Nvidia After Jensen Huang. Are India’s AI Chip Deals on Track?

Jensen Huang’s exit from Nvidia raises a practical question for Indian companies building AI data centres. Will they get the chips they were counting on, and when will customers be able to use them? Nvidia named Colette Kress interim CEO on 12 November. Its announcement gave no revised schedules for its India projects. One deployment offers a useful place to start. Yotta and its partner had already set out different delivery targets for a 20,736-chip project before Huang resigned.

Has Nvidia changed its India plans since Huang resigned?

Nvidia’s transition announcement did not identify any change to those plans. The documents available on Yotta and Reliance set out partnerships, proposed capacity and milestones. They give us dates to test. The clearest proof will be a company confirming how many machines have arrived, how many are operational and what customers can book.

What did Yotta actually announce?

In February, Yotta said it planned to deploy 20,736 Nvidia Blackwell Ultra GPUs in a single computing cluster, with investment exceeding $2 billion. A GPU is a processor that can do many calculations at once, which makes it useful for training and running AI. Yotta said Nvidia’s DGX Cloud business would use part of the cluster under a four-year engagement worth over $1 billion. The investment figure describes the deployment. The engagement figure describes Nvidia’s cloud relationship with Yotta.

Where did the timetable move?

Yotta’s February announcement expected the cluster to go live by August. In May, Yotta and infrastructure partner Gorilla Technology described the 20,736-chip expansion with an expected delivery date of 30 September. Later, Gorilla CEO Jay Chandan told Zacks Small Cap Research that first deliveries for the project were expected in mid-October and that the companies aimed to deliver and deploy all the chips by the end of November. These updates use different milestones. By the time Huang resigned, the public timetable already pointed beyond the original August target.

Can we say how many of those 20,736 chips are running?

The sources above establish targets and work in progress. A September Gorilla announcement said equipment deliveries had begun across its Yotta programmes. Chandan subsequently distinguished the roughly 5,120 GPUs in the first Yotta project, where he said testing and installation were under way, from the 20,736-chip second project. A number for equipment across several projects cannot establish how many chips from this particular batch are available to paying customers.

Why does Nvidia matter as both supplier and customer?

Yotta is building its service with Nvidia hardware. Nvidia, through DGX Cloud, is also expected to use capacity at the facility. Gorilla said the four-year Nvidia engagement was expected to account for roughly half the computing capacity sold under that tranche. That gives the project a large customer from the start. It also makes it important to establish which commitments the parties signed, when payments begin and how much capacity is available for everyone else.

What have Reliance and Jio committed to?

Nvidia and Reliance announced an AI infrastructure partnership in 2023, with Jio involved in developing cloud infrastructure. In June 2026, Reliance told shareholders that Reliance Intelligence was building AI infrastructure in Jamnagar, Gujarat. It aimed to commission its first 120 megawatts by the end of 2026 and said it was bringing an initial fleet of Nvidia GB300 chips into operation. Megawatts measure electrical power. The 120-megawatt figure tells you the scale of the planned facility. A count of machines running tells you what customers can use.

What does a delay mean for an Indian AI developer?

A team training a model needs access to a block of machines for a particular stretch of time. An installed chip still needs working power, cooling and networking before it can be rented out. A delayed cluster can mean rescheduling the run or paying for capacity elsewhere. The government’s IndiaAI Compute Portal lists different chip configurations and hourly rates. Those listings help teams compare options. To know when they can run a job, teams still need a supplier to confirm availability.

Which dates can tell us if the deals are on track?

Gorilla’s stated end-November target is the next checkpoint for the 20,736-chip batch. Reliance’s end-December target is the next one for Jamnagar’s first 120 megawatts. For each, ask for three numbers. How much equipment has arrived? How much has passed testing and is running? How much can a customer book? A revised date from either company would tell us more about its project than the handover announcement from Nvidia.