AI and data centres across the Telugu states

 

AI and data centres across the Telugu states

The next decade will be decided by power, water, connectivity and the ability to turn infrastructure into local capability

Illustrative editorial artwork. It is not a photograph of a specific facility.

A new contest over artificial intelligence is being built in concrete, copper, fibre and megawatts. Hyderabad is expanding its position as a cloud and enterprise technology centre; Visakhapatnam is being pitched as a coastal gateway for large-scale compute and international connectivity. The two Telugu states could form complementary poles in India’s AI economy over the next decade. But there is no automatic “twin-state” digital corridor yet. The proposition will be proved only if announced projects are financed and commissioned, and if power, water, networks, skills and local suppliers develop with them.

Public announcements across both states are large, but most describe commitments, memoranda, proposed campuses or multi-year build-outs—not operating capacity today. Project values can include generation, connectivity and cloud services alongside data-centre construction, while capacity announcements may refer to different phases and technical definitions. Treating every announced rupee and megawatt as delivered would overstate what the region has built.

Two states, different starting points

The complementarity case is a strategy to build, not a settled division of labour.

Hyderabad’s advantage is an existing concentration of software, cloud users, global capability centres, engineering talent and research institutions. Telangana’s official investment portal lists established cloud infrastructure and proposed data-centre clusters at Chandanvelly and Bharat Future City. The State has publicised a 400 MW CtrlS AI data-centre cluster and a separate 100 MW UPC–Volt proposal; it also announced a 1 GW TCS HyperVault campus in September 2026. These are meaningful signals of a pipeline, but the published announcements do not make them commissioned megawatts. Nor should unlike projects be simply added up and labelled live capacity. [1–4]

Andhra Pradesh has made a different, unusually large bet around Visakhapatnam. Google announced a roughly $15 billion investment over 2026–30 for its first AI hub in India, including a gigawatt-scale data-centre campus, subsea cable infrastructure and clean-energy support. Google and Adani broke ground in April 2026. “Gigawatt-scale” describes the announced design ambition; it is not a claim that one gigawatt of IT load is already installed or that the whole investment is data-centre construction alone. [5–6]

Vizag’s proposition combines compute with a coastline, ports, industrial demand and a planned cable landing gateway. Hyderabad offers proximity to a deep technology workforce, cloud customers and a broad services economy. A direct, resilient fibre route between them could support replication, disaster recovery, enterprise services and some latency-sensitive workloads. That is a plausible economic thesis—not a measured benefit already delivered. Cloud traffic can be routed across many locations, and subsea cables alone do not make a city a global AI hub.

The investment pipeline is not the economic outcome

What is on the table	Publicly stated scale	What remains to be demonstrated
Google AI hub, Visakhapatnam	About $15 billion over 2026–30; gigawatt-scale compute announced	Phased commissioning, actual Andhra Pradesh spend, power and fibre delivery, operating load and local supplier share
CtrlS AI campus, Telangana	400 MW cluster; ₹10,000 crore announcement	Land and power schedule, construction and commissioning milestones, final staffing and water plan
UPC–Volt, Bharat Future City	100 MW; ₹5,000 crore MoU	Financial close, construction start, grid connection, renewable supply and operating jobs
TCS HyperVault, Hyderabad	1 GW campus; ₹70,000 crore announcement	Project phases, capacity definition, approvals, energisation and water-neutrality evidence

 

This distinction matters for state competition. Announced capital is not the same as capex actually spent in-state; “jobs” may combine construction, contractors, indirect multipliers and permanent roles; a capacity target may refer to IT load, facility power or an entire multi-phase site. The governments should publish a common dashboard with project stage, MW definition, grid connection date, water source and use, construction workforce, permanent headcount, local hiring and supplier spend. That would make the race legible to residents and investors.

The hard constraint is electricity

AI computing is a power-intensive industrial load. India’s official data-centre capacity rose from about 375 MW in 2020 to around 1,500 MW in 2025, according to a Ministry of Electronics and IT statement. The same national trend is accompanied by increasingly ambitious state proposals. Telangana’s investment pitch highlights strong power infrastructure; Andhra Pradesh’s project depends on new transmission, generation and grid coordination. Neither pitch by itself answers where every incremental unit of electricity will come from at peak demand, how quickly substations and transmission lines can be built, or what the delivered cost will be. [7]

The question is not simply whether renewables are purchased on paper. Large campuses need firm, reliable supply around the clock. States will have to combine renewable generation with storage, grid upgrades, flexible demand and transparent accounting for any backup generation. The right public-interest test is whether new infrastructure expands clean and dependable supply for the wider economy—or shifts costs and reliability risks onto households and existing industry.

Water and land make the build-out local

Cooling design can change a facility’s water footprint, but “air cooled” or “water neutral” is not a complete answer. Operators and regulators should publish site-specific water requirements, expected consumption, source, reuse rate, seasonal peak and discharge quality. WRI India’s 2026 analysis warns that a large share of Indian data centres are located in water-stressed regions, underscoring that location and cooling choices carry operational as well as environmental risk. In Vizag, the project has already faced public-interest litigation and scrutiny around approvals, land and environmental safeguards; the High Court allowed work to continue at the developer’s risk and cost, subject to the case’s outcome, while directing authorities to maintain vigilance. That is not a final judicial endorsement of the project. [8–9]

The Andhra Pradesh High Court record also brings the project’s permissions into sharper focus. A State affidavit reported in September described a clearance for the building proposal and listed 307 kilolitres a day of water for the 160-acre campus filing, including fresh and recycled supplies. These are filing figures for a specified project component, not verified eventual consumption for every campus or the wider district. Transparency across the whole cluster—rather than relying on a single headline figure—is essential. [10]

Land governance matters just as much as water. Tarluvada residents have raised concerns over access to land and rehabilitation; the State and companies have disputed allegations of improper acquisition and said legal processes and engagement are in place. The path from investment approval to a durable local social licence depends on verifiable records: who owned or cultivated each parcel, what compensation or replacement land was provided, what commitments remain, and how grievances are resolved.

The jobs story is outside the server hall

Data centres are capital intensive and operate with relatively small permanent teams compared with their investment value. Construction creates a larger, temporary demand for civil engineering, electrical systems, cooling, logistics, security and accommodation. Once a facility is running, specialised technicians, network engineers, cybersecurity teams and operations staff remain essential, but headline job estimates should state how many roles are permanent, local and directly employed. Recent reporting on the sector has stressed the need to separate these categories; a national projection cannot be treated as a site-level promise. [11]

The larger prize is a cluster effect: local firms that build software and AI services on cloud platforms; cybersecurity and systems integrators; network and subsea-cable operations; power electronics and cooling suppliers; and AI adoption by ports, pharma, manufacturing, agriculture and health services. Telangana begins with a more mature digital-services base. Andhra Pradesh can connect digital infrastructure to port logistics, industrial corridors and its growing technology ambitions. Both need stronger routes from engineering colleges into data-centre operations, applied AI, energy systems and research.

Nor does data residency automatically mean sovereign AI or local value capture. India’s Digital Personal Data Protection Act does not impose a blanket requirement that all personal data stay in India; government restrictions can apply to specified data, and other sectoral rules may have their own requirements. Hosting infrastructure domestically can support control, resilience and compliance, but indigenous models, skilled teams, accessible compute and trusted datasets require deliberate investment. India’s AI Mission is already widening shared compute access through a national portal, so private hyperscale campuses are one layer of the AI stack—not the whole strategy. [12–13]

A strategy for the next decade

For Andhra Pradesh and Telangana, the strongest play is coordination with clear accountability. First, plan data-centre zones around verified power capacity, water availability, flood and cyclone exposure, fibre diversity and land rights—not incentives alone. Second, require every large proposal to report comparable definitions for IT load, facility demand, water use, employment and local procurement. Third, link concessions to milestones: commissioning, renewable-energy delivery, reuse, training, local supplier development and public reporting.

Fourth, build the corridor that makes the two-state thesis real: diverse fibre routes between Hyderabad and Vizag, resilient backhaul to multiple cable landings, and shared standards for cybersecurity, disaster recovery and service quality. Fifth, invest in the surrounding economy: applied-AI labs with universities, paid apprenticeships for technicians, vendor certification, startup access to cloud credits and compute, and adoption programmes in ports, pharma and manufacturing. Finally, publish a yearly public scorecard of commitments versus delivery, including the State’s fiscal cost and the resource burden borne by each locality.

The next decade will be measured in delivery

The Telugu states have a credible opening to shape India’s AI infrastructure map. Hyderabad brings an established technology economy; Visakhapatnam could add a new coastal compute and connectivity node. Their roles may complement one another, but geography does not guarantee integration, and project announcements do not guarantee a broader AI economy.

By 2036, the score should not be the largest announced gigawatt total. It should be how much capacity actually runs, how reliably and cleanly it is powered, how transparently it uses water and land, how many skilled local jobs endure, and how many businesses and public services use the computing capacity to make products and solve problems. That is the difference between hosting machines and building an AI economy.

Sources and reporting notes

[1] Telangana Government, Data Centres sector overview: https://invest.telangana.gov.in/sectors/data-centres/

[2] Telangana Government, CtrlS ₹10,000 crore / 400 MW announcement: https://invest.telangana.gov.in/the-telangana-govt-signs-an-mou-with-ctrls-datacenters/

[3] Telangana Government, UPC–Volt 100 MW / ₹5,000 crore MoU: https://invest.telangana.gov.in/telangana-rising-delegation-signed-an-mou-with-the-upc-volt-joint-venture-at-wef-2026-in-davos/

[4] Telangana Government, TCS HyperVault 1 GW campus announcement, 5 September 2026: https://invest.telangana.gov.in/tcs-hypervault-to-build-%E2%82%B970000-crore-1-gw-ai-data-centre-campus-in-hyderabad/

[5] Google, first AI hub in India, 14 October 2025: https://blog.google/intl/en-in/company-news/our-first-ai-hub-in-india-powered-by-a-15-billion-investment/

[6] Google / Adani, Vizag AI hub groundbreaking, 28 April 2026: https://www.adani.com/newsroom/media-releases/google-breaks-ground-on-india-ai-hub-launching-a-national-industrial-ecosystem

[7] Press Information Bureau, national data-centre capacity, 13 March 2026: https://www.pib.gov.in/PressReleasePage.aspx?PRID=2239616&lang=2&reg=48

[8] WRI India, “More Than Half of India’s Data Centres Are in Water-Stressed Regions”, 18 May 2026: https://wri-india.org/data/more-half-indias-data-centres-are-water-stressed-regions

[9] Andhra Pradesh High Court order, Bolisetty Satyanarayana v State of Andhra Pradesh, 6 October 2026: https://indiankanoon.org/doc/130147101/

[10] ThePrint, State affidavit on Vizag facility clearance and filing details: https://theprint.in/judiciary/google-adani-vizag-data-centre-got-green-nod-on-built-up-area-not-ops-andhra-govt-affidavit-explains/3025598/

[11] India Today, data-centre employment categories and estimates, 29 August 2026: https://www.indiatoday.in/business/story/india-data-centres-ai-jobs-employment-outside-server-rooms-how-many-artifcial-intelligence-2982401-2026-08-29

[12] Ministry of Electronics and IT, Digital Personal Data Protection Act 2023: https://www.meity.gov.in/static/uploads/2024/02/Digital-Personal-Data-Protection-Act-2023-1.pdf

[13] National e-Governance Division, IndiaAI compute access and GPU capacity, 2 April 2026: https://negd.gov.in/press_release/indiaai-mission-expands-ai-ecosystem-with-affordable-compute-and-startup-support/

BIZ MASALA

The business of change

— Biz Masala Editorial Desk