CoreWeave's 240 MW India plan sets a 2028 horizon for AI cloud buyers

CoreWeave's planned arrival in India gives AI teams a future infrastructure option to evaluate, with the first phase expected in mid-2028. It does not provide capacity for a deployment that must start today. The company intends to use 240 megawatts at AdaniConneX's Taloja campus in Navi Mumbai.

The Next Web reported the expansion on October 7, 2026. CoreWeave's investor release is dated October 7, while its newsroom copy carries October 6. Both describe a planned buildout, rather than an operating Indian cloud region.

Three buildings establish the scale of the plan

CoreWeave says it will occupy three 80 MW buildings as the sole tenant and plans to deploy NVIDIA's Vera Rubin platform. Later phases would follow the first opening. An option for another 240 MW leaves room to expand; it is not a commitment that the larger campus will be delivered.

Megawatts describe the project's electrical capacity, not a count of GPUs or an amount of computing a customer can reserve. Turning that headline into a deployment plan requires a service specification. Neither the power figure nor a named chip platform establishes application throughput, price or available cluster size.

A local campus still needs a cloud-service specification

The design includes direct liquid cooling for GPU halls and a non-evaporative chilled-water system, according to CoreWeave. These are infrastructure choices; the announcement supplies no operating measurements from this site.

For customers, the next useful details are the GPU instances and storage services offered at launch, network connections and the failure domains available for recovery. Three buildings alone do not tell a buyer how an application will be distributed or protected.

CoreWeave's Trust Center says customers should discuss regional availability and placement during sales and onboarding, and that detailed compliance evidence is available on request. That makes a specific deployment agreement more useful than assuming that a site in India automatically answers every data-handling requirement.

Use the lead time to define the workload

Our assessment is that teams with a 2028 planning horizon can use this announcement to start a supplier conversation. They should describe the workload they need to run, ask what can be reserved in the first phase and separate promised launch services from later expansion. A project with an earlier deadline still needs an available deployment path.

The broader procurement question is how future hardware, financing and service access fit together. Our analysis of Broadcom's reported Anthropic financing facility explains a different example of that relationship. Neither arrangement, by itself, demonstrates the performance of a customer's application.

Sources

• CoreWeave: investor announcement, October 7

• CoreWeave: newsroom version, October 6

• The Next Web: India expansion report, October 7

• CoreWeave: Trust Center