Meta-Reliance AI Data Centre Jamnagar: India's Digital Infrastructure Leap

On 10 June 2026, Meta Platforms Inc. and Reliance Industries Ltd. (RIL) announced India's first AI-enabled data centre at Jamnagar, Gujarat — a 168 MW built-to-suit facility to be developed by Reliance and leased by Meta, powered entirely by renewable energy and cooled using desalinated seawater. This follows Meta's $5.7 billion investment in Jio Platforms (2020) and a $100 million joint venture (August 2025) to develop Llama-based AI solutions. The deal sits within India's rapidly expanding data centre ecosystem, governed by the Digital Personal Data Protection Act, 2023 (DPDPA), and fuelled by the government's IndiaAI Mission (₹10,372 crore) targeting 38,000+ GPUs for public compute access.

Core Concept & Definition — AI Data Centres

What is a Data Centre?

A data centre is a specialized facility used to store, manage, process, and transmit large volumes of digital data. It houses computing hardware (servers, storage arrays, networking equipment), power infrastructure, and cooling systems in a controlled, secure environment.

An AI Data Centre is an advanced variant designed specifically to support AI training and inference workloads. Unlike traditional data centres (CPU-centric), AI data centres are built around GPU clusters (Graphics Processing Units) capable of massive parallel computation — essential for training Large Language Models (LLMs) and deep learning systems.

AI Data Centre vs Traditional Data Centre — Key Differences
Parameter	Traditional Data Centre	AI Data Centre
Primary Processor	CPU (Central Processing Unit)	GPU / AI Accelerators (Parallel)
Workloads	Web hosting, databases, ERP	LLM training, inference, ML, HPC
Power Density (per rack)	5–10 kW	30–100+ kW (high-density)
Cooling	Air cooling (standard)	Liquid cooling / immersion cooling
Networking	Standard Ethernet	InfiniBand / 800GbE high-bandwidth
Cost	Lower capex	Significantly higher (GPUs expensive)
Example	Government e-governance servers	Meta Jamnagar, Google Visakhapatnam

GPU — Graphics Processing Unit LLM — Large Language Model HPC — High Performance Computing PUE — Power Usage Effectiveness Hyperscale Colocation Edge DC Sovereign Cloud Data Localisation NVLink / InfiniBand

📌 Key Term: PUE

Power Usage Effectiveness (PUE) = Total facility power ÷ IT equipment power. Ideal PUE = 1.0. World-class AI data centres target PUE of 1.1–1.2. Higher PUE = more energy wasted on cooling/lighting.

💡 Exam Tip

UPSC often asks about GPU vs CPU distinction. Remember: GPUs have thousands of smaller cores for parallel processing; CPUs have 8–64 cores for sequential processing. AI = parallel = GPU.

An AI Data Centre = GPU-centric + high-density power + liquid cooling + high-bandwidth networking — purpose-built for LLM training and inference workloads.

Historical Evolution of Data Centres in India

2000s–2010

Government-era: India's early data centres were largely for government e-governance platforms and large PSUs. Limited private presence. NIC (National Informatics Centre) runs government data centres.

2010–2018

Colocation boom: Private operators (CtrlS, Sify, STT GDC India) expand; telecom companies invest in backhaul and inter-DC links. Data residency concerns begin driving demand. CAGR ~24% from 2019.

2020

Meta (then Facebook) invests $5.7 billion in Jio Platforms. Data Centre Policy (Draft) released by MeitY. India's installed DC capacity: ~375 MW. October 2022 — DEA designates DCs as infrastructure under 'Communication' category.

2023

Digital Personal Data Protection Act, 2023 (DPDPA) passed — August 11, 2023. Presidential assent. India's first standalone data protection law. Began operationalising via Rules in November 2025.

2024–25

IndiaAI Mission approved — ₹10,372 crore over 5 years (March 2024). India capacity ~1.5 GW (end 2025). Microsoft pledges $17.5 billion. Amazon commits $35 billion by 2030. DPDP Rules 2025 notified November 13, 2025.

January 2025

Reliance announces plans for potentially the world's largest DC campus in Jamnagar (3 GW target). Nvidia partnership for Blackwell GPU-powered facility reported.

August 2025

Meta–Reliance form $100 million joint venture for Llama-based enterprise AI solutions in India and international markets.

June 10, 2026

Meta-Reliance 168 MW AI Data Centre announced in Jamnagar, Gujarat — India's first built-to-suit AI-enabled data centre for a global tech hyperscaler.

📌 Key Fact

India's DC capacity grew from ~375 MW (2020) → ~1.5 GW (2025) — a 4x increase in 5 years. Projected: 8 GW+ by end of decade (TechCrunch/Government data).

India's data centre journey: Government servers (2000s) → Colocation boom (2010s) → Hyperscale wave (2020–23) → AI infrastructure supercycle (2024–26).

Technical Architecture & Working of an AI Data Centre

Core Infrastructure Layers

AI Data Centre — Infrastructure Components
Layer	Component	Function / UPSC Relevance
Compute	GPU servers (NVIDIA H100/Blackwell, AMD Instinct)	Parallel processing for AI model training; thousands of cores vs CPU's 64
Interconnect	NVLink, InfiniBand, 800GbE switches	High-speed GPU-to-GPU communication for distributed training
Storage	High-bandwidth flash/NVMe, object storage	Rapid access to training datasets (petabytes scale)
Power	UPS, generators, PDUs; 100s of MW capacity	20–40% of operational cost; Meta Jamnagar = 168 MW = ~1.4 lakh homes' power
Cooling	Liquid cooling, immersion, heat exchangers	GPU racks generate extreme heat; air cooling inadequate above 30kW/rack
Networking	Fibre optic backbone, submarine cable proximity	Jamnagar near western submarine cable landing stations + Jio fibre network
Security	Physical + cyber; CERT-In compliance	Critical Information Infrastructure (CII) designation possible

Meta Jamnagar — Specific Technical Facts

168 MW

Initial Capacity

2 yrs

Delivery Timeline

100%

Renewable Energy

Desalinated

Seawater Cooling

~1 GW

RE from CleanMax + 4th Partner

Cooling Innovation: Uses desalinated seawater — addresses freshwater scarcity, leverages Jamnagar's coastal location
Renewable energy: Meta-Reliance announced partnerships with CleanMax and Fourth Partner Energy — together adding ~1 GW RE capacity
Meta bears full cost of energy and water; Reliance acts as single-window solutions provider (design, construction, utilities, connectivity)
Location rationale: Proximity to western submarine cable landing stations; Jio fibre backbone; Reliance's industrial campus; renewable energy access

✅ Technical Fact

A typical high-performance AI server integrates 8 NVIDIA H100 GPUs connected via NVLink, each with High Bandwidth Memory (HBM). A 168 MW facility can potentially host tens of thousands of such servers.

AI data centres differ from traditional ones by deploying GPU clusters, liquid cooling, and high-bandwidth networking — Jamnagar adds desalinated seawater cooling as a sustainability innovation.

Key Statistics & Data Dimensions — India & Global

~1.7 GW

India DC Capacity (end-2025)

53.7 GW

USA Installed Capacity

31.9 GW

China Installed Capacity

8 GW+

India Target by 2030

India's Share of Global DC Capacity

20%

India's Share of Global Data Generated

Major BigTech AI Infrastructure Investments in India (2025–2026)
Company	Amount	Location / Highlight	Year
Meta + Reliance	168 MW DC (lease)	Jamnagar, Gujarat — 1st AI-enabled DC in India	Jun 2026
Google	$15 billion (5 yrs)	Visakhapatnam — largest AI hub outside US	Oct 2025
Microsoft	$17.5 billion (by 2029)	Cloud + AI infra + sovereign cloud across India	Dec 2025
Amazon / AWS	$35 billion (by 2030)	AWS: $12.7 bn; Maharashtra focus	Ongoing
OpenAI + Tata	100 MW → 1 GW target	TCS HyperVault DC; 1st anchor customer	Feb 2026
Adani (AdaniConneX)	$100 billion (by 2035)	Renewable-powered hyperscale AI-ready facilities	Ongoing

India's Data Centre Market — Key Numbers
Metric	Figure	Source / Note
DC Capacity 2020	~375 MW	Government data
DC Capacity 2025	~1.5–1.7 GW	JLL / TechCrunch
Projected 2030	8 GW+	Industry estimates
CAGR since 2019	~24%	JLL India
Mumbai share	~54%	JLL H1 2025 Report
Vacancy rate (H1 2025)	4.3%	JLL — supply constrained
India GPUs (IndiaAI)	38,000+	IndiaAI Mission (Dec 2025)
GPU rental price	~₹65/hour	IndiaAI Compute Portal

📊 Data Point

India hosts ~20% of the world's data but only ~3% of global data centre capacity — this infrastructure gap is the key driver of the current investment supercycle.

India: 20% of world's data, only 3% of global DC capacity — the structural gap explains why Meta, Google, Microsoft, AWS & OpenAI are all racing to build here.

Types & Classification of Data Centres

Classification of Data Centres — UPSC Reference Table
Type	Definition	Scale	India Example
Enterprise / Captive	Company-owned for internal use only	Small–medium	PSU server rooms, bank DCs
Colocation (Colo)	Third-party facility; multiple tenants share space, power, cooling; own equipment	Medium–large	STT GDC India, Sify, CtrlS
Hyperscale	Very large (100+ MW), built by or for cloud giants; highly automated, standardized	100 MW – GW scale	Meta Jamnagar (168 MW), Google Vizag (1 GW)
Edge Data Centre	Small, distributed, placed close to end-users; reduces latency for 5G/IoT	1–10 MW	Telecom tower-co facilities
Sovereign / Government	Owned/operated by or for government; stores sensitive state data	Variable	NIC, MeitY DCs, NDCP proposal
AI-Optimised DC	GPU-centric design, liquid cooling, high-bandwidth fabric; purpose-built for ML/AI	Hyperscale+	Meta Jamnagar, Yotta NM1 (GPU cluster)

Hyperscale DC — Characteristics

100 MW+ capacity
Built for cloud/AI workloads
Single owner or major anchor tenant

High automation, standardized design
Long-term lease (10–20 yrs)
Massive power demand (grid impact)
Often includes renewable energy

Colocation DC — Characteristics

Multiple tenants, shared infra
Tenants own their servers
Flexible — monthly/annual contracts
Internet Exchange Points (IXPs)
Revenue: per-kW or per-rack pricing
India rate: ₹6,650–8,500/kW/month
Example: STT GDC India (28 DCs)

📌 Built-to-Suit vs Colocation

Meta-Jamnagar is a built-to-suit arrangement: Reliance builds a DC customised to Meta's specs, Meta leases capacity. This is different from colocation (tenant brings own servers to a shared facility).

Key classification: Enterprise → Colocation → Hyperscale → Edge → Sovereign. Jamnagar = Built-to-suit Hyperscale AI Data Centre. Mumbai hosts ~54% of India's colo capacity.

Legal & Policy Framework Governing Data Centres in India

Key Laws, Policies & Regulations — Data Centres in India
Instrument	Year	Key Provision / Relevance
Draft National Data Centre Policy	2020 / 2025 (updated)	Infrastructure status; single-window clearance; 20-year tax holiday for qualifying DCs; land bank near IT corridors
Infrastructure Status (Data Centres)	Oct 2022	DEA notification via Ministry of Finance — DCs classified as infrastructure under 'Communication' category (on TRAI recommendation)
Digital Personal Data Protection Act, 2023 (DPDPA)	Aug 2023	India's first standalone data protection law; governs data processing, consent, cross-border transfers; Presidential assent 11 Aug 2023
DPDP Rules, 2025	Nov 2025	Operationalises DPDPA; data localisation provisions; significant data fiduciary (SDF) obligations; Data Protection Board of India
TRAI Recommendations (2020)	2020	Recommended infrastructure status, National Single Window System (NSWS) for DC clearances, policy on CDNs and IXPs
CERT-In Guidelines	Ongoing	Binding cybersecurity compliance for DC operators; incident reporting mandates; critical information infrastructure rules
Union Budget 2026–27	2026	Long-term policy initiative for India as global digital infrastructure hub; AI infrastructure given strategic priority

DPDPA 2023 — Key Concepts for UPSC

Concept	Definition
Data Fiduciary	Entity that determines purpose and means of processing personal data
Significant Data Fiduciary (SDF)	Designated by Central Govt based on volume/sensitivity; extra obligations
Data Principal	The individual whose personal data is being processed
Data Localisation	Requirement to store certain data within India; DPDPA uses "blacklist" model — transfers allowed unless restricted by Govt
Cross-border Transfer	Section 16: Govt may restrict transfer to notified countries (expected: countries sharing land border with India)
Data Protection Board	Quasi-judicial body for adjudicating complaints and imposing penalties under DPDPA

⚠ Common Trap — Data Localisation

UPSC may test: "DPDPA 2023 mandates complete data localisation." — FALSE. The Act uses a blacklist approach — transfers are allowed by default UNLESS the Central Govt restricts specific countries. Only certain SDF data faces localisation obligations.

💡 Exam Tip

DPDPA 2023 is notable for two firsts: India's first standalone data protection law, and first Indian law to use "she/her" pronouns instead of the traditional "he/him."

DPDPA 2023 + Infrastructure Status (2022) + DPDP Rules 2025 + National DC Policy = India's four-pillar regulatory architecture for data centres.

IndiaAI Mission & Government AI Infrastructure Initiatives

IndiaAI Mission — Snapshot

₹10,372 cr

Budget (5 years)

38,000+

GPUs Deployed

₹65/hr

GPU Rental Rate

600

Data Labs Planned

₹2,000 cr

AI Funding Budget 2025–26

IndiaAI Mission — Pillars & Key Facts
Pillar	Key Provision
IndiaAI Compute	10,000+ GPU supercomputer; PPP model; 38,231 GPUs deployed as of Dec 2025; accessible to startups/researchers
IndiaAI Innovation Centre	Develop Foundation Models and Large Multimodal Models; govt-approved funds for indigenous LLMs
IndiaAI Datasets Platform	Curated training data in 22 official Indian languages; enables training India-specific AI models
IndiaAI FutureSkills	Expand graduate/PG AI education; 600 data labs nationwide
Foundation Models	Sarvam AI selected for 120B parameter Sovereign LLM; IIT Bombay consortium (₹988.6 cr) for 1 trillion parameter LLM

📌 Sarvam AI

Selected (April 2025) to build India's Sovereign LLM Ecosystem — open-source 120 billion parameter AI model. Earlier models: Sarvam-1 (2B params), Sarvam-M (24B params). Use cases include "2047: Citizen Connect" and "AI4Pragati."

📌 Semiconductor Connection

Budget 2025–26: Semiconductor manufacturing support doubled to ₹2,499 crore. Tata Group's 910B chip plant in Dholera (50,000 12-inch wafers/month). Deep Tech Fund of Funds: ₹20,000 crore. India building the full stack: chips → GPUs → DCs → AI models.

IndiaAI Mission = India's sovereign response to global AI infrastructure race — 38,000+ GPUs + 600 data labs + Sarvam Sovereign LLM + ₹10,372 crore budget.

Inter-linkages & Connected Concepts

AI Data Centres — Linked UPSC Topics & Connections
Linked Concept	Connection to AI Data Centres
Digital India Mission	DCs are the backbone infrastructure enabling e-governance, cloud services, and digital payments at scale
5G Rollout	5G generates massive data volumes requiring edge and hyperscale DCs for low-latency processing
UPI / Digital Payments	Processed through data centres; RBI's data localisation mandate — payment data must stay in India
Semiconductor / CHIPS Mission	GPUs require advanced chips; Tata + CG Power chip fabs in Gujarat complement DC ecosystem
Renewable Energy / Green Energy	DCs are energy-intensive; India's 9 GW DC target by 2030 = ~3% of India's power; green mandates driving RE adoption
Data Sovereignty	National security concern; data localisation debate; DPDPA blacklist model; critical infrastructure designation
FDI & Make in India	BigTech investments ($17.5 bn Microsoft, $15 bn Google, $35 bn Amazon) = major FDI inflows into India
Submarine Cables	Jamnagar location chosen partly for proximity to western submarine cable landing stations
Artificial Intelligence (AI)	AI training requires GPUs in DCs; inference (using trained models) also runs in DCs; no AI without DC infrastructure
Smart Cities / Urban Infra	Edge DCs needed in smart cities for IoT data processing; land and power availability key location factors

Digital India 5G UPI Data Localisation Semiconductors Renewable Energy Data Sovereignty FDI Policy Submarine Cables IndiaAI Mission CERT-In Critical Infra Protection NIC (Govt DCs)

💡 Exam Tip — Energy Link

UPSC Environment papers may link DCs with energy consumption. Key fact: India's DC expansion to 9 GW by 2030 will consume approximately 3% of India's total power — creating pressure on grid and accelerating RE demand.

AI Data Centres link to: 5G, Digital India, UPI, Semiconductor Mission, Renewable Energy, FDI policy, Data Sovereignty, and Submarine Cables — an integrated UPSC prelims topic.

Global Data Centre Landscape — India vs the World

Global Data Centre Capacity Comparison (2025–26)
Country / Region	Installed Capacity	Key Hub	Notable Feature
USA	~53.7 GW	Virginia (Data Centre Alley)	Dominant cloud: AWS, Azure, Google; hyperscaler HQs
China	~31.9 GW	Beijing, Guangdong, Inner Mongolia	State-backed; separate internet ecosystem
Europe (UK/Germany)	Hundreds of facilities each	London, Frankfurt, Amsterdam	GDPR compliance; strict data sovereignty
India	~1.7 GW (2025)	Mumbai (~54%), Delhi NCR, Hyderabad	Fastest-growing; 20% world data, 3% capacity
Japan	Major Asia hub	Tokyo	Earthquake-resilient; high uptime standards
Singapore	~1 GW	Singapore island	Imposed DC moratorium (2019–22) due to power strain

India's Strategic Advantages vs Challenges

Advantages

900M+ internet users (largest base)
20% of world's data generated
Low cost of land and labour
Strong digital talent pool
Government policy support (IndiaAI, DPDPA)
Proximity to submarine cable landing stations (Mumbai, Chennai)
Large domestic AI demand (fintech, healthcare, govt)

Challenges

Fragmented state policies (no single national framework)
Power supply reliability issues (outages)
Water scarcity for cooling (hence Jamnagar uses desalinated water)
Skill shortage in DC operations
Environmental concern (carbon footprint)
Regulatory uncertainty (DPDPA Rules still phased in)
Competition from China (31.9 GW vs India 1.7 GW)

📌 Google Visakhapatnam

In October 2025, Google announced a $15 billion, 5-year investment for a 1 GW AI hub in Visakhapatnam — its largest AI investment outside the US. Partners: AdaniConneX + Nxtra by Airtel. Includes a new international subsea gateway.

US (53.7 GW) > China (31.9 GW) >> India (1.7 GW) — but India is the fastest-growing market; structural gap between data generation (20%) and DC capacity (3%) drives investment boom.

Current Affairs — AI Data Centres & India Digital Infrastructure

📊 Current Affairs — LatestLY / ANI / TechCrunch · June 2026

Meta-Reliance 168 MW AI Data Centre, Jamnagar (June 10, 2026): Meta Platforms and Reliance Industries announced India's first built-to-suit AI-enabled data centre at Jamnagar, Gujarat. Initial capacity: 168 MW; renewable energy + desalinated seawater cooling; to be built by Reliance (design, construction, utilities, connectivity) and leased by Meta. Delivery expected within 2 years. Meta also announced ~1 GW of new renewable energy partnerships with CleanMax and Fourth Partner Energy. Zuckerberg: "This world-class facility in Jamnagar will help us scale our AI infrastructure globally."

📊 Current Affairs — DLA Piper / Wikipedia · November 2025

DPDP Act 2023 and Rules 2025 operationalised (November 13, 2025): MeitY notified the Digital Personal Data Protection Act, 2023 and the DPDP Rules, 2025 — making India's first standalone data protection law operative. Key provisions being phased in through May 2027. The Rules include significant data fiduciary obligations, breach reporting timelines, and cross-border transfer rules.

📊 Current Affairs — Bloomberg / TechRadar · October 2025

Google's $15 billion AI hub in Visakhapatnam (October 2025): Google announced plans for a 1 GW AI data centre campus in Visakhapatnam — its largest AI infrastructure investment outside the US. Construction began April 2026 through AdaniConneX and Nxtra by Airtel partnerships. Includes a new international subsea gateway.

📊 Current Affairs — PIB / IndiaAI · May 2025

IndiaAI Mission GPU milestone (2025): India's national compute capacity crossed 34,000+ government-managed GPUs under IndiaAI Mission. Sarvam AI selected (April 2025) for 120B-parameter Sovereign LLM. IIT Bombay consortium allocated ₹988.6 crore for 1 trillion parameter LLM. Budget 2025–26: AI funding quadrupled to ₹2,000 crore; MeitY received ₹26,026 crore (48% increase).

📊 Current Affairs — PSU Watch / GK Today · June 2026

Meta-Reliance strategic relationship timeline: 2020 — Meta invests $5.7 billion in Jio Platforms. August 2025 — Meta-Reliance form $100 million JV for Llama-based enterprise AI. June 2026 — 168 MW AI Data Centre at Jamnagar. Aligned with Government of India's policy focus on data centres as strategic national infrastructure.

💡 Exam Tip — What to Remember for 2026 Prelims

Four key facts: (1) Meta-Reliance = 168 MW, Jamnagar, June 2026; (2) Google = $15B, 1 GW, Visakhapatnam; (3) DPDP Rules = November 2025; (4) IndiaAI = ₹10,372 crore, 38,000+ GPUs. Any of these could appear as a statement-based MCQ.

The June 2026 Meta-Reliance announcement + DPDP Rules Nov 2025 + Google Vizag Oct 2025 + IndiaAI 38,000 GPUs = the current affairs cluster for this topic.

PYQ Pattern & Common Traps

Statement True / False Table — AI Data Centres & Policy
Statement	True/False	Reason
DPDPA 2023 mandates complete data localisation — all data must stay in India	❌ FALSE	DPDPA uses a blacklist model — data transfers allowed by default unless govt restricts specific countries
Data Centres were designated as 'Infrastructure' under India's Communication category in 2022	✅ TRUE	DEA notification, Oct 2022, on TRAI recommendation
GPUs use fewer cores than CPUs but each core is more powerful for AI tasks	❌ FALSE	GPUs have thousands of smaller cores optimized for parallel processing; CPUs have fewer but more powerful cores for sequential tasks
Meta-Reliance Jamnagar facility will be cooled using river water	❌ FALSE	Cooled using desalinated seawater, not river water
IndiaAI Mission was approved with a budget of ₹10,372 crore over 5 years in 2024	✅ TRUE	Approved March 2024; includes compute capacity, foundation models, FutureSkills pillars
Google's Visakhapatnam AI hub is Google's largest AI infrastructure investment globally	❌ FALSE	It is Google's largest AI hub outside the US, not globally
DPDPA 2023 uses "she/her" pronouns — this is unusual for Indian legislation	✅ TRUE	First Indian legislation to use she/her instead of traditional he/him pronouns

⚠ Trap 1 — Meta Investment Amount

UPSC may confuse Meta's 2020 Jio investment ($5.7 billion) with the 2025 JV ($100 million) or the 2026 data centre announcement. These are three separate events — learn all three with their years and amounts.

⚠ Trap 2 — Capacity vs Target

India's current DC capacity (~1.7 GW in 2025) vs projected target (8 GW+ by 2030) vs IndiaAI's GPU pool (38,000 GPUs — not MW). Don't confuse units: MW = power capacity; GPUs = compute units.

⚠ Trap 3 — DPDPA Commencement

DPDPA was passed August 2023 but did not immediately come into force fully. DPDP Rules notified November 13, 2025; phased operationalisation through May 2027. Exam may ask about when it was "notified" vs "passed."

⚠ Trap 4 — Hyperscale vs Colocation

Meta Jamnagar = hyperscale built-to-suit (Reliance builds, Meta leases entire facility). Not a colocation (where multiple tenants share a facility). UPSC may present this as a colocation arrangement — reject it.

⚠ Trap 5 — Mumbai vs Other Hubs

Mumbai has historically dominated (~54% of India's DC capacity), but new investments are shifting to Visakhapatnam, Hyderabad, Jamnagar, Maharashtra. "Mumbai is the only major DC hub in India" = FALSE.

💡 How UPSC Tests This Topic

Expect: Statement-based questions on DPDPA provisions; match-the-pair (company → investment → location); correct/incorrect statements on GPU vs CPU; questions on PUE definition; linkage questions (DPDPA + data centres + FDI).

Top 3 traps: (1) Blacklist ≠ complete data localisation; (2) Meta Jamnagar = built-to-suit hyperscale, not colocation; (3) Google Vizag = largest outside US, not globally.

MCQ Practice — AI Data Centres & India Digital Infrastructure

1Consider the following statements about the Digital Personal Data Protection Act, 2023:
1. It mandates complete data localisation — all personal data must be stored within India.
2. It uses a "blacklist" model for cross-border data transfers, allowing transfers by default unless restricted.
3. It was the first Indian legislation to use "she/her" pronouns.
Which of the statements given above is/are correct?

Correct: (c) 2 and 3 only

Statement 1 is FALSE — DPDPA uses a blacklist model, not blanket localisation. Transfers are permitted by default unless the Central Govt restricts specific countries. Statement 2 is TRUE. Statement 3 is TRUE — DPDPA 2023 is the first Indian law to use she/her pronouns.

2With reference to the Meta-Reliance AI Data Centre announced in June 2026, which of the following statements is correct?

Correct: (c)

(a) Wrong — it is a built-to-suit facility, not a shared colocation. (b) Wrong — cooling uses desalinated seawater, not river water. (c) Correct — Reliance builds, Meta leases 168 MW capacity. (d) Wrong — Reliance is the single-window solutions provider handling design, construction, and utilities.

3Which of the following correctly describes the difference between a GPU and a CPU in the context of AI data centres?

Correct: (b)

GPUs contain thousands of smaller cores designed for parallel processing — ideal for AI training and inference. CPUs have 8–64 cores optimized for sequential tasks. AI model training (like LLMs) requires massive parallel computation — hence GPU-centric AI data centres. CPUs still play a supporting role but GPUs are the primary AI accelerators.

4Consider the following pairs (Company / Initiative → Key Figure):
1. IndiaAI Mission budget → ₹10,372 crore over 5 years
2. Google's Visakhapatnam AI hub → $15 billion investment, 1 GW capacity
3. Meta's investment in Jio Platforms (2020) → $1 billion
4. Microsoft India commitment (by 2029) → $17.5 billion
Which of the above pairs are correctly matched?

Correct: (d) — 1, 2 and 4 are correct; Pair 3 is incorrect

Pair 1 ✅ — IndiaAI Mission = ₹10,372 crore, approved March 2024. Pair 2 ✅ — Google Vizag = $15 bn, 1 GW. Pair 3 ❌ — Meta's 2020 Jio investment was $5.7 billion, not $1 billion. Pair 4 ✅ — Microsoft = $17.5 bn by 2029.

5Power Usage Effectiveness (PUE) is a metric used in data centres. Which of the following statements about PUE is correct?

Correct: (c)

PUE = Total Facility Power ÷ IT Equipment Power (not the reverse — eliminates option b). Ideal = 1.0 (all power goes to IT, zero wasted on cooling/lighting). World-class DCs achieve 1.1–1.2. Higher PUE = MORE waste = LESS efficient (eliminates option d). PUE of 2.0 means half the power is wasted on overhead — very inefficient (eliminates option a).

MCQ pattern: statement-based (DPDPA), factual pair-matching (investments), technical distinction (GPU vs CPU), definitional (PUE). Memorise 5 key numbers: 168 MW, $5.7 bn, $15 bn, ₹10,372 cr, 38,000 GPUs.

Quick Revision — AI Data Centres & India Digital Infrastructure

⚡ Rapid Recall — AI Data Centres & India (Science & Technology · Prelims)

Meta-Reliance Jamnagar: 168 MW AI Data Centre announced June 10, 2026 — India's first built-to-suit AI-enabled DC. Reliance builds, Meta leases. Renewable energy + desalinated seawater cooling.
GPU vs CPU: GPUs = thousands of small parallel cores → AI training; CPUs = 8–64 powerful sequential cores → general computing. AI needs GPU.
PUE: Power Usage Effectiveness = Total Power ÷ IT Power. Ideal = 1.0; world-class = 1.1–1.2. Lower is better.
India's DC gap: India generates 20% of world's data but holds only 3% of global DC capacity (US = 53.7 GW, China = 31.9 GW, India = ~1.7 GW in 2025).
DPDPA 2023: India's first standalone data protection law; Presidential assent August 11, 2023; DPDP Rules notified November 13, 2025; uses blacklist model for cross-border transfers (not blanket localisation); first Indian law with she/her pronouns.
Infrastructure Status: DCs classified as infrastructure under 'Communication' category — October 2022 (DEA notification; based on TRAI recommendation).
IndiaAI Mission: ₹10,372 crore, approved March 2024; 38,000+ GPUs deployed; GPU rental ~₹65/hour; 600 data labs; Sarvam AI = 120B parameter Sovereign LLM.
Google Visakhapatnam: $15 billion, 5 years, 1 GW — largest AI hub outside US; partners: AdaniConneX + Nxtra by Airtel; announced October 2025.
Meta-Jio history: 2020 = $5.7 bn Jio Platforms investment → August 2025 = $100 mn JV (Llama AI) → June 2026 = 168 MW Jamnagar DC.
DC types: Colocation (multi-tenant, own servers) → Hyperscale (100+ MW, single tenant) → Built-to-Suit (custom-built, leased) → Edge (distributed, low-latency) → Sovereign (govt-owned).
Mumbai dominance: ~54% of India's DC capacity (JLL H1 2025). Vacancy rate = 4.3% — supply constrained. New hubs: Vizag, Hyderabad, Jamnagar.
Power & Environment: India's 9 GW DC target by 2030 = ~3% of India's total power. Drives renewable energy demand; sustainability mandates now standard for new DCs.

🎯 Meta-Reliance = 168 MW, Jamnagar, June 2026 · DPDPA = Blacklist model, not full localisation · India = 20% world data, 3% DC capacity

· MaargX UPSC · Curated for Civil Services Preparation ·