital raised

A policy target is an objective. A company’s expression of interest is not a signed investment agreement. An MoU is not a financing close. A land parcel is not a campus. Use the precise status label and avoid words such as “secured,” “funded” or “invested” unless the evidence supports them.

Confusing IT capacity with grid demand

A data-centre report may refer to operational IT capacity, while a utility or campus announcement may discuss power draw or connection. These differ because infrastructure needs energy for cooling and other systems, and facilities vary in design. Do not directly compare numbers unless the definitions and units match.

Using a company’s broad India pledge as data-centre capex

Large corporate commitments can include infrastructure, skills, jobs, software, operations, renewable procurement and other activities. Quote the stated scope; do not assign a data-centre-only label to a broader pledge.

Treating press coverage as independent confirmation

Multiple news articles may repeat the same company release. Ten articles repeating one claim do not create ten independent sources. Link the original announcement and use independent reporting for context, not as a substitute for the primary record.

Leaving the map undated

A map without a last-checked date can mislead even if it was accurate when first published. Projects change. Attach a source date and review cycle to every material field.

Suggested editorial method for publishing a credible market map

1. Define the unit of analysis: project, company programme, financing transaction or state policy.
2. Set inclusion rules: geography, project scale, period, capacity threshold and public-source standard.
3. Collect primary sources first: company releases, government policy pages, filings and lender announcements.
4. Use reputable secondary reports carefully: confirm dates, definitions and any stated caveats.
5. Create linked records: parent commitment, project, financing, customer agreement, energy asset and policy can be separate but related entries.
6. Standardise units: preserve currency, capacity metric, period and whether figures are nominal, planned, committed or deployed.
7. Record confidence: primary-source verified, secondary-source reported, inferred or unresolved.
8. Review material claims: especially cost, status, ownership, capacity and overlap.
9. Publish limitations: state where project costs or contract details are not disclosed.
10. Maintain an update log: date, field changed, source and reason.

This method makes a smaller, transparent database more useful than a large spreadsheet full of mixed definitions. Readers can follow the reasoning, challenge an assumption and return to the source.

Turn the public map into a decision brief

A company does not need a perfect estimate of the whole Indian market to make a useful decision. It needs a decision brief that states the choice, evidence, uncertainty and next test. A developer comparing two sites might ask which has a credible route to power and a customer pipeline. A supplier might ask where its product can be qualified. A lender might ask whether a phase can service debt if commissioning slips. A policymaker might ask whether incentives are attracting projects that can actually operate.

A one-page brief can contain five parts:

• Decision: What choice must be made, and by when?
• Evidence: Which primary sources establish project, policy, customer and infrastructure facts?
• Assumptions: Which values are estimates, and who supplied them?
• Risks: What could delay, resize or cancel the project?
• Next test: What source, customer discussion or engineering study would materially reduce uncertainty?

For instance, if a public announcement names a 1 GW campus but no phase-level schedule, the immediate decision may not be whether to sell into 1 GW of demand. It may be whether to invest a few days in validating the site, first phase, local procurement route and product fit. If the next public milestone or buyer conversation confirms those details, the company can increase its research effort. If it does not, the lead remains in a watchlist rather than a forecast.

The same discipline applies to macro totals. A reported commitment aggregate can help show that the sector attracts attention, but it cannot answer which location has the best risk-adjusted return or which vendor will win an order. For those decisions, project-specific evidence is more valuable than a bigger headline number. Keep the macro market view as context and the project diligence as the basis for action.

Related Gridvik guides

• See where potential value flows across the data-centre economy.
• Explore seven business models for companies entering the sector.
• Read the market expansion, capacity and city guide.
• Review grid readiness, state policy and procurement timing.

Frequently asked questions

What is the largest kind of data-centre investment in India?

Large announcements come from different categories and cannot be ranked fairly without normalising their scope. Google’s US$15B AI-hub announcement includes data-centre, energy and fibre components; Microsoft’s US$17.5B plan includes cloud and AI infrastructure, skills and ongoing operations; Amazon’s wider figures include multiple business areas. A project-specific plan such as Digital Connexion’s Visakhapatnam campus is different again. The “largest” answer depends on whether the comparison is a single campus, a company commitment, an operator programme or the wider AI ecosystem.

Is the US$126 billion figure actual money spent?

No. CBRE’s 2026 reporting describes cumulative investment commitments for the data-centre segment through 2025. It is an aggregated commitment estimate, not a bank statement or audited total of capital deployed. It should be attributed to CBRE, defined and not added to individual company announcements without checking for overlap.

Are AWS’s US$8.3B Maharashtra plan and US$12.7B India plan separate?

AWS describes the US$8.3B Maharashtra plan as part of its previously announced US$12.7B India cloud-infrastructure plan. A map may show the state allocation and national plan for different purposes, but it should flag the overlap rather than add both as independent totals.

Does Amazon’s US$35B or US$48B commitment represent data-centre construction?

No. Amazon’s US$35B 2025 announcement covered all its businesses in India. The company’s June 2026 update described an additional US$13B focused on AI and cloud infrastructure and a broader US$48B total for 2026–2030. These amounts are not a single facility capex budget. AWS cloud-infrastructure plans should also be tracked separately and checked for overlap.

Are Gujarat’s 7.5 GW and ₹6 lakh crore already secured investments?

They are policy targets and public ambitions under the Viksit Gujarat Data Centre Policy 2026–29. They are not, by themselves, built capacity, private financing closes or cash deployed. Individual projects and eligibility should be verified against the policy and project sources.

Are all MoUs binding contracts?

No. An MoU can express intent, facilitate coordination or define a framework. Its legal effect depends on its terms. Unless the binding status is publicly confirmed, label a project as an MoU or plan and do not describe it as a fully financed construction contract.

Can I estimate project cost from planned MW?

Not reliably from MW alone. Cost depends on site, design, redundancy, cooling, power infrastructure, rack density, phasing, construction market and what is included in the amount. Any estimate should state its assumptions and source. Do not publish a project cost as fact if the company has not disclosed it.

How do state incentives affect a project’s return?

They may reduce eligible costs or improve feasibility, but the value depends on eligibility, caps, approvals, timing and disbursement. A project should model the economics with and without the incentive and verify whether the policy applies to the specific project size, location, phase and activity.

Which investment types are most relevant to a vendor?

The vendor should follow the actual project and package rather than only the investor. A new equity announcement may not indicate an open procurement; a financing close, design issue, EPC appointment, tender or phase schedule may be more useful. The right signal depends on the product and buyer.

What does “investment by state” mean?

It can mean a company’s planned spend in a state, an MoU value, project cost, state incentive, market-research estimate or a government target. The label should specify which one. State-level totals can double-count national programmes unless each item is assigned to a geography consistently.

How often should an investment map be updated?

Update it when a material source changes the amount, scope, ownership, status, phase or schedule. A weekly or monthly review cadence can be useful, but the key is to record the last-check date and not claim a figure is current when it has not been reverified. A large commitment may change only once; a construction status can change more frequently.

Is Gridvik’s free sample the complete Indian project database?

No. The public resources page describes a 8-row preview and a larger paid product. A sample is a demonstration of structure, not a complete national inventory. Users should check Gridvik’s current methodology and service description and verify source links before relying on a record.

Is this an investment recommendation?

No. The article summarises publicly reported announcements and general capital categories. It does not recommend a project, security or financing. Professional investment, legal, tax, engineering and environmental diligence is necessary for a real transaction.

Sources and further reading

• CBRE — India’s Data Centre Market in a New Era, 2025 YTD update
• CBRE — India Alternate Sectors Outlook 2026
• BusinessToday — coverage of CBRE’s 2026 commitments estimate
• Savills — Data Centre Market Watch, H1 2026
• Cushman & Wakefield — India Data Centre Market, H1 2026
• Google — US$15B AI hub commitment
• Google — Visakhapatnam groundbreaking
• Microsoft — US$17.5B India cloud and AI commitment
• AWS — Maharashtra plan, described as part of its India plan
• Amazon — US$48B India investment update, June 2026
• Digital Connexion — Visakhapatnam MoU, Data Center Dynamics
• Meta — Reliance data-centre lease announcement
• Gujarat policy analysis, NASSCOM
• Uttar Pradesh Data Centre Policy 2026, Invest UP
• Gridvik Resources, About and methodology, Services
• Related guides: market expansion, power and policy bottlenecks, profit pools, and business models

This article is informational and based on public sources. Gridvik is an independent research firm and does not provide investment advice. Announcements and policies may change; verify current project status and eligibility with the original company or government source.