CoreWeave Plans 240MW Nvidia Cloud Expansion in India
CoreWeave plans to enter India with 240MW of data-centre capacity at AdaniConneX’s Taloja campus in Navi Mumbai, using Nvidia’s Vera Rubin platform for its AI cloud. The project comprises three 80MW buildings, with the first phase expected in mid-2028 and later phases following. CoreWeave would be the sole tenant of those buildings and has an option to add a further 240MW at the campus. The announcement creates a substantial future deployment route for Nvidia systems in a major regional market. Its commercial status remains a planned expansion, with the timing of operating capacity dependent on construction and commissioning rather than the announcement of the relationship alone.
The facilities are being designed around the requirements of dense AI computing. CoreWeave describes direct liquid cooling in the GPU halls and a non-evaporative chilled-water system, with provision for later generations of hardware. Those choices affect how much computing equipment can be supported within a building and how the site handles heat. They do not by themselves disclose the campus’s total electricity demand, water consumption or rack count. The companies have not provided enough information to turn the announced megawatts into a defensible number of Rubin processors. What they have identified is a physical deployment plan, a named infrastructure partner and the computing platform CoreWeave intends to make available to customers.
“India has the talent, developers, and ambition to become a major center of AI innovation,” CoreWeave chief operating officer Sachin Jain said. The company also plans an Indian office and local hiring, giving the expansion a commercial and service presence beyond the equipment installation. It positions the campus as supporting training, inference, reasoning and agent workloads, and links the investment to India’s wider AI ambitions. Local operations can help an infrastructure provider work with enterprise procurement teams, developers and counterparties in the market it serves. They do not establish a government procurement award: the announcement’s reference to the IndiaAI Mission is strategic context, rather than disclosure that the programme has contracted for this capacity.
AdaniConneX is a joint venture between Adani Enterprises and EdgeConneX, combining local infrastructure development with data-centre operating experience. Specialist reporting places Taloja within a wider Indian estate and development pipeline that includes other major cities. CoreWeave would bring its specialised AI cloud operations to that physical base. The division of roles is commercially useful: the cloud provider can expand through a local facility partner while concentrating on the systems and services customers consume. It also means successful delivery depends on coordination across organisations. Building readiness, electricity, cooling and the arrival of expensive computing equipment have to align before the cloud operator can turn committed site capacity into usable customer services.
The initial 240MW plan and the option for another 240MW should be kept separate. The first describes the announced campus deployment; the second preserves a route to expand if demand and delivery justify it. Exercising the full option would bring the combined capacity to 480MW at this campus. CoreWeave reported approximately 4.2GW of contracted power in August, offering a reference for the scale of the India project without proving that Taloja is incremental to that earlier total. Contracted power also differs from operating capacity. Across the portfolio, the relevant progression is from a site commitment through construction and equipment installation to reliable service that customers actually use and pay for.
Analysis
The 240MW plan is equivalent to about 5.7% of CoreWeave’s previously reported 4.2GW contracted-power base, a scale comparison rather than an addition to it. Nvidia gains a route to a geographically broader customer base through a provider already organised around its systems. The mid-2028 start places the opportunity well beyond an immediate sales catalyst and makes delivery sequencing economically important. The expansion option is valuable because it permits further growth at the same campus without requiring all 480MW to be built at once. Returns improve when additional capacity follows durable customer demand rather than preceding it by too much.