Marvell has one-fourth of its workforce in India: Country head Bishnoi

Marvell's India country head on how India has become the company's second-largest R&D hub, contributing to 90% of product lines and driving custom silicon for AI infrastructure

Marvell Technology's India operations have quietly become a cornerstone of the $70 billion semiconductor giant's global R&D strategy. In two decades, the company has scaled its Indian workforce to nearly 1,700 employees—one-fourth of its global engineering headcount—contributing to more than 90 per cent of Marvell's product lines, including AI infrastructure chips powering hyperscale data centres. In a video interview with Peerzada Abrar, Navin Bishnoi, India country head and AVP of data centre engineering, discusses how India has evolved from a support centre to a product-creation hub, the challenges of building advanced AI chips, and why Marvell sees India as critical to competing with giants like Nvidia and Broadcom in the custom silicon race.

 

Marvell has been in India for about two decades. What's the scale of your current operations here—headcount, investment levels—and what are your expansion plans over the next 2-3 years?

 

In January, we'll celebrate the 20th anniversary of Marvell India. During our global pivot from a “do everything” company to an enterprise company, India gained strong sponsorship.

 

Back in 2017, we were fewer than 100 people. Today, we have close to 1,700 regular employees, and including partner resources, our workforce crosses 2,000—for a global company of about 7,000 employees. India has quickly become almost one-fourth of the company’s workforce and drives R&D across every product and technology.

 

Some product lines are completely designed, developed, and delivered from India for global markets. Over the next three years, we aim to continue this growth. This includes advancing competencies, scaling teams, and creating products for global markets. We have grown at an average of 15 per cent annually and expect this momentum to continue. Our Bengaluru and Hyderabad offices will also expand next year. We are also exploring how our products can address Indian consumer use cases and how India can further strengthen its role as a creation hub for global semiconductor needs. For the next three years, we expect annual hiring of over 200 people, taking our India headcount from 1,600–1,800 to well beyond 2,000—maintaining our position as more than one-fourth of Marvell’s engineering workforce. Our annual investment in India has already reached about $300 million, covering headcount, advanced R&D, state-of-the-art labs, tools, and product development. This will continue to grow as we scale.

 

What percentage of Marvell's global R&D and product innovation now comes from your India GCC? Can you quantify India's contribution—whether in terms of patents filed, chips designed, or revenue generated from India-origin IP?

 

India is our largest R&D centre outside the US, accounting for one-fourth of Marvell's global engineering workforce. With $8 billion in revenue and 7,000 employees globally, India contributes significantly to our $1 million revenue-per-employee metric.

 

In terms of product contribution, India plays a role in more than 90 per cent of Marvell’s product lines, with some products entirely owned and built from India. For example, security processors for data centres, including both hardware and software, are developed completely in India. India’s impact spans security, fibre channel, wireless, communications, and several other key domains.

 

How does India compare to other global R&D hubs like the US or Israel?

 

Security is a major example, and another is wireless—particularly for carrier networks across 4G, 5G, and 6G. Marvell supplies key DPU (data processing unit) technology for these infrastructures, and a significant portion of both hardware and software development takes place in India.

 

The US is our largest centre, followed by India. Other centres are significantly smaller. While other global centres focus on specific product lines, India contributes across all products and technologies.

 

Are you finding the skills you need domestically, or are you having to upskill aggressively, especially in areas such as AI and chip design?

 

India has a strong talent foundation, and for decades, around 20–25 per cent of global chip designers have been Indians. However, three areas are now evolving. The first is global collaboration skills, since chip design is inherently global and requires teams to work seamlessly across geographies and communicate effectively.

 

The second is AI-enabled engineering—AI is transforming workflows from coding to documentation. The third is industry–academia collaboration, which we support through curriculum feedback, faculty internships, and joint research projects.

 

US-China tensions have reshaped semiconductor supply chains, with export controls on advanced chips. How is this affecting Marvell's global operations and India strategy specifically?

 

In general, semiconductors are not heavily affected by current tariffs. Tariffs typically apply to products physically manufactured in a country and stamped “Made in India.” Most of us—including Marvell—operate as R&D centres. Manufacturing happens in Taiwan, Korea, the US, or elsewhere, so it doesn’t carry a “Made in India” stamp. Semiconductors are outside tariff structures for now.

 

So, it doesn’t impact us today; we continue to execute and grow as planned. But yes, we monitor announcements closely to see if anything could affect us in the future.

 

Is India becoming a 'China+1' alternative for design and manufacturing, and are you seeing customer demand shift geographically?

 

Yes, the shift is already happening. That’s why so many investments—both in design and manufacturing—are coming to India. It started as a China-plus-one strategy but has expanded beyond that.

 

It’s not just about de-risking supply chains but also about leveraging the massive consumer opportunity in India. This trend will continue to grow. Semiconductor development, however, is a long-term game. R&D investment is huge, and technologies take years and decades to mature.

 

India is currently building foundational blocks—technologies that are relevant, affordable, and have faster ROI. Those are the priorities. Over time, more pieces of the ecosystem will fall into place.

 

India has ambitious semiconductor manufacturing plans, but the ecosystem—from materials to packaging—is still nascent. How is Marvell looking to participate?

 

India’s semiconductor plans currently focus on mature, legacy technology nodes. The most advanced node planned is 28 nm, which is still not advanced enough for mobile processors or AI infrastructure—those are several generations ahead.

 

But 28 nm fits well for industrial automation, home automation, automotive supply chains, mission-critical systems like satellites and defence, and more. These sectors don’t need 3 nm or 5 nm. So, India’s investment is correct—smaller, faster ROI, and commercially viable.

 

For Marvell, which works on advanced AI-infrastructure-grade technology, there is a mismatch today. So, we’re observing. If advanced-node capabilities come to India in the future, we will engage.

 

Regarding the India semiconductor mission, we are evaluating which ecosystem players can support advanced-node manufacturing for AI infrastructure and how we can integrate with their supply chains.

 

AI workloads are fundamentally different from traditional computing. How is this reshaping chip architecture priorities? What's the lifespan of current AI chip designs before they become obsolete, and how is Marvell staying ahead?

 

We build chips for AI powering large language models with trillions of parameters. These chips use 2 nm and 3 nm technologies, carry 200–300 billion transistors, consume 700–1,000 watts, and cost $10,000–$30,000.

 

Previously, chip design cycles took 24 to 36 months, but these timelines now need to shrink significantly. AI supports this acceleration by helping optimise power, improve performance, reduce cost, and shorten overall design cycles. Chip life cycles remain long—typically between three and 10 years depending on the segment, such as compute or networking. Data centres continue to add new GPUs and XPUs without discarding older ones, which means demand only keeps increasing.

 

Marvell competes with giants like Broadcom, Intel, and Nvidia in data centre chips and hyperscalers like AWS and Google designing more custom silicon in-house. How are you differentiating in an increasingly crowded AI chip market?

 

Companies like Nvidia, AMD, and Intel build GPUs and related chips and sell complete racks to hyperscalers. But hyperscalers now want to optimise costs because today’s AI infrastructure is predominantly an investment—very few end users are directly paying for AI services. To recover their massive investments and reduce infrastructure expenses, hyperscalers must customise their compute environments.

 

AI infrastructure is becoming highly customisable. OTT platforms, e-commerce, search engines, and enterprise applications all have fundamentally different compute requirements.

 

These companies, however, are primarily software-driven. While they maintain internal hardware teams, hardware is not their core business. Chip development demands world-class talent, costly EDA (electronic design automation) tools, extensive IP libraries, deep supply-chain partnerships, and significant capital investment. Marvell, by contrast, builds dozens of chips every year and has strong partnerships and foundational IP—advantages hyperscalers simply cannot match when they build only a handful of chips annually.

 

As a result, hyperscalers focus on developing only their “secret sauce” internally while partnering with custom-silicon companies like Marvell for full-chip development. They then brand, package, and deploy the final silicon as their own product. This collaborative, co-creation model is the way forward for the industry.