Vultr Launches 20 MW AI Supercompute Cluster In Chennai With 10,000 AMD GPUs

Vultr has launched a 20-megawatt AI supercompute cluster in Chennai, powered by more than 10,000 AMD Instinct MI355X GPUs and AMD EPYC processors.

The privately held cloud infrastructure company said the cluster is its second-largest supercompute deployment globally, after its facility in the United States. It is built for large-scale AI training, fine-tuning and inference workloads.

What 20 Megawatts Buys

For Vultr, power capacity is the number that matters in AI infrastructure is the number that matters. A 20-megawatt site can run thousands of accelerators continuously. That is the unit in which cloud providers now plan.

Read More: IISc Spin-Out Venomics AI Raises $2.7 Mn To Turn Venom Into Drugs

Start UpMylapay Bags $1 Million Funding to Scale Unified Payment Infrastructure for Banks and Fint…Start UpUS-based VideoDB acquires Chennai’s AI testing startup Devzery

More than 10,000 GPUs in one location makes Chennai one of the larger AI compute sites in India. Vultr says it plans to bring successive generations of GPUs and CPUs into the country.

Read More: Desiya Suraksha Party Unveils National Mission for Safety and Citizens’ Welfare

The choice of AMD hardware is also notable. Vultr has built much of its AI story on AMD Instinct accelerators. Gives customers an alternative to the Nvidia supply queue.

Why Chennai

Kevin Cochrane, Vultr’s chief marketing officer, framed the decision around the market itself. India is not just consuming AI, the buildout of AI is happening here.

That is partly a talent argument and partly a latency argument. Indian developers building for Indian users need compute close by. Training runs that involve large local datasets are cheaper to run where the data sits.

Data residency rules are the third factor. Enterprises in regulated sectors increasingly want workloads on infrastructure they can point to, inside the country.

The StartupTN Deal

The launch came with a partnership between Vultr India and StartupTN, the Tamil Nadu Startup and Innovation Mission.

Under it, eligible startups in the state get cloud credits and preferential pricing. Credits are the cheapest form of customer acquisition in this market, because a founder who trains a first model on a platform tends to stay.

It also gives the state a concrete offering beyond incubators and office space. Is what most startup missions actually provide.

The Competition for AI Compute in India

Power availability is the real constraint. A 20-megawatt cluster needs a substation and a cooling plant as much as it needs GPUs. Tamil Nadu has been courting data centre investment for exactly this reason.

Vultr is not alone in this buildout. Global clouds, domestic data centre operators and GPU resellers have all announced India capacity in the past two years. Several Indian conglomerates have committed capital to AI data centres.

What differentiates a provider now is not the size of the announcement but three things: how fast capacity becomes usable, what it costs per GPU-hour. Whether the software layer makes deployment easy.

Vultr’s own pitch has been simplicity and pricing rather than enterprise services. Appeals to developers and smaller AI companies more than to large IT departments.

What to Watch

Sovereign AI is the other thread here. Governments want domestic compute for public sector models. A provider with local capacity is better placed to win those contracts than one serving India from Singapore.

Three things will decide whether the Chennai cluster matters. The first is utilisation. Announced capacity is meaningless if the GPUs idle, and utilisation data rarely gets published.

The second is pricing in rupees. Indian AI startups compare local GPU-hour rates against overseas clouds. A 10% difference changes where they build.

The third is the next generation. Vultr has promised successive hardware generations in India; whether the newer chips land here quickly will show if Chennai is a strategic site or a single deployment.