Vultr targets March launch for Chennai’s 20 MW AI cluster
Domestic computing access could reduce reliance on scarce overseas capacity
Overview
Vultr is building a 20 in Chennai with about 10,000 . Indian labs, startups, companies and researchers could rent this computing power instead of buying their own machines. The site is meant to , improve existing ones and . Vultr aims to start operations in March 2027.
The project supports India’s effort to expand local AI computing as demand rises and powerful chips remain scarce. Vultr targets 600 of Indian capacity by 2028 and 1 GW by 2029. These are goals, not working computing power yet. Meeting them will require electricity, cooling, fast networks, suitable sites, construction and GPU deliveries. The main test is whether customers can rent working systems at a price they can afford, and whether Chennai opens on time.
Timeline
Vultr announced a 20 MW Chennai AI cluster and longer-term India capacity targets
Vultr announced a 20 MW Chennai cluster expected to house about 10,000 , with a March 2027 operational target. It also outlined targets of 600 MW of Indian AI infrastructure by 2028 and 1 GW by 2029; these figures describe planned expansion rather than demonstrated customer-accessible capacity.
Vultr’s Chennai cluster added planned commercial capacity to India’s AI infrastructure
Vultr said its planned would provide AI compute services for Indian enterprises, startups, developers and research users. The announcement was part of a broader expansion of commercial AI infrastructure, but does not by itself establish convergence with public compute programmes.
Vultr targeted 1 GW of Indian AI capacity by 2029 and began evaluating additional sites
Vultr said it aimed to reach in India by 2029, with capacity spread across locations and additional sites under evaluation. The plan raises execution questions around secured power, land, permits, construction progress, . Progress toward the target will be measurable only through disclosed sites, commissioned megawatts, deployed GPUs and customer-usable capacity.
Vultr targeted 600 MW of Indian AI infrastructure by 2028
Vultr said it was targeting 600 in India by 2028. Reaching that figure would require substantially more than adding GPUs, including multiple suitable sites, grid connections, high-density cooling, internal networking, construction and financing. The target is announced infrastructure capacity and should not be treated as usable compute until sites are built and services are available.
Vultr said Chennai would support training, fine-tuning and inference workloads
The company said the cluster would support and foundation-model training, fine-tuning, and high-throughput inference. Training uses large to create models, fine-tuning adapts existing models to specific data or tasks, and inference runs those models for users. The distinction matters because each workload imposes different requirements for GPU memory, networking, storage, scheduling and response time.
Vultr targeted March 2027 to operationalise Chennai’s 20 MW AI cluster
Vultr said it was targeting March 2027 to the 20 MW Chennai AI cluster. The date was a company target, not confirmation that the facility had completed commissioning or was already available for customer workloads.
Vultr positioned Chennai to serve AI labs, startups, enterprises and research institutions
Vultr said the cluster would provide cloud access to AI labs, , AI-native startups, enterprises, and academic and research institutions. The design covers training new models, existing models, and high-throughput after models are deployed. Renting access through the cloud could reduce the need for customers to buy and operate their own GPU clusters, but pricing, availability and service performance remain unreported.
Vultr said the Chennai cluster would deploy about 10,000 AMD MI355X GPUs
Vultr said the initial Chennai configuration would contain about 10,000 GPUs, making it India’s largest AMD GPU deployment. Each MI355X has 288 GB of memory and 8 TB/s of memory bandwidth, specifications intended for large-model workloads. The announced hardware scale is not the same as operational customer capacity until installation, commissioning and service availability are demonstrated.
Vultr announced a Chennai cloud region and a 20 MW AMD AI supercluster
Vultr announced Chennai as its 34th and said the cluster would begin with 20 MW of . The company described the deployment as one of the world’s largest AMD AI superclusters and said further expansion was planned for 2027. The announcement established the project as a commercial cloud service, not merely a privately operated research installation.