CoreWeave and AdaniConneX logos side by side on a light grey background — CoreWeave and AdaniConneX announced their India partnership on October 6, 2026. Image: CoreWeave.

India just got a big new name in AI cloud computing. CoreWeave, the US company that rents out GPU power to AI labs and enterprises, says it will build its first Indian data centers in Navi Mumbai.

The deal is with AdaniConneX, the data center joint venture of Adani Enterprises and EdgeConneX. CoreWeave plans to use 240 megawatts of capacity there, with an option to double it.

I'm Nitin from the Debugged Pro team. We build software and AI features for clients in India and the USA, so where GPUs sit matters to our budgets and our designs. Here is what was announced, what it does not change yet and how agencies can plan for it.

What CoreWeave Announced for AI Cloud in India

CoreWeave shared the details in its official announcement, dated October 6, 2026. The main points are simple:

Location: AdaniConneX's Taloja campus in Navi Mumbai.
Size: 240 MW across three buildings of 80 MW each.
Room to grow: CoreWeave can add up to another 240 MW at the same campus.
Tenancy: CoreWeave will be the only tenant in all three buildings.
Hardware: NVIDIA's Vera Rubin platform, for training, inference, reasoning and agent workloads.
Local team: An India office and local hiring.

Sachin Jain, CoreWeave's Chief Operating Officer, said: "India has the talent, developers, and ambition to become a major center of AI innovation."

The company also says the investment supports national goals, including the IndiaAI Mission, the government programme for AI compute, datasets and skills.

The timeline is the part to remember

The first phase is expected to be ready in mid-2028. More capacity will follow in phases after that.

So nothing changes for your client projects this quarter. This is a 2028 story with planning work that starts now.

How big is the money?

CoreWeave did not give a figure in its release. Bloomberg, in a report carried by NDTV Profit, said the investment is expected to total multiple billions of dollars over the life of the project.

Two server racks packed with GPU systems and yellow and blue network cables — GPU racks shown in CoreWeave's announcement of its NVIDIA Vera Rubin NVL72 deployment. Image: CoreWeave.

Why CoreWeave Picked India Now

CoreWeave has been expanding outside the US to add capacity and serve customers closer to home. In August it announced its first Asia projects, in Indonesia. India is the next step.

Demand is the other reason. As of its earnings call on August 11, 2026, CoreWeave had about 4.2 gigawatts of contracted power. Bloomberg reported that CEO Mike Intrator said last month the company is "largely sold out" of capacity.

India's own growth adds to the pull. Bloomberg cited a July report from Wood Mackenzie with these numbers:

India built roughly 2.2 GW of operational data center capacity over the past three decades.
It is expected to reach 12 GW by 2030.
AI-specific capacity alone is projected to grow almost 24-fold, to 6.5 GW.

AdaniConneX is also one of the companies working on a large India data center project for Google, according to the same report. The partner already has big AI builds in progress.

Bar chart of India data center capacity: 2.2 GW operational today, 12 GW expected by 2030 and 6.5 GW of AI-specific capacity projected — India's data center capacity outlook. Chart by Debugged Pro using Wood Mackenzie figures cited by Bloomberg.

What an AI Cloud Provider Like CoreWeave Actually Sells

Not every team has used a GPU-focused provider, so a quick explainer helps. AWS, Azure and Google Cloud sell almost everything, from email to databases. CoreWeave focuses on compute for AI.

Its pitch is GPU clusters, fast networking, storage tuned for AI and managed Kubernetes built for training and serving models. Big AI labs rent this capacity to train models. Enterprises and startups rent it to run inference, which is the work of answering user requests.

The Taloja buildings are being designed for dense AI hardware. CoreWeave lists direct liquid cooling for GPU halls and a non-evaporative chilled water system. That matters in a city where water and power are both under pressure.

What This AI Cloud Deal Means for Agencies in India and the USA

Most agencies will never rent 80 MW. But the effects reach smaller teams in a few practical ways.

1. Lower latency for Indian users

Today, many AI features built for Indian users call models hosted abroad. Every request makes a long round trip. Local GPU capacity can cut that delay for chat, voice and search features.

2. An easier data residency story

Some clients in finance, health and government want data and processing kept in India. More local AI compute gives them more choice. We covered how this plays out with model providers in our guide to AI data residency in India.

3. More pricing competition, later

More capacity usually means more pricing options. But do not promise clients cheaper GPU costs based on a 2028 opening. Prices depend on demand, and CoreWeave says it is already close to sold out.

4. More GPU and MLOps jobs in India

A local office means hiring. Agencies with engineers who know Kubernetes, inference tuning and GPU cost control will find more work as enterprises move AI projects into production.

A 6-Step AI Cloud Plan for Agencies

Here is what we suggest to clients and partners right now. None of it depends on the Navi Mumbai site being live.

Map where your AI calls go. List every model, API and region your client apps use today.
Measure latency and cost per request. You can't judge a new region without a baseline.
Keep workloads portable. Use containers, standard APIs and clean configs so you can move regions without a rewrite.
Tag data by sensitivity. Know which data must stay in India before a client asks.
Budget for inference, not just build. Running a model every day often costs more than the first build.
Review options every six months. Hyperscalers, GPU clouds and model APIs are all adding India capacity.

If you need help shaping that plan, our software development consulting team works on exactly these choices with clients.

A Simple Example From Agency Work

Picture a typical project we see often. A lending startup in Pune wants an AI assistant that reads loan documents and answers customer questions in English and Hindi.

Today, a team like ours would likely call a model API hosted outside India. That works, but it raises three questions from the client:

How long does each answer take for a user on a mobile network?
Where do the documents go while the model reads them?
What will this cost each month once 50,000 customers use it?

Local AI cloud capacity does not answer these on its own. It gives you one more option to test. The teams that already track latency, data flows and cost per request will be the ones who can switch quickly when it makes sense.

What US Clients Should Know

This is not only an India story. Many US companies run support, engineering or analytics teams in India. Many also hire Indian agencies to build their AI products.

For them, more AI compute in India could mean a few things over time:

Region choice: Products that serve both countries can run models closer to each user base.
Vendor reviews: Security and compliance teams will want to know which region processes which data.
Same provider, two regions: CoreWeave already runs sites in North America, Europe and Asia-Pacific. A team using it in the US may be able to use the same tools in India later.

Our advice is the same on both sides. Write down where data lives today, and keep that record current.

Questions to Ask Before Choosing an AI Cloud Region

When a client asks us to pick where an AI workload should run, we go through a short list:

Where are most of the end users?
Does any contract, regulator or client policy require data to stay in a country?
Is the workload training, batch processing or live inference?
How sensitive is the app to delay?
Can the app fail over to another region if one goes down?

The answers usually make the choice obvious. Live inference for Indian users with strict data rules points to local capacity. Batch jobs with no residency rules can run wherever GPUs are cheapest.

AI Cloud Options Compared: Today vs After 2028

Question	Today	After CoreWeave Taloja opens
Where do most AI models run for Indian users?	Often outside India, through global APIs	More local GPU capacity becomes available
Who can rent CoreWeave GPUs in India?	No local CoreWeave site yet	CoreWeave customers, once phases go live
What hardware?	Varies by provider	NVIDIA Vera Rubin, per CoreWeave
When?	Now	First phase expected mid-2028

What Is Still Unclear

The announcement leaves some open questions. Agencies should watch for answers before making promises.

Pricing: No India pricing has been shared.
Customer mix: CoreWeave has not said which customers will use the India capacity.
Smaller buyers: It is not clear how much capacity will be open to startups and mid-sized firms.
Power: Large data centers need steady power, and phased openings can slip.

Demand for AI inside Indian firms is still catching up, too. We looked at that gap in our post on AI adoption in India.

Key Takeaways

CoreWeave is entering India with 240 MW of AI cloud capacity at AdaniConneX's Taloja campus in Navi Mumbai.
The first phase is expected in mid-2028, with an option to double capacity later.
NVIDIA Vera Rubin hardware will power training, inference and AI agent workloads.
Bloomberg reports the investment will total multiple billions of dollars over the project's life.
Agencies should measure current AI costs, keep workloads portable and plan for data residency now.

About the Author

Nitin Kumar is the Lead Project Coordinator at Debugged Pro. He works with development teams and clients in India and the USA to plan software and AI projects, set delivery processes and keep quality high. He writes about practical AI adoption for agencies and growing businesses.