For Indian AI builders, the decisive milestone will be available capacity with clear commercial terms, rather than the headline number alone.
The announced plan
CoreWeave announced its India entry on October 6 through a collaboration with AdaniConneX. It plans 240 MW of data-centre capacity across three 80 MW buildings at the Taloja campus in Navi Mumbai. The announcement identifies NVIDIA Vera Rubin as the planned platform and expects the first phase in mid-2028.
CoreWeave says it will be the sole tenant and has an option for another 240 MW. That option should not be counted as committed capacity. The company also plans an Indian office and local hiring. These are forward-looking commitments; the announcement does not establish that the campus is already operating.
Foundra analysis: megawatts are not a booking
For an AI startup, the practical unit is a workload that can run at a known price and service level. Electrical capacity is an important infrastructure measure, but it does not specify the number of accelerators available to a particular customer, the timing of deployment or the economics of a training run.
The gap between announcement and service is therefore worth watching. A campus needs power, cooling, installed hardware and operational readiness. Customers also need contracts, support and a way to move data without introducing expensive bottlenecks. A local facility may create options, but it is not by itself evidence of cheaper compute for every Indian startup.
What to watch next
Plan
240 MW announced; another 240 MW is an option.
Timing
First phase expected in mid-2028.
Customer test
Track usable capacity, service terms and workload economics.
What builders should track
Founders with immediate workloads should plan against capacity that is available now. A future campus belongs in a longer-term supplier assessment, with milestones for installation, customer availability and service commitments. Conflating the two could turn a promising infrastructure development into a fragile operating assumption.
The more interesting question is whether this investment eventually supports a wider set of buyers beyond the largest training customers. Watch for published service availability, transparent deployment dates and workloads that benefit from proximity. Also distinguish a customer's right to use a particular facility from a broader promise about data location. Those details will determine how useful the expansion is to businesses with specific operational constraints. The 2028 clock gives the market a clear point of reference, while leaving plenty of execution still to demonstrate.
Sources
- CoreWeave announcement ↗ CoreWeave · 2026-10-06 · Primary source
Have additional information? Request a correction.
