Good morning, welcome to Marvell's 2020 Investor Day. My name is Ashish Saran. I'm the head of investor relations at Marvell. Let me give you a quick preview of today's presentation. Matt will summarize our key growth drivers. Sandeep will take you deep into our technology platform. Dan and Raghib will flesh out the growth opportunities in our storage and networking businesses respectively. We'll note that we'll have a short 10-minute break after Dan's presentation. Jean will provide an update on our long-term financial model. We will wrap up our presentation today with a Q&A session. Participants can submit questions through the chat window in the webcast. All right. Now we get to the fun part, our safe harbor statement.
Except for statements of historical fact, this presentation contains forward-looking statements within the meaning of the federal securities laws, including but not limited to statements related to market trends and to the company's business and operations, business opportunities, growth strategy and expectations, and financial targets and plans that involve risks and uncertainties. These statements are not guarantees of results and should not be considered as an indication of future activity or future performance. Actual events or results may differ materially from those described in this presentation due to a number of risks and uncertainties. For factors that could cause Marvell's results to vary from expectations, please see the risk factors identified in Marvell's quarterly report on Form 10-Q for the fiscal quarter ended August 1st, 2020, as filed with the SEC on August 28th, 2020. During this presentation, we will also be making certain non-GAAP financial measures.
A reconciliation to GAAP is provided at the end of this presentation, which will be filed on our website. Let me now explain the syntax behind our fiscal calendar and revenue reporting. Marvell financial measures in this presentation correspond to our fiscal calendar. Our fiscal year typically starts at the beginning of February and runs through the end of January of the following year. As an example, for fiscal year 2020, it started in February of calendar year 2019. Moving on to revenue reporting. We typically report our total revenue broken out by three product groups: storage, networking, and other. However, to provide additional context during this presentation, we will also provide revenue and addressable market trends by our four end markets. With that, we're going to kick things off with a short video, followed by Matt's presentation. Thank you.
[Presentation]
Good morning, everyone. My name is Matt Murphy. I'm President and CEO of Marvell, and it's great to see all of you here virtually. I'm very excited to talk to you today about Marvell's opportunity in data infrastructure and accelerating our growth in 5G, cloud, and automotive. Before we get into all of that, let's step back and take a look at the market. It's pretty amazing how critical our data infrastructure is to our world today. It's the essential infrastructure that our economy is built upon and bringing you this presentation right now. If you go back 10 years, what if we had a global pandemic? Could we have sent so many people to work from home and still remain productive? There's no way. Recent events have put our infrastructure in the limelight, but at Marvell, we have been focused on this market for some time.
We selected this market coming out of our very first portfolio review in 2016, and we talked to all of you about it in our March 2017 Investor Day. I believe that this would be the most attractive market in semiconductors, with the largest and fastest growing TAM, growing at multiples of the consumer industry, and with a number of near-term drivers for potential long-term acceleration of growth. It's hard. This is not a market that just anyone can jump into in infrastructure. Large customers dominate the market, and they demand trust, partnership, and execution. I felt it was a market we could focus on and build a strong position for the long term. Let's define the market. Last year, total semiconductor revenues globally was $420 billion. If you take out memory, that leaves a $310 billion market for integrated circuits.
Whether you look at Gartner or WSTS. They all cut the market in a similar way. Automotive, consumer, and industrial, those are all pretty intuitive markets. The other two, communications devices that consumers purchase, like mobile phones, and they follow a consumer upgrade cycle. It also includes telecom infrastructure, 5G base stations, enterprise LAN equipment, service provider core routers. This is capital equipment. Data processing, again, this is more of a technology. It includes PCs, tablets, printers. These are things consumers buy. It also includes servers, data storage infrastructure. Again, capital equipment, routers deployed by telecom service providers. Combine it with the servers and storage deployed in data centers, add in industrial and automotive, that's the infrastructure market. You might say, well, isn't automotive a consumer market?
It's true, automobiles are purchased by consumers today, but they follow a market trend much more similar to infrastructure. It's a long-term, stable market we believe is going through a major transition. It's adopting similar technologies to other infrastructure, and some would argue that in the long term, transportation will be provided as a service. Let's be clear, we're not addressing the traditional part of the automotive market. We're focused on the emerging opportunities. If you take that $310 billion semiconductor market, split it into consumer and infrastructure's $140 billion, consumer is $170 billion. Infrastructure is growing a steady 6%, or 3x the consumer market, which is growing much slower with, by the way, lots of volatility, ups and downs, quarter-to-quarter and year-over-year with consumer demand trend.
That's why we believe that data infrastructure is the best market to be in semiconductors. Let's summarize. You take the $310 billion semiconductor TAM without memory. Data infrastructure is $140 billion. Clearly, we don't address all of that. There's about $30 billion that's in analog and discrete components. These are typically broad market products that aren't part of what we consider to be our TAM. That leaves about $110 billion, which is Marvell's TAM. It's an enormous market, and we are right in the middle of it. We can't service all that right now. The portion that we address today is about $16 billion, growing 9% a year, which is 50% faster than the TAM. Which gives us line of sight to about a $20 billion opportunity and serviceable market over the next few years. Data infrastructure truly is the best market in semis.
They're not short-term trend type of markets. We see this as a multi-decade secular growth opportunity. It's the foundation on which next generation economies will be built. To be clear, we've got the $110 billion TAM. We're sizing our SAM at $20 billion. Over time, as we expand into adjacent markets, as we have been doing, we'll be able to address larger and larger portions of the TAM, particularly in cloud. You'll hear more about this later. Marvell's mission is aligned to this growing opportunity. Our goal is to move, store, process, and secure the world's data faster and more reliably than anyone else. We set this as our mission four years ago. As I said before, you can't just set this as your strategy and then go and start working on it.
The market is incredibly sticky, has a very high barrier to entry, and we needed to pivot our own business at Marvell to drive this organic opportunity over the last few years and look for inorganic opportunities as well, which align with our vision and view of the world. We made a key decision very early on of what we were going to focus on and what we were not going to focus on. We let that strategy drive our actions. Let me take you back to 2017. Marvell had strong markets. We were committed to the infrastructure strategy, and I believe in the power of focus. We divested businesses that didn't fit our strategy. We found homes where those businesses and those teams could be more successful. Ultimately, we unlocked almost $2 billion in cash, and we turned our focus to infrastructure.
As I said, Marvell had the networking and storage pieces. In many systems, the heart is the compute element. Marvell did have an upstart processor business. We were focused mostly at the low end with two and four core products. We didn't have the high-end multi-core processors required by big infrastructure. If we were going to be successful there, we needed to acquire. We did. That's where Cavium came in. Cavium was always a leader in compute with its innovative OCTEON platform. We added that to our portfolio in 2018. In 2019, we added Aquantia to increase our overall strength in networking. As we engaged with the customer base, we also saw a trend very clearly where customization capability was really important. We added Avera, decades of history building wired and wireless infrastructure, custom chips from their heritage at IBM.
This significantly increased our SAM, woven across all of our end markets. Importantly, we've integrated now into one company, one team, one culture, leveraging all of these acquisitions into a common technology platform, which ultimately culminated in our recently announced five nanometer platform. Every single one of these businesses will benefit from it. Let's take a look at now how all this comes together in terms of our portfolio. If you want to build any piece of data infrastructure, you really need three core technologies, storage, networking, and compute. Marvell provides the full spectrum today. This includes the key components of both hard disk drive and solid-state drive, and storage networking solutions like Fibre Channel.
It also includes our full Ethernet connectivity portfolio, including switches, PHYs, and adapters to serve the enterprise, data center, and automotive markets, each of which has unique requirements and special technologies associated with them. Our multi-core processing franchise, we call this OCTEON. This category is also called a DPU, for Data Processing Unit. Finally, our full and custom ASIC solutions have a unique business model in the industry and leverages all this IP from the other products in the portfolio and our know-how. All right. Let's see how this now comes together from a customer point of view. As I said, there's a relatively small number of large customers in this infrastructure market. Building their trust is hard. Typically, you start out with one or two small projects to test you out. It takes years to build a significant relationship.
When you get to $100 million revenue with a customer, you're in a different league. It's a big deal. You become integral to their strategy, and they become integral to Marvell's strategy and success. We interlock roadmaps, technologies, architectures, and drive senior-level relationships between the two companies. Just a few years ago, we had five large customers, over $100 million in revenue, three of which were the big three hard drive OEMs. We knew if we were going to be an infrastructure company, we needed to diversify and build strength in the other key markets, and we have done that. As of this year, we have almost doubled the number of $100 million customers in Marvell. We added two large networking OEMs and two large base station OEMs. This is huge progress in such a short period of time.
Most exciting, we are poised to add four more $100 million customers in the near term. This is based on designs we've already won. This speaks volumes for the company we have built and for our future growth potential. Another key component of our strategy is our co-investment business model. The majority of our projects have some form of NRE funding. For those of you that aren't familiar with the term, it stands for non-recurring engineering. It means a customer is funding part of the project themselves, and it's treated from an accounting point of view as an offset to our R&D expenses. Of course, all of our ASIC projects come with NRE. Even for our non-ASIC projects, over one-third of those have some portion of the development cost funded by NRE.
We utilize this framework not just because of the funding, which is helpful, but because we know that when they have to write a check, they've got skin in the game. Writing a multimillion-dollar check typically requires senior level approval. This significantly de-risks those design wins and creates focus on the ultimate success of the program. Yes, the funding helps, okay? It lets us punch above our weight and deploy an R&D footprint that is larger than what we could normally organically afford. This is important as costs continue to rise as you go to the advanced nodes and only a few companies are able to keep pace. Now let me tell you how we execute these programs. We have our own playbook, and look, it takes discipline, it takes hard work, it takes planning.
Now we've built a set of tools and processes around it to enable our teams to be successful. The ability to execute consistently is a huge driver of success in this market. Knowing this, very early on, we decided we had to get this right. In fact, it's one of the first things that we put in place when the management team joined a few years ago. This is one of our core behaviors in the company that every employee knows, which is execute with thoroughness and rigor. Now, I'm not going to walk through each of these in details, but at a high level, the goal of this framework is really to do two things. One, make sure that we're working on the right projects, the ones we believe could be successful and can deliver an attractive financial return.
Two, make sure we deliver them on time and with high quality. I'm telling you all of this because this is where I spend my time. I'm in all these reviews, building accountability, setting the tone. This is also what we spend all of our money on, research and development. It's very important to get this right. Look, if you look over the past three years, we've cut five to six months of lateness out of the average project, which means now we're tracking to 90%+ accuracy to every one of our schedules. It's an incredible execution on projects that often exceed three years in development. Look, we built a strong execution machine and customers have noticed. All right, beyond executing on projects and delivering for customers, we also believe that companies have a higher purpose.
We take our corporate social responsibility very seriously, and it's fundamental to our long-term growth and competitiveness. I have a strong belief that these efforts drive business resilience because our customers are looking for this, our suppliers too. Employees are looking for this. It improves talent attraction and retention, and I strongly believe we get much better outcomes with a diverse workforce. We do a lot of these ESG efforts internally, but historically, we have not systematically disclosed it externally, and we will do more of that over the coming year. For example, we are making progress on reducing our carbon footprint, and we are particularly proud that during the recent time our cafeteria was closed, we donated all of our food budget to the Second Harvest of Silicon Valley.
We've also built a very strong board over the last few years with deep financial backgrounds, several with deep technical backgrounds, and most recently, we added Marachel Knight from AT&T to provide us with the voice of the customer. Our ESG program is very important to Marvell and has my personal support and personal involvement. When you take a step back and look at what we've done over the past four years, we've pivoted to infrastructure, developed a culture of excellence, and built deep partnerships with the key customers in this industry. That's how we knew it was time to launch our new brand, because in this industry, your brand is really your reputation. We have now the reputation for excellence. I'm sure you noticed we launched our new brand back in May, Essential Technology Done Right.
It speaks to the criticality of the technology that we enable. Done right speaks to our execution and also how we partner with our customers. I heard loud and clear when I came to Marvell that our customers were looking for a strong partner. We found that the reason we win is when we're dynamic. We find ways to solve problems. We're attentive to their needs. We're not just pushing our architecture. We're co-innovating and inventing with our customers. We're real with them. We tell the truth, and we deliver on what we say, and we're tenacious and we don't give up. Even if there's a setback, our team is extremely resilient. All of this is resonating really well with our customers, and we see directly in our design win pipeline, the largest one that's been in place since I joined the company. Let's quickly recap.
We talked about how we're focused on the most exciting market in the semiconductor industry, one that is growing faster and that has a very high barrier to entry. We talked about our growing blue chip customer base, our deep partnerships with them, and our co-investment strategy. We're going to talk about our new 5 nanometer technology platform and how we are going to be able to leverage that across our portfolio to drive the next generation of growth. I'll finish with our very exciting key growth drivers in 5G, in cloud, and automotive. All right, let's talk about our technology platform and take a look back. Marvell, Cavium, Aquantia, and Avera, each individually were leaders in product definition and architecture. All very creative teams, very innovative, even without access to the latest process node. Each of these companies was somewhat limited by scale.
Now we're putting our foot on the gas with our new 5 nanometer technology platform. Not just anybody can do this. There's only about five or six companies that have the scale to invest in the latest process node. This is where our co-investment model I talked about really helps. We've placed our bet on 5. I hired Sandeep to help do that. In fact, I asked him in the interview, "How would you do it?" As I'll show you today, we have working five nanometer silicon now in our labs here in Santa Clara. The performance improvement of this 5 nanometer platform are dramatic. 40% lower power than seven nanometer, which is the cutting edge today. We're able to pack 30% more content into these SOCs and more features. That's just the benefits we get from the process node.
If you couple that with our unique architecture and design, we can push the improvements, and we'll push the improvements even further. Look, you'll hear today about multiple growth drivers near term that we're going to talk about next. Our five nanometer technology platform accelerates growth for Marvell a few years out. We have silicon in the lab today. We're going to sample our first product by the end of FY 2022. We're starting to win designs, and you'll hear about that later. We'll start to see revenue in FY 2024, and then growth beyond that. Turning to our market opportunity. When I went through our SAM earlier, I showed it was $16 billion growing 9% to $20 billion. We categorize our SAM into really four end markets, enterprise, carrier, data center, and edge, each with their own growth dynamics.
Within each of these markets, some subsegments are growing much faster than the total. Let's examine that in more detail. First is carrier, where we have wireless infrastructure. This includes all the 4 and 5G base stations. This market is growing very fast. It's important to note, 80% of the SAM is in networking, 20% is in storage. Raghib and Dan will talk to you about the cloud opportunity in their sections. Finally, you have edge, which is really three markets. We're focusing our efforts on automotive, which has the highest or the fastest-growing parts by far, over 4x the overall industry. Here's what that means in terms of revenue. Keep in mind, this is just the last 12 months. The growth segments at the top are ramping now and are already at a higher run rate today.
For consumer, we're very opportunistic, and we believe we'll stay flat with that market. For enterprise, wired, and industrial, we have a $1.5 billion business there today. We've just refreshed our complete portfolio with features and capabilities optimized for these markets, and we expect to continue to take share and outgrow the market. For the high growth segments of 5G Cloud and auto, we've been building our position and have about 10% share today. This is growing from very little just a few years ago. In a moment, I'm going to walk you through each of these segments in detail, but the summary is that we have the products and the design wins to take significant share and grow 2x the market over the next few years.
Taken as a whole, this positions Marvell for long-term growth of 10%-15%, much higher than we said at our 2018 Investor Day. Let's take a look at 5G Cloud and auto in more detail. At our last Investor Day in 2018, we outlined for the first time our platform strategy for 5G. We had one customer, and shortly thereafter, we won a second customer for baseband only. Shortly after that, we announced Avera, which gave us incremental revenue and customers in 5G, and that formed the base case of our 5G revenue opportunity that we've articulated. That's what everybody's been using in their model. Now, let me let you know what's happened since then. First, we've strengthened our position at the second OEM. We have, going forward, the full share of the baseband as well as the transport processor in the future.
Second, on Ethernet switch and PHY, we have started to drive attach across the customer base. Our base case did not include much contribution from Ethernet switch and PHY, but now we have won designs at three different customers. Third, we're starting to address the radio head portion of 5G in a much more meaningful way. We've won an additional customer for a radio head processor, and we see a trend forming where more intelligence is needed in the radio head, particularly for Massive MIMO. What started as a single opportunity is now an engagement across every OEM. Fourth, the Massive MIMO itself growth trajectory is faster than we expected. Our base case assumed a very modest attach rate for Massive MIMO, but that is looking much higher. Fifth, we started winning tier two, more regionally focused players. We are platform in the market.
The alternative is a discrete solution. We've now secured two design wins within the tier two customer base, and we're chasing more. Finally, we had not included any potential share shifts, and that is looking increasingly likely. Just to give you a sense, every point in share shift to our customers opens up $60 million of opportunity for Marvell. If you even assume a modest 5% shift in share to our customers, that would open a $300 million incremental opportunity for Marvell. In summary, we're very pleased with our progress in 5G, and now we see the opportunity is significantly larger than we had previously communicated. Turning to our cloud opportunity. Earlier this year, we announced that cloud revenue had ticked over 10% of total revenue, and that surprised many and generated a lot of interest.
Just for reference, we see the cloud opportunity as big or bigger than the 5G opportunity. The data center TAM is huge. It's $28 billion, and most of that is in cloud. Be even larger than 5G. Let me tell you about our business in the cloud. Over 10% of revenue, as I mentioned, is coming from the segment. About half of that is in storage, and half of that is in networking, with the networking piece growing much faster. In storage, Dan's going to talk to you about our strategy about data center SSDs and nearline HDD products. Platform and our custom ASIC business. The cloud revenue opportunity is similar to 5G in the near term. The exciting thing is that in the long term, the opportunity in cloud could be much bigger just given the size of the overall TAM. Finally, automotive.
It's a new market for Marvell. When we had first talked about this in 2017, it was really a startup business. Today, it's become a real business. There's already an existing legacy automotive connectivity market that exists. It's already very large. These are primarily slow legacy point-to-point connections. In the future, all cars will have Ethernet as a backbone. Every single OEM is doing this now or has a plan to do so in the future. This takes time to overhaul the overall architecture. The automotive Ethernet market is poised to be a billion-dollar market for Marvell. The range is anywhere between $5 and $50.
Even if just 1/3 of the market adopted Ethernet and used the midpoint of the range I just gave you on content, you'd already be at a $1 billion market that we can see in front of us, and we're investing here to be the leader. We've signed up 24 OEMs, seven of which are in the top 10 car manufacturers in the world. Based on the designs we've won, we see a clear path to drive our market share to over 1/3. If you look out over the next 5-10 years, automotive could be the single biggest incremental opportunity for Marvell across the whole portfolio, not just Ethernet, but compute, storage, and security. Let me conclude by summarizing everything we've covered. First, we believe the market matters. We have selected the best opportunity in semi.
$110 billion TAM is growing at multiples of the overall semiconductor market, with line of sight for Marvell [to a twenty. Second,] w e built a significant set of customer relationships with all the big blue-chip infrastructure customers, and they are awarding us with more and more of their business. Our co-invest business model de-risks those design wins and allows us to deploy a much larger R&D footprint than we could have organically. We've assembled a leadership portfolio of unique products and technologies. In each of these markets that we compete in, there's typically us and maybe one or two at the most other players. With our new five nanometer technology platform, we are putting our foot on the gas. There's only five to six companies in the world on the leading edge node. Even fewer of those are in infrastructure.
You'll hear from the team later, we're already getting design wins, which will be very meaningful to the company. Fourth, we have multiple drivers for outsized growth, both near and long term. Our 5G opportunity is significantly larger than we previously thought. Our cloud revenue opportunity is similar to 5G in the near term and much larger in the long term with SAM expansion. We've secured design wins to drive our market share in automotive to over 1/3. Hopefully, you can see we're building something very special here at Marvell. I'd like to now turn it over to Sandeep, who will take you through our technology platform. Thank you very much.
Thank you, Matt. Good morning, everyone. I hope you're all doing great today. My name is Sandeep Bharathi. I'm the Senior Vice President of Central Engineering at Marvell. As the leader of central engineering, my team and I are responsible for setting the strategy for the process technology for Marvell, building the IPs that are necessary to go on these process technologies, as well as developing the tools, flows, and methods that require the chips to be built to address storage, server processors, networking and compute products that Raghib and Dan will go over. I joined Marvell in February of 2019. I just wanted to take you through some of the experience before coming to Marvell.
I started my career 25 years ago in 1995 at Intel, and for the first 15 years, I was responsible for various engineering and leadership roles doing x86 process development for server and client market segments. Prior to coming to Marvell, for about eight years, I spent my time at Xilinx and Intel leading FPGA product development at various process technologies. I have had the experience of doing and developing products, in an IDM model or an integrated device manufacturer model or a fabless model. Before I jump in to talk about the technology, his vision to lay out a clear strategy to lead Marvell to the data infrastructure market. It was immediately apparent to me how this is very critical to map product architectures to advanced process technologies and how to scale to address the needs of the data infrastructure market.
No conversation in the semiconductor industry is complete without a reference to Moore's Law. You might all have heard that Moore's who addressed that and what part of Moore's Law is slowing down. In the last 30 years, Moore's Law has kept pace. What did that mean? Basically, going from advanced generation one after the other, you got lower cost because the number of transistors that you could pack in for a square millimeter of die kept increasing, cost decreased, you got higher performance, and thereby also getting lower power because of the advances in semiconductor manufacturing. At about nanometer, I'm going to three nanometer. Why do we have to go to the advanced process geometries? The need for compute and performance and lower power keeps increasing because of the advent of various applications with explosion in data.
What that means is in advanced geometries, you can pack more transistors. More transistors is a proxy for talk to you about the singular focus and process choice is just not the main thing that we should focus on. It really becomes a platform play, and I'll go over that on what it means to be in a platform play. In the past 50 years, advances in communication and compute have really defined several eras. Initially, it was the advent of the invention of the transistor, then came the mini computer generation with more complex chips that were being added to the portfolio, followed by the IBM PC and the internet era. In the early 2000s, it was all about cell phones and mobile communications.
All through this time, there were certainly a lot of advances in semiconductor manufacturing that made these transitions possible. We have hit an explosion of data that continues even to this day, and as I mentioned before, you really need to go to the advanced process to give you the density that you need. However, process by itself is not going to give you all the performance and power. It becomes a platform play, which means you need to have the right architecture and right design optimized for the right process. The platform focus delivers the highest performance per watt. In summary, migrating to the advanced geometry is imperative because first, you can deliver the advanced features with increased densities, and that is enabled by the advances in process generation going from one geometry to the next.
Of course, you need to deliver the performance to deliver the same or lower power. As you can see, our own products across three generations, which is the OCTEON DPU, has been able to increase between the three generations to greater than 7x performance per watt. That's an enormous increase in performance per watt, if we had just depended on the process, that alone by itself would not have given us the increase that we have realized. Evolution to a platform leadership is not automatic. As you know, we have acquired, in the past few years, Cavium, Aquantia, and Avera, and Marvell ourselves, independently, all these companies could not have had the scale independently to go to the advanced geometries. Standalone, each of these companies were already good and leading with products that took advantage of architecture and design.
In order to address the data infrastructure market, you really need the scale to deliver the right product with the right performance per watt. Thereby, we have achieved scale and expertise with the combination of all these assets, and now we are able to realize the market segments with advances in process, and we are leading with the process technology, coupled with our existing competence in architecture and design. Process. Let me take you through what historically our strategy was on process. Prior to the current advanced geometries, our explicit strategy was to be a fast follower. Meaning, if you had the foundry ready at a given geometry, we were very good at wringing out the performance and the cost from that particular geometry and explicitly be a fast follower. That was the strategy, and we were very good at doing that.
Now, with the advent of new application in compute and networking segments, we really need to be at these advanced nodes. Hence, with the scale that I've already talked about, we are able to go and be the leader in five nanometer. Today I'm very excited to present before you the very first five nanometer chip that we have from the lab, and it was working within a week of its arrival in our labs. This is not just any five nanometer node, it's the latest process addition that we have been able to get from, which is N5P from TSMC.
While we have the five nanometer portfolio right now, we are not sitting still because we want to establish a firm foundation to the product development going forward that will get revenues three to five years down the line, and that's how we are able to have a small team working on the next generation, which is three nanometer. We have established this platform really to be able to do that, and I'll talk next about how we do this key platform features for leading-edge process nodes. First and foremost, in order to do this, I've already talked about why you need to go to the next advanced process node. Now, how do we do this at Marvell? First, you need access to leading-edge process collateral from foundries.
We are a handful of companies, maybe five to six, that are playing in five nanometer, and it requires a extremely good partnership with our foundry partners. You need to build the analog and digital IP that's very hard to build at these advanced nodes. In order to build the chips, you need advanced packaging to put out the products. Before anybody embarks on the product development, whether it's in Dan's group or Raghib's group, we need to have test prototypes built such that it's greases the skids on getting this tools, flows, and methods, and a development flow has to be put together. These test prototypes cannot happen throughout the year. There are only specific shuttles that the foundry runs, and that's why it is important to focus on the test prototypes. However, standalone, we are not able to do this by ourselves.
We need a very strong industry ecosystem partners, whether it is foundry, EDA software to build our chips, wafer chip, test chip and assembly, external IP portfolio of interacting and collaborating with our partners, and the supply chain that is necessary to bring the products into volume production. This is how we do this, and these are all important, very complex and hard to do. That's why we are one of a handful few of companies to be able to, as I showed you, have the ability to bring these products into the market. Scale enables the delivery of a complex technology platform. I've talked about process, why we need to do that, be there, how we are going to do that. Advances in the data infrastructure market really means we have to be first to the market. Being first really matters.
Why is that? Because you have to get the features, you have to get the performance and lower power. In 5G or any of these segments, the life cycles are really long. First-mover advantage captures the market share that we will need because these are long life cycles, and first-mover advantage is very helpful to establish incumbency, and incumbency leads to market share. Once you have locked in a socket, it's very difficult for our customers and partners to requal different parts, and that's why this being first to any market segment in these advanced nodes is really imperative, and it matters. I've talked about process. I've talked about why being first matters. Let me just talk through what are the products that we build. What is the anatomy of a modern SoC? SoC here really means a system on a chip.
Chips have gotten very complex in the last few years. In a chip, you've heard about our vision, it is to process, move, store, and secure data. For processing, you will have a CPU, or many of them for that matter, and it has to act on some data which you need to bring data from the outside and store it on chip caches or SRAMs. In order to move the data, you have to interact with a DRAM interface or a memory interface externally or a high-speed I/O interface. Purpose-built functions like AI engines, packet processing engines, accelerators, and graphics are all important in the anatomy of a modern SoC or building a modern SoC.
On top of all of this, there are foundational blocks that are necessary to be built, which may be thermal sensors for power management, voltage regulators, so on and so forth. As you can see, we have built a complex subsystem, which is our OCTEON multi-core processors, which have all these ingredients. It's tens of billions of transistors, and it has to work perfectly and ramp to volume production. Now, as a result, we also, for this anatomy of a modern SoC, we have to build a comprehensive IP infrastructure portfolio. That basically means you have to have a foundational IP, which I talked about earlier. It may be voltage regulators, memories, thermal sensors, so on and so forth.
On top of this, a critical analog and mixed signal IP, whether it's high-speed I/O, SerDes, which I'll be talking in a moment, and DRAM interfaces and so on. You have processors, complex CPU processors, multi-core subsystems. On top of it, system IP, because we already talked about it. It's a system on a chip. You have to have the systems that are necessary for either storage, networking, or processing. By building this comprehensive portfolio, we are able to service all the business units at Marvell. This scale is so crucial and critical for us being a competitive player in the market space. Now, I'll switch gears and talk about the data movement challenge. What is the data movement challenge?
If you look at any data center, you will have several racks of server processors, and there is a switch on top of the processor racks, and you have rows of these. To get data from a switch chip to a server chip or between servers, you have what are known as SerDes. These are an important and a fundamental building block. If you have a data center that has a 10 gigabits per second SerDes, that's the only data rate that it can move data through. Now, if you want higher data rates to establish and increase the scale of the data center in bandwidth, then you have to have different data rates, which is 28 gig, 56 gig, 112 gig, and so on. This is crucial. This is the most important building block, and it's very complex.
Let me talk through what roadmap we have in this fundamental building block. The critical things that we have in the past, as we talked about fast following in the process, we used to focus on 10 gigabits per second. Now we are one of the very few players who have 56 and 112 at multiple nodes, 16 and 12 nanometer. We have the same capability in seven nanometer. What I just showed you with a chip, we have this, which is brand new, which is working in our labs, 112 and 56 gigabits of SerDes, and that's a key defining IP and a key competitive differentiator for us. We are not resting here. We have the technology platform that has established the foundation for our next generation, which is at three nanometer and also 224 gigabits.
This is how we can scale the data center up from where we are today. Now, I just walked you through process, why being first matters, platform, the Marvell comprehensive portfolio, and one of the critical building blocks, which is SerDes. I want to end this with how we are uniquely positioned to lead. Being first matters, that leads to market share. Best-in-class technology platform, which includes process dominance, architecture, and design, is critical for any product that Dan's team and Raghib's teams build. We have been able to accomplish this with the scale from our acquisitions and leveraging the platform to address the data infrastructure market, which is the main focus for us that Matt has already outlined for you in the past. I was very excited to share our technology transformation with all of you for the past few years.
Now I want to introduce you to Dan Christman, who will talk about the data center storage.
Thank you, Sandeep. Hello, everyone. I'm Dan Christman, and I run the storage business group here at Marvell. Today, I will cover the huge opportunity in data center for storage. I'll talk about Marvell's strengths and our strategies to address this large market and how we've pivoted our storage revenue to the data center. I'll also discuss the many growth vectors for our storage business, both in the data center and as a whole. Look, data is rapidly moving from our devices to the infrastructure and primarily the data center, right? Primary storage used to be on our smartphones or on our notebook computers or desktops. In fact, 10 years ago, the pinnacle of storage was the hard disk drive in your notebook computer. That has now moved to the data center. It's being stored on SSDs and high-capacity nearline HDDs. This is Marvell's focus.
It's our storage focus. We are focused on the infrastructure, not devices. We are the leader in both HDD and SSD controllers, and that is where the vast majority of data will be stored in the future. What do data center customers care about? What are their priorities around the storage infrastructure? Well, simple. It's performance, capacity, security, and efficiency. Now, you think performance, AI, data analytics, 5G. It's all about speed and low latency, getting the data in and out as fast as possible. That's where the SSD shines. It's all about performance. Capacity. It's massive capacity that's needed, and you need to be able to scale that capacity. That's HDD. It's all about dollars per gigabyte and providing that capacity for the data centers. Security. These are multi-tenant environments, multiple users accessing the same drives at the same time. You need state-of-the-art drive level encryption and acceleration.
Efficiency. Look, the customers want full visibility of these drives. They want to build provision and manage the flash memory. They want to increase the utilization, get the full throughput, and maximize the capacity. How does Marvell help? Well, first, we have the greatest storage franchise in the industry. We are the market leader in SSD controllers, addressing the performance segment in the data center. We also provide storage accelerators for security and efficiency to offload the CPUs. In HDD, I told you, it's all about capacity, dollars per gigabyte. We've been doing this for 25 years. We are the market leader in HDD controllers, and now we've started shipping preamps in volume as well. Last, we have a very strong Fibre Channel business. We have almost half the market, and as you know, this is a long-lived and stable business.
Marvell is uniquely positioned in the data center for storage. As I mentioned, we were founded on storage, we are the market leader in storage, and we've shipped over 5 billion controllers. Look, 25 years, it's a very long time. We've built long-lasting, multi-generational relationships with our customers. We provide thought leadership. We influence and drive technology roadmaps with our customers. As Sandeep showed you earlier, having the right IP in the right process at the right time is incredibly differentiating. Last, we have a very flexible business model. This allows us to address the largest SAM possible. Let's talk SSD for a moment here. I'm going to start off by saying size and scale matter in this business. It's only going to matter more in the future. Marvell has the size and has the scale.
We're the leading SSD market share. We have business across all the swim lanes at our customers. We're able to address multiple swim lanes because of our technology. We have a focus on the data center as our primary focus. Strong relationships with the NAND vendors, the HDD customers, as well as system OEMs. Technology-wise, look, things have moved from SATA and SAS. They've moved on to NVMe, which is PCIe interfaces. You need to be first. We were first with PCIe Gen 4. We are now shipping in high volume with Gen 4. We have multiple new design wins ramping next year in PCIe Gen 4. We'll also be first with PCIe Gen 5. We've already secured multiple design wins for our PCIe Gen 5 data center storage controllers. Process. Sandeep talked about being a leader in process.
Well, in storage, we were actually first with 28 nanometer then first with 12 nanometer. I'm excited today to also say we will be first with five nanometer for our storage controllers. Last, we get to leverage the overall Marvell IP. We have storage, networking, security, and compute, and that's an advantage for our solutions when we can combine them. I'll talk about that towards the end of my presentation in more detail. Last, I mentioned our flexible business model. We're the only supplier that offers both merchant and custom solutions and controllers. We have very strong firmware capability, and we've pioneered the do-it-yourself SSD. Let's talk about that a little more. Today's SSD is really just a black box. It's a PCB, it's a controller.
On there, it has flash memory, but it's built by the flash vendors, and really, there's no visibility or control on this drive for our customers. They can't customize or optimize for their workloads or their applications. Do it yourself solves those problems. As I mentioned, this is a business model pioneered by Marvell. You buy a Marvell controller. It could be merchant, it could be custom for your solution. You get access to all the latest cutting-edge NAND, and you get to work with multiple NANDs at the same time. Then you get complete access to customize the firmware stack for your application, for your workload, for your environment. I'll say one thing that's critical here is you need to be NAND-agnostic to offer this model to your customers. It's not possible for a NAND OEM to do this and offer more than only their NAND.
Marvell has all of the necessary elements to enable this do-it-yourself model in the market. We're the leaders. We're pioneers, and we are the leaders. We have had a strong multi-year engagement with our first edge do-it-yourself SSD customer. It's a fully customized solution architected for their unique application. Now, I'm also excited to announce for the first time a tier 1 data center customer SSD design win. Now, this customer will use our high-performance data center SSD controller. They see great value in the real-time telemetry and the ability to customize the platform across various workloads and applications. Now, this will start ramping next year, and it'll be a multi-year ramp as they expand this program throughout their SSD ecosystem. They're using our merchant data center controller. They see great value in the real-time telemetry that this device offers them.
They will start ramping into production next year. What's more exciting is if you look out multi-years, this platform will only extend itself more and more into their data centers and be used more and more as a percentage of the drives that they buy and use. Now, as I mentioned, we have this do-it-yourself leadership. We have this because we have leading technology. We have the best interfaces, the NAND PHYs, the error corrections. We're on the best processes, the latest processes. Also, we're independent, as I mentioned, as a provider of NANDs, NAND agnostic. We also have a very strong reputation for quality and for leadership. That creates this kind of virtuous loop here with the NAND vendors. It strengthens our relationships there.
We get early access to the latest and greatest NAND. We help our customers get their NAND to market, we also help influence their roadmap. At the same time, you look out in the market and you see, look, things are changing fast. The PCIe interfaces are going from Gen 3, Gen 4, Gen 5. You've got the new NANDs. They require new NAND PHYs. You've got new features being added in, it's actually kind of hard when you look at the NAND vendors to keep up with all of this. It's very hard to stay cutting edge here. When you look and say, "Hey, Marvell has these great relationships already with the NAND vendors. Marvell has this technology leadership." This actually creates new opportunities and again, very excited to say today that we have now won a new tier one NAND OEM.
This is a new customer for Marvell Storage. It's very significant. It's a large volume revenue opportunity, it's a strong proof point that not only are we winning designs with the NAND vendors, but we're able to win a brand new NAND vendor that we've never had before. This will start ramping next year. It's a multi-year engagement. Let's take a closer look now. kind of step back a little bit, look at where all the world's data is being stored. I'll just make two points here. First, the vast majority of the world's data is being stored on either an HDD or an SSD. Second is that HDD is not going anywhere. This is still critical to our infrastructure. You need HDDs for the capacity. If you look and say, "Okay, it's interesting, right?
HDD, SSDs, this is where all the capacity is being stored in the data center or where all the data is being stored. Again, very clear that HDDs are about the capacity. Flash or SSDs are about performance. The opportunity within a data center is actually quite large. As our data shifts, we actually see the exabytes on HDDs also moving into the data center. Client has largely transitioned at this point. It's all about $ per gigabyte. You need to maintain this gap between HDDs and flash, and that is actually happening. It makes HDDs, again, the best selection for capacity drives in the data center. Another interesting point is as the capacity goes up in a nearline HDD, one of the ways they do that is they add more disks.
They add more platters, that actually creates the need for more heads and thus more preamps. You actually see the opportunity for preamps in the data center going up. Let's take a closer look at that for a moment. The preamp is a very large opportunity. It's a $400-plus million SAM. I showed you it's increasing, driven by the number of increasing heads and disks. It actually creates an opportunity where the content of preamps starts approaching the same value as the content for an HDD controller in a nearline hard disk drive. If you remember, this is an emerging opportunity for Marvell. I'm pleased to report that it is now starting to ramp. It's very little share today. The bulk of the growth is ahead of us, we are targeting more than 30% of share of this market.
Look, Marvell's technology is essential for the data center, for the HDD within the data center. Our leading edge read channel and preamp technologies enable our customers to expand capacity and maintain that dollar per gigabyte. They're doing it primarily three ways, right? Extending traditional ways or conventional recording. This requires, again, Marvell's read channel. These are very small signals, very hard to pick up. When you look at adding more platters, not only does the controller have to change to be able to control and write data and read data from multiple disks, but also the preamps need to be faster. You need to provide more channels. Last, you've all heard terms like HAMR and MAMR. These are energy-assist technologies to pack more data onto a single disk, again, require technology innovations by Marvell, both in the read channel and the preamps.
Let's switch gears a little bit, and let's talk about a newer area of business from Marvell. It's data center or storage accelerators. Look, CPU cycles are very expensive. Data center customers are always looking for ways to offload a CPU in a server. Marvell can provide tremendous value by offloading the CPU with our storage accelerators. Just this week, HPE announced a new native NVMe RAID boot SSD. Our storage accelerator in this application sits between the host CPU and the two NVMe SSDs. This enables reliable RAID boot for the operating system without burdening the CPU. We also have other large enterprise OEM customers designing in this same solution. Another example is our virtualized storage accelerator. This will begin its initial deployments in Q4 this year by a cloud data center customer.
Here, instead of the CPU managing a bunch of drives in a multi-tenant environment, our storage accelerator handles the offload for the CPU. It's the first merchant-based solution for IO virtualization of NVMe flash storage. Initial deployments, as I mentioned, will be later this year, but this is a very exciting opportunity for Marvell. Last, let's talk Ethernet storage. This is the next big interface using Ethernet as a storage fabric. We've had a partnership with Kioxia for quite some time, and they announced in September the first Ethernet SSD. You've heard of JBOD, you've heard of JBOF, and you take this a step forward and think about what is EBOF. It's an Ethernet bunch of flash, another innovation by Marvell, and this is fully disaggregated storage or flash storage.
You get full throughput of the drives that is not attainable by using a SmartNIC or an x86 processor. We leverage Marvell's IP across all of our businesses to enable this solution. We use our state-of-the-art Ethernet switches, our storage accelerators, Marvell security IP, as well as our compute complexes. This enables any standard NVMe SSD to directly connect into the network. It's in evaluation right now with multiple tier 1 enterprise customers, HPC customers, as well as storage vendors. In summary, the data center is the largest, most important market for storage. We continue to execute on our pivot to the data center. My goal when I was in front of you two years ago, I promised to get more than half of our revenue into the data center. You can see we're already beating this.
You can see that our client exposure is now very low. We have multiple growth vectors for our storage business. I'll say again, first, scale matters, and that has helped us in this business. I mentioned our DIY wins. I mentioned our new Tier one NAND vendor customer. I mentioned the opportunity in nearline HDDs as well as the preamp opportunities that we've now started ramping, and I also mentioned how we're able to leverage Marvell's technology to provide these storage accelerators and offload the CPUs. We're targeting 30%-35% of the $1 billion plus cloud data center opportunity. As I stood up here in 2018, I told you I would pivot this business to the data center, and we did this. Today, I'm going to tell you that we will continue to grow our storage business, both in the cloud and overall.
Thank you for your time today. We're now going to take a 10-minute break.
[Break]
Welcome back. As you know, my name is Raghib Hussain, and I'm Chief Strategy Officer at Marvell. I also run networking and processor business. We have three product lines in networking. Ethernet networking and processor are established businesses, and we recently added ASIC business as a part of Avera acquisition. I'm very pleased with the progress of that business, and it is tracking over our initial estimation. That is mainly driven by the upside growth in 5G and data center. We use our products to address four end markets. First, I'll talk about enterprise. You must be thinking, why am I talking about enterprise when we are all sitting at home and attending this video conference? Well, enterprise is not about connecting offices and cubes. It is about connecting peoples and resources, and people are still at work. The pandemic is, of course, changing the enterprise networking.
For IT groups, it is even more challenging because now they have to connect thousands of houses and make sure the security and quality of service is maintained for the productivity. Not only the volume of data is increasing, but also the data sources are evolving. It is going beyond the boundary of the enterprise campus. The process of stretching enterprise was already underway, and pandemic has just accelerated it. We call it Borderless Enterprise, where mobility and cloud has made enterprise network completely virtual. Now, we have people, devices, data sources, and apps everywhere. We need to connect and network all of them in a secure manner. It can be all kind of businesses. It can be schools, it can be hospitals, it can be retail stores, or it can be stadiums. All these entities communicates with their workers, whether working on-site or at home.
They communicate with partners and customers and various kind of devices. We need to network and secure all of them everywhere. As I said, that pandemic has accelerated the trend, and the need of high bandwidth and security in enterprise networking is even stronger. If you are a CIO and you want to implement a borderless enterprise, what do you need? You need to have visibility. If you cannot control the network, you need to assume there are threats and bottlenecks everywhere. You need to have network intelligence, because all that data that is generated at various nodes and end devices cannot be brought to the center for processing and decision-making.
That is why you need more intelligence in every network node to be able to detect and parse the useful information of data. Then send only that useful information to the center to be processed and new policies to be implemented. You need to assume zero trust and make sure every node in the network is fully protected. All this needs to be done with ever-increasing data volume. As an example, we have been talking about video traffic in the network for over a decade. This pandemic has dramatically increased the amount of video traffic in our network. We used to have only a few employees using VPN. Now all of them are using VPN. All this is driving the need of a lot more compute, visibility, and flexibility all over the network.
Our switches and PHYs are designed with all such features and are best suited for borderless enterprise networking. Our OCTEON processor family provides further intelligence in every node of the network to make it more effective and visible. Let me share with you a case study of a fully automated retail store. This customer wanted to implement an automated retail store where customers could simply walk in, pick up the item, and walk away without the need of any checkout process. This customer needed automated inventory management, tracking, shelfing. It required electronic labeling and sensing, and a lot more. We use efficient, smart, and optimized network solution implemented on top of feature-rich Marvell silicon. Special security, telemetry, and traffic analysis feature of our product were used to achieve the goal of fully automated retail store. All modern stores are moving in that direction.
Such market requirements place Marvell in a unique position to take leading position in that specific verticals. How are we addressing everything that I told you about? Matt shared with you that we are gaining market share in enterprise networking. Why are we gaining market share? Well, because we are focused on this market. We have developed optimized silicon solution for borderless enterprise. In the last 18 months, we have completely refreshed our switches, PHYs, and processor product portfolio. Three years ago, we presented at the Analyst Day that we have complete refreshed product portfolio in 28 nanometer. That drove value, higher margin and ASPs and gave us return of investment. Now, we have yet again refreshed our complete product portfolio, combining best of the breeds from Cavium, Aquantia and Marvell.
This is going to give even more value to our customer and in turn drive higher ASPs and margin for us. These silicon are optimized for the new needs of the enterprise. Combination of Aquantia further strengthen our position in multi-gig PHY solution. Today, Marvell has the most comprehensive and feature-rich portfolio for the entire borderless enterprise. We are winning and gaining market share. Let me share with you two examples. We have been partnering with one of the tier 1 networking OEM in North America. That OEM, for the first time, chose our switches and PHY platform and ported their own networking operating system. This is significant. With the lack of their own OS support on our product platform, our solutions were limited to certain segments of the market. By being the part of their integrated product portfolio, we have broader market reach.
In addition, that removed the hurdle of our product being adopted for additional design in their product portfolio. The second example is of yet another tier 1 networking OEM. They chose our switches and PHYs to implement in their design in their product portfolio. With these initial design going in production, we are winning even more design in that product line. These examples show the value that we bring to our customer with our latest and innovative products. These initial designs are in production and ramping. With these wins, we are poised to win more designs in those product families. The last example that I showed you was exciting, but even more exciting are the one that I'm going to share with you now. Our OCTEON product line are very successful in the networking appliances. We have market leadership position when it comes to data plane and security processing.
With our next generation five nanometer OCTEON product family, we have won a major design in a tier 1 networking OEM in North America. This OEM has moved their entire enterprise router platform from x86 to OCTEON. Sandeep talked about the value of accelerating process node technology. This is an example of winning due to time to market. This is a testament of our ability to provide most performance and optimize multi-core processors solution to our customers. This will be over $100 million opportunity for Marvell. This is also a validation that 5 nanometer platform is not only needed for cloud, 5G or ASIC. It is equally applicable for enterprise platforms. I'll talk about automotive. Traditional car used to be a means of transporting from point A to point B. Today's car are not just engines on wheels. They have all kind of features.
These features are evolved over time. From window control to cruise control, to backup camera, to lane assist. Every one of those features were added one at a time, resulting in a dozen of disparate systems not able to intercommunicate. This is the reason it is very difficult to add a new feature in a traditional car. When a smart car vendor decided to build a car from scratch, it decided to build based on modern technology. They built all those features as an integrated piece of software running on a centralized compute engines. They connected all those cameras and radar and sensors in a high speed, secure network. The result was fully integrated in the field upgradable system. Now you can pick your own horizon, just like all phones became a smartphone, one day, all cars will become a smart car.
To enable the smart car, you need a new type of networking technology. Let's see what is that networking technology. Today, the industry spend about $1.5 billion on car connectivity silicon and additional 10x more in the cabling in the car. These are archaic point-to-point, low performance, low speed, minimal network, not secure system. These are not efficient. Automobiles are adopting networking technology, fully secure, standard-based, and scalable, not only to handle that data bandwidth within the car, but also to optimize the overall cost. All the data that is generated by the cameras and radar and lidar in the car has to be brought to the central compute engines to be processed and to take decision in real time. This requires a special type of backbone in the car that can scale efficiently, that is low latency, that can provide reliability, that is well deployed and established.
That backbone is Ethernet. This is where Marvell is leading the curve. You may be surprised to know that automobile Ethernet has been around since 2007. It started with basic connectivity. The real adoption of switches and PHYs happened around 2017. This chart shows the average Ethernet content in a car. The actual content in a high-end car is much higher than this. If you pick in 2023 $46, the higher end of the range is over $100. When the content is in those ranges in a few years, it will be over billion-dollar opportunity for Marvell. Marvell is investing to be market leader in automotive Ethernet. Let me explain you what does it take to be leader in automotive Ethernet networking. First, you need to be an expert in Ethernet and silicon technology. There are only two companies with those capabilities.
Marvell has 25 years of experience of Ethernet technology. Combined with Aquantia, we have complete solution from 100 megabit all the way to 25 gigabit, we also have leadership in process node technology. That is not all. It is one thing to be able to deliver silicon for the control environment like data center or enterprise. This is completely different ballgame to be able to deliver silicon, which can address the needs of reliability and quality of an environment like car, which is exposed to very harsh environment. The cost of failure is huge. You need automotive excellence and expertise, you need to have capabilities to be able to deliver a high-quality product. We have established a special lab at our Santa Clara facility, which is CISPR 25 compliant.
Labs like these and other specialized labs are needed to deliver the auto-grade DPPM, ESD, and EMI product. We have shipped millions of silicon to our automotive customer with a DPPM close to one, and that is world-class, and that is what is needed to be an auto-grade silicon provider. Today we have design win in 24 OEMs. We have seven out of top 10 car manufacturers in the world as a customer. These seven OEMs drive 55% of global car production. With the initial wins, we are poised to win even more design within those OEMs. We already have established ourselves as a leader when it comes to automotive Ethernet networking, it is underway. If you look further, all these cars need solutions for a high-performance compute, highly sophisticated security, as well as storage. All that are exactly Marvell's core technology that can be leveraged.
We are engaged with major OEMs, and those discussions are ongoing. While this progress will take time, it is bound to happen. That is why we see automotive as one of the largest growth driver for Marvell in the long term. Let's talk about our 5G progress. Two years ago, I shared with you our journey in 5G. We had one customer who had adopted our 5G platform. That story has progressed on all fronts. Today, we have complete platform for 5G base station, and we have expanded into radio by providing processor for Massive MIMO processing. In addition, the combination of Avera has given us ASIC solutions for DFE, that not only increased our SAM, but also completed our product offering to cover every aspect of 5G RAN. Finally, we have expanded our customer base from one to four.
As I discussed, we have growing momentum in 5G. We have now four out of top five tier 1 OEM in the world as a customer. Our customers today own about 65% market share in 5G RAN. With evolving geopolitical situation, any share shift will be a gain for our end customer and as a result will benefit us. Matt outlined about various incremental changes that had happened in the last 18 months. Let me give you some more details. We had a lead customer with our complete 5G platform. When another tier 1 was looking for a partner, they chose us for performance and time to market. When we delivered on our initial baseband processor commitment, and they looked at the value that we bring in terms of technology, leadership and performance, they decided to partner with us for the entire platform.
Let me remind you, our full 5G platform include transport processor, baseband processor, switches, PHYs, Massive MIMO radio processor, and DFE ASIC. With increase of data between radio head and base station, traditional interfaces were not able to keep up. As a result, Ethernet was adopted. This is where our expertise in 5G as well as Ethernet made us the right partner for Ethernet solution for front haul. Traditional radio used to be analog devices. However, with the increased processing requirement driven by Massive MIMO and beamforming in 5G, you need a lot more digital processing in radio. These solutions are also power-constrained. As a result of that, traditional FPGAs solutions are not efficient. The type of processing needed in these radio are similar to what is done in base station. That is why the adoption of our technology for the Massive MIMO radio processor was the natural choice.
The attach rate of these Massive MIMO radio varies from country to country and carrier to carrier. However, the initial deployment shows that it is higher than what it was expected. Finally, with the geopolitical situation, tier two are getting traction. It is just a natural choice for them to partner with us because of time to market, technology leadership, and software enablement. We are winning. We are engaged and winning multiple designs at those tier two. As you can see, not only 5G is happening, but we are in a much more stronger position with complete product portfolio and the customer base. The large design win that we have will drive revenue for the Marvell. There have been a lot of talks about RAN architecture. Let me walk you through that. A 5G RAN consists of radio unit, baseband unit, and central unit.
Today, we have processor to address the needs of all of those components. For some RAN applications, O-RAN or vRAN may be a right architecture. In O-RAN and vRAN, the baseband processing is split between radio unit and distributed digital unit. With this split, more and more digital processing, which traditionally used to be in base station, is moving in what is called an O-RAN radio unit. The pressure of compute is even higher in radio unit, and it is very similar to what it used to be in baseband, base station baseband unit. That is why our radio unit processor is much more natural choice to be deployed in those applications. Baseband processing is a complex task. In a baseband unit, we used to have a special processor, which is our OCTEON Fusion processor.
Implementing those processing in those compute in a general purpose CPU is not efficient. That is why these distributed DPUs require specialized RAN accelerators. The technology which is used and being deployed today in base station will be used in those RAN accelerators. Transition like this take long time, and it can be even over a decade. Marvell has built complete portfolio of 5G RAN. The flexibility and scalability of our technology solution makes it right choice for every RAN architecture. As a result, we can provide solution for 5G RAN, O-RAN, or even cloud-based vRAN. This chart shows you 5G overall carrier deployment cycle, which is long. If you look at the yellow line, which is 4G, it shows it has already been going for 10 years, over 10 years, and it still has to go few more years.
5G is in initial stages with the deployment in China and Korea. 5G is going to have similar or longer cycle. All the growth which are going to be driven by 5G is still in front of us. Based on the design wins, market dynamics, as well as our core expertise, we expect 5G to be even bigger growth driver for Marvell revenue. Next, I would like to talk about the exciting opportunity in cloud. Data center is going through a big transition. If you go 10 years ago, data center used to be in enterprise. Today, the massive growth is in cloud. Let's just take a look how these architectures differ. A traditional data center used to have a rack of servers, and they were front-ended by a dedicated appliances.
These appliances offload all kind of networking protocol processing and security processing for these servers. When it comes to security and data plane processing, Marvell has a strong DNA. If you open any one of those devices, you will see that it has our processor in it. In the cloud, all this functionality needs to be done. However, the architecture of the cloud requires elasticity, scale, and virtualization. That is why all these applications are implemented in a distributed manner. In other words, every server, every node in the network will be doing this kind of processing. Initially, cloud CPUs are doing some of these function in software, but obviously it is not efficient. It require new cloud-optimized technology. Marvell has made this pivot very successfully. Our silicons are being used in both scenarios.
To be able to address the need of data processing in the cloud, there's a need of a specialized processor called Data Processing Unit. To understand what is Data Processing Unit or DPU is all about, let's go through the evolution of compute. We all know CPUs are designed for general purpose compute. GPUs are specialized for parallel processing. However, with the advent of internet, this is the first time networks were connected and data had to be shared between the networks. It sounds simple, but all this data movement requires a lot of hand-holding and processing. This is the first time when data needs to move over various kind of physical mediums. Most of these entities' infrastructure are owned by various entities and cannot be trusted. That is why there's a need of protocol processing and security processing at every node.
CPUs are good for general purpose application processing, but they are not good for high bandwidth data movement as well as protocol processing or security processing. We at Cavium identified that need and developed OCTEON processors. Early OCTEON was the first incarnation of DPU. We developed OCTEON DPU for networking appliances, and we have an established position in that market. As the data volume increased in 5G, there was a need of DPU. This is where OCTEON platform expanded to serve that need. Today, OCTEON-based DPUs are being deployed in transport processing, baseband processing, as well as Massive MIMO radio processing. With the increase of data in cloud, similar processing needs are emerging. Eight years ago, we pioneered server offload technology for the data center, working closely with one of the largest data center in North America. Today, that is called SmartNIC.
We recently announced that we have shipped over one million OCTEON DPU-based SmartNICs. Similarly, as the volume of data increases in smart cars, similar DPUs will be required. Let's take a look at the architecture of the DPU. DPUs are a specialized processor to move, process, secure and manage data and convert it into the form of information that can be effectively utilized by higher layer applications. DPU architecture requires high bandwidth network interfaces. It require memory subsystem to store data while it is being processed. It requires various kind of processors and accelerators, which are specific to line rate protocol processing, security processing, as well as various application-specific processing. In addition, it requires a scalable compute subsystem, which can work hand-in-hand with the accelerators and application-specific processor. This is what is the architecture of OCTEON.
The most unique value that OCTEON brought was the line rate data processing and protocol processing, seamless data sharing between various type of processing engines, and a highly scalable compute platform. With the increase of data volume in cloud, the percentage of data-centric application is surpassing the compute-centric, application-centric compute. This is very similar to transition in AI, where CPUs were not efficient to process the AI specific compute, and as a result of that, GPUs or AI specific hardwares were deployed. Similarly, the server CPUs are not efficient solution to handle the data-centric compute, hence over time, more and more compute will shift from CPU to DPU. That is what is the opportunity. We discussed how DPUs are effective for various kind of application in the cloud. How does it all matter to us? Here is what is already in front of us.
We have four DPU design wins in four tier one. These are starting to ramp. When we acquired Avera, they came with their own designs in the cloud data center. Those are in the early stages of ramp and going to grow in the next few years. You already heard from Matt that we have over 10% of our revenue in cloud, half of which is in networking. The cloud networking SAM is growing to be greater than $4 billion by fiscal year 2024. We are targeting 10%-15% of that share. That is not all. I've showed you in the previous slides how the compute is transitioning from CPU to DPU as a result of transition of compute from application centric to data centric. This is all driven by the growth of data in the cloud. Hyperscale cloud customers want optimization.
They have already implemented their own AI accelerators, compute engines, as well as even DPUs. Why? Because they have scale, they have unique workloads, and they have resources. Each of these trends will drive a massive expansion of SAM for Marvell in cloud data center. We see cloud revenue opportunity in the near term, similar to 5G, but much larger than that in the long term, driven by the SAM expansion because of all these trends. Sandeep shared with you about our investment in five nanometer technology. Today, Marvell has one of the strongest IP portfolios in the market that is best fit with the emerging data processing needs of the cloud. We have already delivered DPU in the cloud and also DPU in 5G. This is the proof of our ability to add platform flexibility and bring the optimized solution specific to target applications.
The richness of our IP portfolio together with process node technology leadership and the unique ASIC design capability and the long track experience of our ASIC team to deliver very complex high-end ASICs has no match in the industry. Arm server is a perfect example of customer-specific optimized silicon. The promise of Arm was to bring efficiency. We build Arm server processor of Xeon class and prove that we can compete on overall performance. However, the real value of Arm server comes from our ability to optimize solution for large cloud applications. For example, some application may need higher compute, higher memory bandwidth, and lesser IO. Others may need various kind of higher IOs, various kind of co-processors, and lesser compute engine. Yet another may need very highly multi-threaded, a lot number of cores and a lot of memory bandwidth.
We are engaged with multiple customers in such discussions, and those talks are ongoing. In summary, enterprise is becoming borderless. Marvell has most feature-rich product portfolio to address the needs of that emerging enterprise. Marvell is gaining market share and winning designs. Car architecture is going through a revolution driven by the explosion of data in cars. Marvell has long history in automotive and well-established leader in automotive Ethernet networking. We have the most complete product portfolio for 5G and well-positioned for radio and RAN architecture evolution. We see 5G even bigger growth driver for Marvell. The data economy is driving an increased need for DPU, and Marvell has the longest history of developing DPU. We have been shipping DPU in production for a long time. Cloud data center requirements are well-aligned with Marvell's core expertise, and as a result, it is a large opportunity for Marvell.
Two years ago, I stood here and said, "If you want 5G in your portfolio, Marvell is your stock." Today, I would like to say, if you want to invest in 5G and cloud with automotive as an upside, Marvell is your stock. Thank you. Now I would like to invite Jean to share our financial model.
Good morning, everyone. Thank you all for joining us virtually today. My name is Jean Hu, Chief Financial Officer of Marvell. I joined the company four years ago. It has been truly exciting to be part of the transformation journey. This morning, you heard from our team about the tremendous market opportunities we have. You also heard how we have been building Marvell to capitalize those opportunities, which are just unfolding in front of us. I'll spend the next 20 minutes to talk about our financial roadmap going forward. My presentation will be centered on three key elements of our path to creating long-term value. First, accelerating revenue growth, like you heard this morning. Second, expand the earnings and the free cash flow generation even faster than revenue growth. Third is our disciplined capital allocation approach to invest for the long-term value creation, also maximize returns for shareholders.
Let's get started. The first and foremost is about accelerating revenue growth. As Matt discussed earlier, it all starts with our market opportunities and our technology leadership position. We have chosen the best market data infrastructure. Not only these markets are substantial and fast-growing, they also have long product cycle and very high barrier to entry with extraordinary economics, which can offer us the best returns on our investment. The market we serve is positioned to grow 9%. Within the market we serve, you heard our team talk about multiple secular growth trend in cloud, data center, automotive. Those opportunities I expect to grow 16% going forward. Now let's look at our product portfolio, how our team has transformed the company completely. The left chart shows our revenue mix since our first Investor Day in 2017.
As you can see in fiscal 2018, our revenue from data center, carrier, enterprise networking was below 40%. Today, if you look at the last 12 months revenue, it's already more than 70%. We also transformed our portfolio within the edge market. In fiscal 2018, our edge market include HDD controllers for clients, multimedia product, Wi-Fi business. Today, our edge portfolio include automotive, DIY flash solutions, industrial. Those are fast-growing market in data infrastructure segment. We're so proud that what our team have achieved during very such short period of time. We have built the most complete product portfolio to serve the best end market in semiconductor industry. Let me quickly do a recap what our team talk about this morning, our top line revenue growth drivers. We expect our 5G, cloud, and automotive business to grow two times faster than the market.
Today, if you look at our last 12 months revenue, 5G and cloud business, they are roughly the same size. Automotive is very small. Going forward, the largest incremental revenue opportunities coming from 5G, followed by cloud. As you heard from our team, that's equally important market for us and have a great potential going forward. Our automotive business today is very small, but growing rapidly. We do expect, based on design wins we have, we'll get to 1/3 of market share. On enterprise, wired, and industrial market, Raghib talk about our refreshed product portfolio with the new features and capabilities. He also talk about our success to win new designs with key leading OEMs. All of those set us up to grow much faster than market.
On consumer side, the DIY flash controller business are expected to grow fast to offset the decline of both our HDD client business and the legacy printer business. Let me translate the major market growth drivers into our product group of storage networking growth. We expect our storage business to grow low single digits, driven by double-digit flash business growth offset HDD client decline. Our Fibre Channel business will be stable, flattish. When you look at the storage business growing low single digit, our networking business is the engine of the company. With significant growth of networking business, it will drive the overall company's revenue growth at 10%-15%. We're super excited about the inflection opportunities ahead of us. Let me shift to how we are going to expand earnings and free cash flow even faster than top line revenue growth. First, on gross margin.
A question I got asked often is, "When you accelerate revenue growth, what kind of gross margin impact do you expect?" I'll take this opportunity to talk about major puts and takes that impact our gross margin going forward. As many of you know from our discussions in the past, gross margin is one of the most critical metrics we follow inside the company. Our team spend a lot of time on improving gross margin. Operationally, you should expect us to continue to drive efficiency and leverage our scale. We actually have multiple programs ongoing to improve gross margin. As a result of all those effort, we do expect to improve our gross margin from current 63% level in the coming fiscal 2022. Our primary driver of the gross margin range of 63%-65% will be revenue mix.
At highest level, we actually see ASP across the company going up as the IP, the content our team is selling to customers increasing. We also have a very disciplined internal process and a team dedicated to overall pricing strategy and execution. Our merchant product, they tend to have higher than corporate average gross margin. When we ramp the business in enterprise networking, cloud, automotive, we're going to see meaningful gross margin improvement going forward. A good example is our Ethernet networking product portfolio. As many of you recall, two years ago, we refreshed the whole product portfolio. Not only we see revenue growth, but also we improve the gross margin by 350 basis points. There are two areas of the company the gross margin is below corporate average. One is the ASIC, the other is the customer solutions.
As many of you know, our revenue ramp could be very lumpy quarter-over-quarter going forward. The way to think about our gross margin range of 63%-65% is if we have large revenue mix of ASIC and semi customer solutions, our gross margin could be more around 63%. On the other side, if we have a cloud business, enterprise networking merchant product, which will push our gross margin more toward the high end of 65%. Just as a reminder, both our ASIC business and semi-customer business, as Matt mentioned earlier, we actually get customer funding. Both of those business have excellent economics. For instance, our ASIC business today, the growth margin is around 50%, but their operating margin actually is better than corporate average. Let's move on to operating expense.
Given the tremendous opportunities Marvell has, we have been investing aggressively to build the technology leadership position. Going forward, it's all about operating leverage. We plan to manage our non-GAAP operating expense increase annually around 3%, which is primarily just merit and salary increase. We're going to continue to drive operational efficiency and also leverage our technology and IP portfolio within the company and with customers. The math actually is very simple if you look at our scale curve. If we can grow top line 10%-15% improve and maintain our strong growth margin and grow OPEX at 3%. I'm sure you can do the math quickly. You're going to see very significant operating leverage. When you look at the last 12 months of operating margin, 21%, that's really at the very beginning of this scale leverage curve we are showing you here.
Let me put it together to look at our overall long-term non-GAAP target financial model. We assume a very normalized operating environment and the macro environment. Our product line tend to have a very long design cycle. The revenue opportunities we discussed here, we already have design wins on most of them. Some of the design opportunities our team talk about with five nanometer, they are even longer term upside opportunities. We expect to grow our revenue at a 10%-15% going forward. We expect to improve and maintain our gross margin in the range of 63%-65%. We're going to continue to invest to target our operating expense at a range of 28%-30%. We expect 35% or more on operating margin side.
We have a very efficient free cash flow conversion as evidenced during our first half of fiscal 2021 free cash flow generation. We expect our free cash flow to be more than 30%. Overall, we are building a company we can grow top-line double-digit, and we also can expand earnings and free cash flow even faster than top-line revenue growth. Let me shift to capital allocation. As Matt discussed here earlier, our disciplined capital allocation approach has helped us to build the company to become today's stronger Marvell. Going forward, our capital allocation approach will remain the same. Our number one job and the number one priority is to invest organically or through acquisition. If you look at the fiscal 2020, our largest capital allocation is $1.2 billion operating expense. We invest in R&D and the supporting function of the company.
We also acquired two companies to expand both market opportunities and our capabilities for $1.1 billion. Our team is also not afraid of divest business if that means it will generate higher return for shareholders. During the year, we divested the Wi-Fi business for $1.7 billion. We used the cash to pay down the debt for the $1.1 billion two acquisitions, and immediately, we returned more than $300 million cash to shareholders through share buyback. We continue to be committed to return more than 50% of free cash flow to shareholders through both share repurchase and a dividend. As you can see, in fiscal 2020, we returned more than $520 million cash to shareholders, more than 100% of our free cash flow. Our driving principle continue to be the same, to focus on highest return on invested capital.
Now let's look at our team's track record since we started the transformation journey. Not only we build a stronger Marvell to drive top line revenue growth and generate earnings and free cash flow, we also returned $1.8 billion cash to shareholders, more than 100% of our free cash flow. Our stock with price appreciation also offered a great return on our share repurchase program, which is significantly higher than cost of capital. Today, we have a very strong balance sheet. At end of Q2, our cash is $832 million. We also have an undrawn credit facility of $500 million. Our business continue to generate a strong free cash flow, and that we have ample access to liquidity. Our near term plan is to pay down our term loan. Once we pay down our term loan, we'll start the share repurchase program.
As you can see, as a company, we have strong financial flexibility to invest for the long-term value creation through the economic cycles. In summary, as you know, we have built Marvell with a broad portfolio to serve the best end market in data infrastructure. All the building blocks are in place for us to outgrow the attractive end market we serve. Our strong business model and disciplined execution will drive operating leverage and expand earnings and free cash flow even faster than our top line revenue growth. Our disciplined capital allocation approach will fund long-term growth strategy and maximize shareholder returns. Before we start our Q&A session, I want to take the opportunity on behalf of the whole Marvell team to thank you all for your interest in Marvell. We also want to thank all our shareholders for your strong support during our transformation journey in the past.
Now we will set up for the Q&A session. Thank you.
Okay. Thank you everybody for attending our virtual Investor Day. It's the first one that we've ever done. I hope you enjoyed it. We have a great story to tell, and we tried to provide a lot of information and transparency in our business. I'd also like to thank the entire team at Marvell who did a great job of putting this together, all the presenters and all the people behind the scenes. With that, we're going to turn it to Q&A for about 20 to 25 minutes. Ashish, who runs our investor relations, will be moderating the Q&A. Then I've got my team here standing by, and I'll answer some questions, and then they'll also answer some questions. With that, Ashish why don't we start the Q&A?
Great. Thanks, Matt. For the first question, for the co-investment model, what factors drive decisions by customers to pay NREs? Are these more technical in nature, or is it purely a financial consideration?
Great. Actually, I'll take that question. I think it's a very good question. The first is that the co-investment model that we articulated is very good for customers, I think, on a number of levels. The first is, the resources that we have in place in this company to do the kinds of work that we do in these advanced geometries with the IP set we have, it's a very scarce few number of companies that can actually pull these types of large, complex SOCs off. First, with the co-investment model, our customers get access to a world-class team that they know is dedicated to their program. The second is that these NRE engagements are typically milestone based.
We set a schedule together working with the customer, it holds our team accountable, it also holds the customer's team accountable, we can drive the schedules to meet these critical time-to-market requirements that are out there. From our standpoint, obviously we benefit because it creates a de-risking to these key programs. It puts skin in the game from the customer and ultimately drives really a program where we can drive mutual success, everybody benefits from it. I think those are some of the key factors from the customer point of view and also Marvell's point of view as to why this business model that we've really developed over the last few years is working well and it's a benefit to our company and to our customers. Ashish, next question.
Great, thanks. The next question is around our DIY model. Can you provide some more details on why customers, especially cloud customers, are looking for DIY SSD controllers?
Great question. I think I'll turn that one over to Dan. Dan, why don't you come on up and answer that question?
Great question. When you get these large cloud data center customers, it's really about control. They want to be able to optimize their infrastructure around their unique applications and workloads. With the do-it-yourself, they can now manage their drives, whether it be for capacity, performance, efficiency, even real-time telemetry, which allows them to kind of see the flow of data and optimize at any given time. It's also very important to them to have access to the latest cutting-edge NAND technology from multiple vendors, for supply reasons as well as technology reasons. That's why being agnostic to the NAND is so important for our DIY model. I'd say last, just the technology that Marvell is able to bring. Sandeep talked a lot today about latest technology nodes, the IP. We talked about various storage acceleration blocks, et cetera.
That's all super critical to have the leading solutions for these cloud data center customers and very differentiating for Marvell.
Thanks, Dan. I think that was a great answer. When we've met customers together, I remember many of them saying, "All roads for DIY end up leading to Marvell, and so we really want to work with you guys." We've pioneered this business model, and as Dan said, there's a tremendous number of benefits that our customers see. Great question. Ashish, how about the next question?
Right. The next question is on our long-term revenue growth rate we outlined in our model. Relative to your long-term CAGR of 10%-15%, the Street's already modeling the next two years slightly higher, closer to 16%. Was your long-term CAGR trying to signal that those growth expectations may be a tad high?
That's a great question. Jean, why don't you come up and answer the question around our long-term growth rate and how we think about it?
Thank you for the question. When you think about our long-term target, that really means long-term, 10%-15%. When you look at our current performance, we have been accelerating both cloud and the 5G revenue, both of them about 10%. Even look at the last 12 months. When we guided the Q3, it's accelerating again. Of course, we will have a higher growth rate sometimes, and in the longer term, we do think with the multiple growth drivers we have, we can perform really well. Okay, thank you.
Yep. Great. Thank you, Jean.
Great.
All right, next question.
The next question's about the exciting world of DPUs. How does Nvidia's new DPU change the competitive landscape for Marvell's products in this space, for example, LiquidIO, LiquidSecurity?
Okay. Well, I think that question is ideally suited for Raghib, since he's been working on these types of products for probably the last 20 years. Raghib, please answer.
Thank you. This trend of compute moving from data centric to application centric, we identified it early on, and I'm pleased to see others are also recognizing it. When it comes to processing data and networking protocol and security, we have established expertise. We have already deployed OCTEON-based DPU in cloud, which is in early stage at the moment, but we expect it to grow rapidly. In addition, in cloud, the overall importance of customization is very critical. The combination of our investment in 5 nanometer technology, combined with ASIC capabilities and in combination with the expertise when it comes to offload, makes us the ideal partner for cloud provider. Thank you.
Very good.
Great. Thanks, Raghib. The next question is around 5G, specifically on Massive MIMO. You said Massive MIMO attach rate is increasing. Could you talk about what your original expectation was and where you see it going? More broadly, can you talk about the competitive landscape and what your products can uniquely provide for customers?
Sure. Maybe I'll answer a part of it and then I'll have Raghib come up and we'll team up on this one. I think we initially, again, in our base case, had very modest assumptions around Massive MIMO. The reason for that is just upfront, I think we wanted to set a very conservative and achievable baseline. As countries have started to deploy, we can start seeing the real numbers. Korea was a great example in 2019 as the first major geography to start deploying 5G networks. We saw in that rollout, the attach rate actually being quite high, and I think others are going to follow. I think that's the first is that we wanted to see some proof points, and we've definitely seen that.
The second is that, as we talked about in the presentation, more and more of the processing and intelligence with the new 5G standards in the splits is moving functionality that was normally in the digital unit or the base unit up into the radio head. That started off for us early on with really one engagement, which we initially thought was going to be maybe limited to one customer and not that significant. It's turned out that that's a trend now we've seen with all the major OEMs. We believe that the radio head opportunity in massive MIMO for us, in terms of content, is much larger. We also think that the attach rate of massive MIMO as well will be more significant. I think I'll leave it at that unless Raghib, you have anything to add. Okay, perfect.
The next question is around process technology. How long will five nanometer be a workhorse process for Marvell, and when should we expect new products using three nanometers?
Sure. I'll have Sandeep come up and address that question.
Good morning. Thanks for that question. As I talked about our five nanometer technology and the comprehensive portfolio that we have built up, this is going to be our key workhorse technology for the next three to five years. There will be initial products that Raghib has talked about that will be there. Over time, other products at Marvell will migrate to five nanometer. It's going to be a really key one for us, and will continue for three to five years for sure. On the three nanometer, it's early in the technology cycle, as these usually are. I do believe we will get some of these test chips roundabout in the next two years, and then that forms the basis of the next platform that we will invest in. Matt, I don't know if you have anything to add to it.
That's great. Thank you, Sandeep.
Thank you.
Yeah, I would just add that, as Sandeep said, obviously we have leading-edge products adopting five today. The way we architected the platform as it relates to our product roadmaps from all the business units in the company, many of them are going to make the leap directly to five. Some of them aren't even in seven today. They're even in older legacy nodes. I think there's going to be a tremendous benefit that our customers are going to see as we start migrating the entire Marvell portfolio across the board directly into this technology where we've really optimized it, as Sandeep described, for the type of performance that's required for the infrastructure market. I believe that's going to last quite a long time. First products are going to sample at the end of next year, production a few years after that.
This really will be a workhorse process, but we also are very committed to technology leadership. That's why for the most leading edge, cutting edge, we will be there on 3 for our customers when they're ready to go.
Great. Thanks, Matt. The next question is around our expanding customer base. Matt, can you provide any more color on the four additional OEMs you expect to become $100 million-plus annual revenue customers? Are these customers focused on 5G or cloud or automotive?
Yeah, great question. As I described, we've had this very nice build where just a few years ago, our three of our top customers were the big hard drive OEMs, we obviously have been on this mission to add and balance that out with additional customers. The four that are coming, there's really two in the 5G area, okay? That's a result of obviously all of our organic efforts that we've put in place to build that market out. Plus, we had Avera come in. On the networking side, there's two very prominent world-leading OEMs as well that were additionally benefiting from both the organic efforts as well as some contribution from the acquisitions we did of Aquantia and Avera. When you think of the progression, these are already meaningful accounts for us.
We're now seeing a path to them graduating to the $100 million club. What I would also say is there's a number of promising accounts behind that we're driving additional growth with. I think this is where we see the next year or two in terms of that large customer set. I actually think there's multiple candidates in the company where we're developing relationships that we could add additional $100 million type of accounts and relationships as well in the future.
Great. Thanks, Matt.
Sure.
The next question comes on our automotive opportunity. It's a two-part question. The first is, it looks like you're making significant progress on the internet connectivity. Can you explain a little bit on why customers are selecting the Marvell platform? The second part of the question is, you also talked about further growth opportunities beyond Ethernet. Again, it'll be great to get some details on that.
Great. I'll take the first part of that question around why are we winning in Ethernet today for automotive. I'll let Raghib answer the second part of the question, which is all the adjacent opportunities that we have in front of us, given the portfolio that we have in compute, in storage, in security. On the original question in terms of automotive Ethernet, I think our progression comes from a few things. I think first off, we've been investing in this area, and we've put a dedicated team in place several years ago. We actually created an entire business unit dedicated to automotive Ethernet. We carved it out of the overall Ethernet organization that we had, and we did that for focus.
The second is very early on, within the first few months after I joined Marvell, it was pretty clear to me that we had great technologies and great IP in this area, we didn't have a lot of people that knew automotive. Very quietly behind the scenes, we've added key personnel, technical personnel, quality assurance personnel, reliability, people that came with automotive experience, some of them with a decade or two decades or even three decades of designing chips in high volume for automotive. We're really pleased. If you look at the initial results, we've shipped millions and millions of units already with a failure rate that is almost out the shoot at a world-class level. I know this from my prior background in spending a lot of time at my prior company in automotive.
We're down close to one DPPM, which is an amazing result. I think it's a combination really of investing in the business, the focus, and having the quality mindset and the quality performance that these OEMs matter, and the scale and the commitment to be there over the next 10 years plus as these OEMs expect a long-term commitment. It's a number of factors. It takes persistence, obviously, and focus. Maybe Raghib, why don't you come up and describe some of the other opportunities which are a little bit farther out, but yet from an opportunity standpoint, we're right in the middle of a number of them.
Thank you, Matt. As Matt described, we have already established ourselves as a leader when it comes to automotive auto-grade Ethernet networking. If you look at the overall evolution of the architecture in car, as I have described in a smart car, the increase of volume of data in the car requires a lot more compute, right? That will require a really high-end, strong processor to be able to do that in a very power and efficient way, in a very optimized implementation. This is where our OCTEON experience and our OCTEON-based DPU experience comes into picture. We have expertise and track record of delivering these optimized multi-core, high performance, power efficient solution in the network appliances, in the 5G, now going into the cloud and data center.
It's very natural for auto vendor when they are looking for a partner who has the capability to handling these heavy data centric compute, it makes us the right partner. In addition, with these architecture, with these cars now, they are going to require much more security, as well as storage to store all these data and all these services, which is all the right core technologies of Marvell. Having that relationship with the OEMs established because of the quality, because of the complete portfolio of the Ethernet and having the track record of delivering the compute, the storage and security makes us the best partner. Thank you.
That's great. I would just add at the end, a lot of investors have asked us, "Hey, this 5G opportunity looks great." Now you've got this cloud opportunity, what's next? As you can hear from our presentations, one of the big opportunities beyond that to keep driving and accelerating our growth is automotive. I went through the infotainment revolution in my prior company, where consumer-based technologies and IPs made their way into automotive, and it was a huge boon for the industry. There was a massive growth and all kinds of new players and entrants flourished during that time. The traditional suppliers, in some ways were replaced by the new suppliers. When I look out over the next 10 years, I'm very sure the next big opportunity in automotive semiconductors is going to be for those companies with the data-centric technologies.
While we're starting with Ethernet, as Raghib mentioned, on multiple domains, we see opportunities for all of the IPs and all of the technologies we have in Marvell to make them automotive grade, optimize them for those applications, and drive a whole new potential future leg of growth that really goes on out in sort of call it the five to 10-year range, if you start looking longer term with where we can go with this thing.
Great. Thanks, Matt. The next question is around our 5G opportunity. Matt, you laid out last year that you saw a path to $750 million in annual revenue, post the Avera acquisition. You laid out a number of additional drivers today. Can you say how comfortable you are in exceeding that target? Can you also lay out a little bit more around, it sounds like you won more in Massive MIMO, and you talked about Ethernet potentially being a bigger opportunity. It'd be great to get some color on both of those.
Sure. To take everybody back in time, the reason that we sized the opportunity the way we did was to set a baseline early on in our journey in 5G. Everything that I showed in my slide, then Raghib showed something very similar in his, which was, here's where we were before and here's where we are today. The today is all incremental, above the base case, Ashish, that you mentioned of the $750 million. We're confident it's going to be larger than that, and depending on the dynamics in terms of attach rate of Massive MIMO, in terms of the potential share shifts that we do see occurring, depending on different countries and their preference for which vendors they use, as well as in Ethernet, et cetera. I think all of those are going to contribute.
We're not going to put a number on each exact one, but when you add them all up, it certainly points to an opportunity that's much larger than we had previously articulated.
Great. Thanks, Matt. The next question is around the OCTEON win we announced. Can you give any more color on what drove the five nanometer win with OCTEON at a North American networking customer? Intel's process issues have led to share loss in traditional compute, but what are you seeing from your infrastructure customers?
Well, I think I would say at a high level one of the key beneficiary product lines early on in our five nanometer portfolio will be the OCTEON line, will be our processor line. Those are the flagship products that we're developing on the technology. When customers now are benchmarking our performance in terms of our OCTEON processor family versus what's out there and the roadmaps that our competitors are showing, it's very compelling. Not only the improvements we're showing from customers who were OCTEON customers before that were migrating. That story's resonating really well, and we showed you some performance data on that.
I'd say for customers historically that haven't used an Arm-based platform and haven't used OCTEON, it's giving us a new set of opportunities, candidly, to address because the combination of process and technology leadership, architecture, and 20 years of track record and 10 generations of OCTEON processing has gotten customers to wake up and really, I think, make some potentially very different decisions about their processor choices. I think that's a great example of an early proof point. I believe there's more to come, and I believe there's a very large market out there of customers who have historically used non-Arm to move to our platform on OCTEON and five nanometers. We hope that that's the first of more to come.
Thanks, Matt. The next question is around our storage business, specifically on SSDs. You said you would be the first to five nanometers in storage. You compete in SSDs versus both merchant and internal solutions. Do you see competition also moving to five nanometers?
I think the short answer is no. The reason is twofold. One is there are internal efforts going on, which have been going on for some time at various levels of magnitude with the different NAND vendors. It's very difficult to afford to invest in a process node like five, or candidly even seven, if all you're doing is basically a couple of chips per year. I think the internal efforts obviously will be difficult for those companies to invest what's necessary to drive a full platform like we are.
From a merchant standpoint, a lot of the other companies that are out there, they're smaller in scale, and so again, I think there's going to be a scale challenge and there's really no way for them to amortize all this investment across a wide number of tape outs and chips over a multiyear period. I think there will be some effort there. Dan and his team are driving a very compelling vision, not only to compete very vigorously against the merchant suppliers, but also to go and continue to try to convince. We've had great success in this, by the way, of our partners in the NAND area who are doing their own developments to consider diversifying internal and external and using a mix of both.
We're starting to see some of those open up, especially with our very aggressive and compelling technology process and product roadmap. I have very high hopes for Dan and his team on their ability to really take advantage of these types of things that we're doing as a whole company.
Thanks, Matt. The next question's on enterprise networking. In addition to the OCTEON win you just described, it appears that you have two additional wins, one with an existing customer, one with a new customer. Can you provide some more color on both of them?
Sure. Why don't I have Raghib come up and give a comment about that? It would be great.
As I described during my presentation, one of the win is in one of the large North American tier 1 OEM. We have been working with them for a long time, and they have chosen to use our Switch+ platform, porting their own OS. Now, the reason it is a huge significance from the overall opportunity point of view, because being part of the integrated portfolio allows us to reach the broader market as well as it also extends our applicability to be adopted in additional designs. The second one is also very critical because this is also a tier one North American networking OEM. They never used our platform before, so this is the first time they adopted our Switch+ platform.
We have successfully executed it, as a result of that, now we have established our reputation and partnership, we are winning more design in the same product line. Both of these products are already going into production in initial ramp, we expect it to ramp in the next few years. It's going to drive our revenue and add more on top of our revenue. That is going to drive our growth in that networking. Thank you.
Very good. Thank you.
Thanks, Raghib. We'll take our last question on storage. Very similar to the product question, it sounds like you announced there's two wins in this space, one with a new NAND OEM customer, and the second one a new cloud DIY customer. Could you elaborate on both of them? What does this mean? When does it start to ramp?
I'll let Dan answer that one since he and his team worked extremely hard to generate these opportunities. Go ahead, Dan.
Thank you for the question. Yeah, in both cases on the new cloud data center customer, they're qualifying drives now. We expect ramping to start up next year. I think what's major, even more exciting, is this should be a multi-year ramp as the DIY model within their business expands and the percentage of drives they'll use that come from the Marvell solution. That is a major, tier one U.S.-based cloud data center customer. For the NAND vendor, there's a handful of NAND vendors. I can't say who it is. I will say it's a new customer for Marvell Storage. Again, that'll start ramping next year. It's a very meaningful design. It's a very large revenue design. We also believe this will lead to additional engagements with the same customer over time.
We're super excited about both opportunities and what it means for our SSD business.
Great.
Thank you.
Thank you, Dan. In closing, I just want to thank all of our investors, current investors, and shareholders, employees who are watching, all of our constituents. We really appreciate your support of the company and the journey that we've been on. It was our pleasure to tell our story today. As you can hopefully heard from our words, from the way that we're expressing our enthusiasm, that we've got something very special here at Marvell. We look forward to future outreach with our shareholders and continuing to tell you about our story and all the progress we're making. I hope everybody stays safe, stays healthy, and again, thank you very much for attending today's virtual Investor Day for Marvell. Thank you.