Nvidia’s ₹8,000-crore move: one brain, one team, one bet on the future

Let’s say it plainly. Nvidia just put down roughly ₹8,000 crore to grab a person, his team, and the blueprint to wire the world’s biggest AI clusters so they don’t choke on their own size. The person is Rochan Sankar. The blueprint is Enfabrica. The goal is ugly simple: make tens of thousands of AI chips act like one giant mind, not a thousand expensive islands waiting on data.

Who is Rochan Sankar?

Indian-origin, global operator. Electrical Engineering from the University of Toronto, MBA from Wharton. Two and a half decades in the trenches of data-center networking. He led Broadcom’s data-center Ethernet switching business — the Tomahawk/Trident chips that fed the early cloud.

He’s not a headline-chaser; he’s the guy who makes packets move faster. His resume carries multiple patents in networking and systems. He knows how the plumbing of the digital world works — and how to fix the leaks.

What did Enfabrica actually build?

Enfabrica didn’t make another GPU. They built the fabric that lets GPUs breathe.

• Their technology connects a massive number of AI chips together without slowing them down.
• It makes sure expensive hardware doesn’t sit idle, waiting for data.
• It allows up to 100,000+ chips to function like one unified supercomputer.

Two big innovations stand out:

1. SuperNIC (Accelerated Compute Fabric – SuperNIC): a powerful chip that moves data across clusters at lightning speed.
2. EMFASYS (Elastic Memory Fabric System): a memory system that saves costs by pooling and virtualizing memory across chips.

In plain English: Enfabrica built the plumbing that keeps the world’s biggest AI brains alive and firing.

What exactly did Nvidia do?

This isn’t the usual “buy the company and slap on a new logo” move. Nvidia:

• Hired Rochan Sankar and his top people.
• Licensed Enfabrica’s technology.
• Paid out over $900 million (around ₹8,000 crore) in cash and stock.

This was a talent grab plus tech license, designed to save time. In AI, a single microsecond delay at scale costs millions. Nvidia didn’t want to wait — they absorbed the people and the know-how to build faster now.

Where is Enfabrica based? Do they have people in India?

Enfabrica’s headquarters is in Mountain View, California — right in the heart of Silicon Valley.

But India is not absent from the story. They have engineering roles in Hyderabad and Bengaluru, mostly in firmware and chip design. That means while the brain trust was in the US, India was already part of the engineering muscle.

How many employees?

Enfabrica wasn’t a giant. Public data puts them between 51–200 employees, with estimates around 160 people at the time of the deal. This is typical for deep-tech hardware startups: fewer people, but each one a heavyweight in their field.

What’s the status now?

As of late September 2025:

• Rochan Sankar is now at Nvidia.
• Nvidia has licensed Enfabrica’s tech and absorbed staff.
• Whether the Enfabrica brand continues or fades into Nvidia’s stack is still unclear.
• But the direction is obvious: Nvidia will fold this “fabric” into their GPUs, switches, and data-center systems.

Why this matters

AI clusters today waste billions of dollars’ worth of silicon because chips often sit idle, waiting for data. Enfabrica’s fabric fixes that.

Nvidia’s moat is no longer just GPUs. It’s GPUs + NVLink + NVSwitch + high-speed networking + memory fabrics + software orchestration. By bringing Enfabrica inside, Nvidia just tightened its chokehold on the future of AI computing.

The Indian angle

Here’s the blunt truth: this is proof that skill beats postcode. An Indian-origin engineer — with global exposure, deep knowledge, and zero interest in hype — became the pivot for a multibillion-dollar corporate chess move.

For India, the message is clear:

• Stop idolizing “apps.” Fund hard tech.
• Build labs in Bengaluru and Hyderabad where engineers can create infrastructure breakthroughs, not just outsource code.
• Support the ones who move electrons smarter, not just those who market louder.

When India backs compute fabrics, chip design, and system plumbing, we’ll stop exporting genius and start keeping the gravity here.

The man vs the myth

Don’t turn Rochan Sankar into a saint. He’s not here for slogans. He built things. From leading Broadcom projects, to holding patents, to co-founding Enfabrica, to selling brains and blueprints to Nvidia — it’s a straight line of execution.

That’s the lesson: not noise, not PR. Execution.

What happens next?

• Nvidia’s future data centers will feel more like one single machine than thousands of loosely linked GPUs.
• Customers will get faster training, cheaper per-token costs, and deeper lock-in to Nvidia’s ecosystem.
• Rivals like AMD, Intel, and Google will scramble to build or buy their own “Enfabricas.”
• Expect more billion-dollar “talent grabs” in the networking and chip infrastructure space.

Nishani Verdict

This wasn’t about a logo or a startup exit. It was about buying time and buying certainty. Nvidia paid ₹8,000 crore because the stakes are cosmic: whoever controls the fabric controls the pace of AI.

And here’s the punchline: Nvidia didn’t just buy tech. They bought the people who actually know where the wires break — and how to keep them alive when 100,000 hungry chips are screaming for bytes.

If you’re an Indian founder reading this: stop pitching “the Uber of X.” Start pitching the Enfabrica of Y. The deep, boring, beautiful plumbing of technology is where the next ₹8,000-crore moves will be made.

Comments

comments