Neysa says its AMD-powered AI cluster in Mumbai is now live, with global customers already having contracted capacity. The November 18 announcement marks an operating milestone rather than another proposed data-centre project. It does not, however, disclose how much of that capacity customers are actually using.
The Breakdown
In its November 18 release, Neysa says the cluster was brought online in a single phase on Tytan, its dedicated bare-metal infrastructure offering. Bare metal means dedicated computing hardware rather than simply access to a managed application. The system uses AMD Instinct MI350X accelerators, EPYC processors and the ROCm software ecosystem.
The company describes support for training, fine-tuning and inference: building an AI model, adapting it and running it to produce responses. It says customers can select architecture and operating requirements, while Neysa provides the deployment and operating layer. This differs from its Velocis managed AI cloud offering.
Why This Matters
For a business buying dedicated computing infrastructure, the procurement question extends beyond the processor model. Network and storage design, software compatibility, service commitments and operating responsibilities determine whether the system fits the intended workload. Those are evaluation questions, not evidence that one provider delivers better performance or lower costs.
A live cluster and customer contracts are meaningful but different milestones. Deployment concerns infrastructure readiness; a contract concerns a commercial commitment. Neither alone measures usage, completed computing work or profitability. The release does not identify the customers or disclose contract values, pricing or utilisation figures. It also does not establish that the entire cluster is available for immediate on-demand rental.
The Big Picture
This Mumbai announcement should not be mistaken for completion of Neysa’s separate planned Noida deployment. Our earlier coverage of STTGDC’s Noida facility described a future GPU and Velocis deployment expected from the first quarter of 2027. The projects have different locations and reported delivery stages.
Likewise, the Andhra Pradesh power and compute proposals remain separate from this Mumbai operating announcement. Their planned capacity should not be added to or treated as delivered by this cluster.
The Crunch
The new evidence is Neysa’s report of live, contracted infrastructure in Mumbai. Buyers still need workload-specific terms and testing; readers should look for subsequent evidence of usage and service delivery. The announcement supports a deployment story, not a claim that utilisation or commercial returns have already been demonstrated.






Leave your thoughts