Joined the US-based chipmaker about a decade ago, Mark Papermaster, Chief Technology Officer and Executive Vice-President (Technology and Engineering) of AMD, has led the re-design of engineering processes and turned around AMD’s strategies.

In an interview with BusinessLine , Mark speaks on the broad contours of the semiconductor industry, supply chain issues that the industry is facing and newer challenges and opportunities for the industry.

It is almost 10 years since you took over as CTO of AMD, and it also, incidentally, coincides with AMD’s turnaround story. What are the key milestones in this turnaround and what is the road ahead?

What I really brought to AMD 10 years ago was a vision that we could put these building blocks, i.e., the CPU and GPU, together with a modular approach to bring efficiency in delivering high performance to the industry. We set about a roadmap that would be multi-generational, and not just one big splash. It was about creating a family of generations for CPU, generations of GPU.

Also read: India awaits first investment in semiconductor space even as Intel, Samsung set up chip factories in the US

It was about creating a roadmap and changing the engineering processes with AMD to become a dependable supplier to the industry. That was the vision and as a team, AMD has really come together and executed extremely well on this vision.

An important phenomenon is happening in the semiconductor space, where top consumers of chips are building their own. How do you think this will impact established players like you, and is it going to be a challenge for you, or do you see an opportunity there?

That is a great question. The explosion of demand for compute capacity is so high that it is spurring unique solutions. Rather than one type of computing for all the problems out there, the demand is creating room for innovation and targeted optimised solutions. So, we do expect some of our larger customers to want to tailor their own devices. Even as we grow our roadmap in terms of our compute capabilities for each generation, we do think that in the long term, there is a strong market for these types of customised designs that some of our clients will need.

We see strong future growth opportunities in our traditional markets across our CPU and GPU, but also for tailored specialised compute.

So, for them, you’ll have this opportunity of building semi customised chips..?

When a customer with a large demand would like a tailored solution, we have historically been able to work with these customers to build them tailored chips. Historically, we have done this with traditional chip designs, like the work that we did with Microsoft on Xbox and with Sony on PlayStation.

It will be less expensive for our customers to tailor a design for their specific needs by working with us.

At present, the semiconductor industry and all the industries that are dependent on chips are facing a major supply problem. The automobile companies are facing huge losses and they have cut down on production. When do you think normalcy will come back?

The automotive industry does not use the newest semiconductor foundry nodes, but typically those that are well established, those that have been around already for several generations.

We are not using the same semiconductor manufacturing lines at AMD as the automotive manufacturers. We use cutting-edge foundry semiconductor nodes. So, we have not had the same shortfall that the automotive industry has had with legacy nodes.

Also read: Semi-conductor shortage impacts premium two-wheeler sales

What we have found is that by establishing a deep partnership with our foundry suppliers and combining that with a long lead time forecasting capabilities, we have been able to inform our suppliers, well in advance, of the need for increased capacity. This has served us well at AMD.

In 2021, we have projected a 65 per cent revenue growth over 2020. We can achieve such growth, only because of the excellent work on our supply chain side.

Can you throw a bit of light on how the overall situation will pan out in the next three to four quarters? When can we expect normalcy, if not for AMD, the overall industry chip supply?

In our markets, we see an easing of supply in the second half of 2022 through 2023. That is when we are projecting the normalcy of supply and demand balance.

What are the broad contours of the current semiconductor space. Can you give me a broad overview?

Let me start with a view from AMD. Some time ago, we had set the goal to become leaders in High-Performance Computing (HPC). To do this, we revamped our CPU roadmap and, introduced the first of our Zen family processors in 2017.

Recently, we shared details about our fourth-generation Zen processors. We also announced new innovations in our third generation of Zen processors.

We are going to use advanced techniques of 3D packaging to further extend the performance of the third-generation AMD EPYC processors. The need and demand for computing is skyrocketing, because we have so much data that is being generated around.

AMD has come a long way from facing crises, facing competition from industry biggies like Intel to revenue growths of over 50 per cent year on year. Do you think this expansion is overheating the system? Is this growth sustainable?

Well, we have established excellent momentum in the industry on the strength of our products. Our products have leadership performance. The total cost of ownership, the performance that we deliver for the cost and the energy of that product are of outstanding value to our customers.

We have always expected our competitors to respond. We believe that the best defence is a strong offence.

We have made significant share gains, but there is much more in terms of the market opportunity for AMD to grow across our traditional segments and the expanding opportunity of artificial intelligence. AI will drive an expansion in the market for high-performance computing and we are very well positioned to supply the computing capabilities that will fuel the growth of AI.

Over the past one or two decades, the semiconductor industry, has followed a path of ups and downs. The cycle goes up and then hits a slump. Do you think that this has changed because of the explosion of data on account of mobile and other digital devices? Have you seen a change in this cycle?

For the leading-edge semiconductor nodes, we are seeing ongoing high demand for computing capacity. The reason for that is the explosion in data and the advent of advanced artificial intelligence algorithms. AI requires high-performance computing.

While on some commodities like memory and storage, you may continue to see cycles - for leading-edge semiconductor compute nodes, we see consistent growth in demand, commensurate with the growth in compute capacity need.

What’s the contribution of your Indian operations to AMD?

For AMD, our India design centres are central to almost every product that we design and develop. We have been invested in India for over 15 years, and we have grown the talent and the team in India to where they, now have deep expertise.

When we started in India over 15 years ago, we had more of junior engineers, who assisted the design activity that centered out of North America. But, if you look at where we are today in 2021, we have now built deep expertise in India. Projects are led and executed, by senior leaders based out of our India development centre. And this team has done an excellent job of building their expertise and growing their contributions to AMD.

How is your data centre business evolving and growing?

AMD data centre business is booming. If you look at our most recent quarterly announcements, we doubled our revenue in the data centre over the previous year. So, there has been significant growth, fuelled not only by the demands from hyperscale and the largest public data centres, but also our enterprise business.

With our third generation of EPYC, our enterprise business along with the hyperscale customers, have achieved high growth rates.

We see very strong growth in the data centre, and key for the data centre business is not only increasing performance each generation, but also the efficiency of computing.