Yotta Data Services Raises $150M at $3.9B Valuation to Build India's Largest AI Compute Platform

Yotta Data Services, a subsidiary of India’s Hiranandani group, has raised approximately $150 million at a $3.9 billion valuation (₹37,000 crore) from non-institutional investors, according to the Economic Times. The entire proceeds are being injected into the business to support growth in AI and cloud operations.

The capital raise positions Yotta to establish India as a producer of AI infrastructure and intelligence, not only a consumer of it.

Where the Money Goes

Yotta is accelerating expansion across four verticals: AI infrastructure, sovereign cloud, managed services, and data center operations, ET reports. The company is also engaging with global AI model developers and inference providers, signaling a play for hosting foundation model training and serving workloads.

The expansion is expected to position Yotta among the world’s largest AI compute platform operators outside the US and China, creating a third pole for GPU-dense infrastructure in a market where data residency and sovereign compute requirements are becoming non-negotiable for governments and regulated enterprises.

A Third Pole for AI Compute

AI compute capacity is overwhelmingly concentrated in the US and China. For agent deployments and model inference serving Indian enterprises or regulated industries with data localization requirements, that concentration creates latency, compliance, and dependency problems.

Yotta’s $3.9 billion valuation reflects the market’s bet that sovereign AI compute outside these two poles will command premium pricing. For agent builders and AI model providers looking for non-US, non-China deployment options, India’s infrastructure buildout is creating capacity that did not exist two years ago.

The question is execution speed. The global AI infrastructure buildout is capital-intensive and hardware-constrained. Whether $150 million is enough to compete at the scale Yotta is targeting depends on GPU procurement timelines, power availability, and whether global model providers choose to deploy training runs in India rather than just inference endpoints.