Semiconductor Insiders - Podcast EP269: A Broad View of Semiconductor Market Dynamics with Rajiv Khemani

Episode Date: January 10, 2025

Dan is joined by Rajiv Khemani, a serial entrepreneur and an industry leader with 25 years of experience in building and scaling businesses. He is currently co-founder & CEO of Auradine. He is also an investor and board member in the data infrastructure, AI and software/platforms space. Previously, Rajiv was co-founder &… Read More

Hello, my name is Daniel Nenny, founder of SemiWiki, the open forum for semiconductor professionals. Welcome to the Semiconductor Insiders podcast series. If you have a topic you'd like us to cover, please post it on semiwiki.com and we'll get right to it. My guest today is Rajiv Kamani, a serial entrepreneur and industry leader with 25 years of experience in building and scaling businesses. He is currently co-founder and CEO of Oradine, a web infrastructure company. He is also an investor and board member in the data infrastructure, AI and software platform space. Previously, Rajiv worked as co-founder and CEO of Innovium, a leading provider of cloud-optimized network switching solutions that was acquired by Marvell for over a billion dollars in 2021. Prior to that, Rajiv was COO at Cavium.

He also worked at Intel, Sun Microsystems, and Network Compliance. Welcome to the podcast, Rajiv. Thank you, Daniel. It's a pleasure to be talking to you today. The first question I'd like to ask is, how did you get started in semiconductors? Do you have an interesting story to share? I do. When I was going to college, both for my undergrad and master's, frankly, I was not that interested in EE and not interested in semiconductors.

I was really more focused on software. So I started my first job after I finished my master's as a software engineer at Sun Microsystems, which was a hot company at the beginning of the computing era. And after I did Sun, I went to business school and I wanted to get into business side of things. I was later on working at a company called NetApp, did an acquisition there, and the CEO of that company enjoyed working with me

and hired me into it. And that company was supposed to be a security appliance or a system company now as part of that the company built a asic technology and you could say that that was one of the first network processors silicon but our goal was not to sell chips but was to sell systems and security solutions as it turned out in the late 90s, what happened is that we found a lot of interest in customers wanting to use that chip instead of the system or the solution. And we started, we morphed that company into selling chips and board level products, eventually getting bought by Intel. So you can imagine that I got into semiconductors almost by chance.

Those days, Intel was an amazing company. I learned just incredible amount of stuff at Intel and slowly got, if you will, excited and interested in semiconductors, really enjoyed that experience. And that's how I got into semiconductors. So really, it was by chance that fate took me into semiconductors, even though I wasn't planning to do so. Interesting story. So let's talk about the semiconductor industry a bit. What are the risks of monopolization in the semiconductor industry, and how do they affect innovation and competition? You know, the natural tendency of semiconductors in almost all areas,

with the exception of certain commodity components, is that there is a huge amount of upfront investment needed. And then you have to make bets that play out over a long period of time. So whether you look at the x86 architecture or you look at what's happened more recently with NVIDIA's AI, it has taken, it takes tremendous amount of capital and long amounts of time to get to a point where it takes off in a big, big, big way. And that actually ends up being a monopoly-like situation. All these companies will argue they're not. But what it does is that it actually locks in companies into these vendors. And it also makes it very hard for new companies to enter the space. And that actually is not a good thing. Monopolization, not just in the semiconductor industry, but in any part of the overall economy is a bad thing. And so when this monopoly,

you know, the innovation slows down, the prices are very high. And essentially, people effectively go to customers and force them to use their products instead of using alternative solutions. So it's really a bad thing for consumers. It's a bad thing for the overall economy. Right. So as a follow on, what's your take on the future of AI and the semiconductor infrastructure? How can the industry ensure innovation thrives

without falling into a monopolistic situation? You know, I think the way it has happened, I mean, we can look at history and learn from it. We can look at first principles and learn from it. They are tremendous guides in figuring out how to deal with these situations. Clearly, regulation plays a big role in this, and that's why the FTC exists as an example. But I think if you go back in the history of the computer industry, it started out with companies like IBM that built amazing products that were all vertically integrated, and frankly, drove a lot of the adoption of technology in enterprises and so forth. So they used to build, for example,

chips. They used to build systems. IBM wrote the OS for those systems. They also did the DB2 database, and in many cases wrote applications. So they wrote the full vertically integrated stack and brought it into market. The way that got broken, that monopolistic hold got broken was that a set of companies together over several, several years built an alternative, what I would say, open stack that really led to the tremendous innovation we have enjoyed in the last few decades. And the way that happened was that, you know, there was the Intel architecture, the x86 became a standard, but it became, you know, standard interfaces. So things like PCI Express, things like Ethernet happened. And all of these interfaces allowed innovation

to thrive around them and across various companies. Later on, Unix happened and Linux happened, and that broke the stranglehold of companies that would do vertically integrated software hardware and enabled a broad range of applications to happen without the lock of the vendor. And I think that's how it will happen in the future as well. And we see that brand new innovation happens in vertically integrated stacks. And then there is a open ecosystem that develops that allows for innovation to thrive. Yeah, I agree. You know, one of the questions I have about the AI infrastructure is, you know,

are we overbuilding? You know, we're building a huge amount of AI infrastructure. Everybody wants data centers in their own backyard. Yet the revenue from AI, you know, generative AI and stuff maybe isn't there yet you ask anyone, they will tell you that this is going to transform how we live, work, and play, right? So there's no debate on that. records. So people see that tremendous potential. But what happens with any new exciting technology, and again, looking at history and from first principles, people tend to overestimate its value in the short term, and they underestimate its value in the long term, right? And so is there incredible amount of infrastructure built? You could argue, yes, there there is but I would say that this infrastructure that that's been built is not the best infrastructure that we will see five or ten years from now it's still suboptimal in many many ways whether it's cost whether it's power and whether it's capabilities and so what I see is that again going back to like the beginning of Internet, there was an incredible Internet infrastructure build out.

And five years after Internet started, there was a big crash that happened. Will we see such a crash in AI? Maybe. Will it be five years or is it shorter? You know, we don't know exactly, but there could be a correction. But I think in the long term, what we are building is actually going to be many times what exists today. Yeah, I agree. Power is my big concern, but we are addressing it. So, Rajiv, what advice would you offer? I mean, as an entrepreneur, what advice would you offer to startups and innovators looking to navigate the complexities of the semiconductor ecosystem today?

The semiconductor ecosystem is not an easy place for new entrepreneurs, right? You cannot have like three guys coming out of undergrad or dropping out and building a chip company at the leading edge semiconductors. It's very rare, if any okay and so semiconductor entrepreneurs typically have had to demonstrate that they have a team as opposed to just by themselves they have to demonstrate a track record you also have to have relationships with large customers. In general, semiconductor companies sell to tens of customers, not millions of customers, because these companies actually build the systems of one form or another and then either sell it or deploy it if you're a cloud vendor, for example. So you need to have some set of relationships. You need to be able to do a significant amount of fundraising on day one. That becomes an important thing because the upfront costs of taping out chips and leading-edge process technologies is very, very high.

So I would say that for leading-edge technologies, especially in the digital domain, you need all of those things. Now, there are exceptions to that. And those exceptions are typically with especially with regard to capital are in, you know, if you're doing analog stuff, if you're just doing IP and just going to do test chips rather than silicon sales. Or if you're in lagging edge process technologies, you may be able to get by without a lot of capital and be able to bootstrap it with less money. So those are things to think about as you are looking at being an entrepreneur in the semiconductor space. Right.

Yeah, I've done startups most of my career. You're absolutely right. It doesn't get easier, that's for sure. So how do you see U.S. regulation impacting the semiconductor industry today and, you know, in the coming months with the new administration? Well, I mean, just to step back, I think U.S. has been at the forefront of semiconductor innovation as far as I can tell. OK. And especially when there's tremendous innovation, it ends up being mostly in the U.S. OK. Once things get commoditized, you know, there's other countries like DRAM actually used to be in the U.S. And now it's, you know, it's in Korea and Japan and various other places.

The U.S. regulation, I think, is very important. and Japan and various other places. The US regulation I think is very important. Semiconductors have now become the oil of the new economy. And they are incredibly critical for consumers, for our lives, for our security. You know, they are incredibly important for climate change, and so on and so forth. Some of the biggest initiatives that we all care about, they've become very important. In a world where for the last 10-20 years, free markets and free trades drove a lot of what I would say globalization, you know, what people have started to realize that geopolitics is important. And if you're too dependent on any product or technology from a country that you

view as potentially competitive in the long term. And they have started to do it. I think it's still moving slow. And you can do a combination of incentives to companies to do this. You can do it through a combination of tariffs that President Trump likes a lot, or sometimes he threatens tariffs to bring people to the negotiating table. So I think that's very important. And I expect that the new administration, and frankly, any administration would look at this very seriously to drive semiconductor innovation and manufacturing back to the U.S. and to its friendly partners. So what regulation would you suggest the U.S. adopt to support domestic semiconductor manufacturing? I think that the answer varies depending on the types of products. If it's really a commodity and the sources are many

across many parts of the world, then you can let free markets play out, if you will. In segments of the industry where there is concentration among suppliers in geographies. That's where the U.S. needs to pay attention. You know, the ability to build these things with meaningful capacity to address local demand is very, very critical. And so in those areas, regulation becomes even more important. Now, what types of regulation you asked and I would say the regulation you got to make sure that you know you can provide incentives to local manufacturers either through tax benefits or through government funding although we all know that the new administration doesn't like to fund things they want to cut government and not increase government. And then tariffs play an important role in that as well. So it has to be thought through very

strategically, segment by segment on how to make it happen. Now, there is yet another way, which I think that this new administration is going to employ, which is to go to the large companies and really push them towards bringing manufacturing. So, for example, going to some large system companies and saying you ought to purchase a certain number of your semiconductors from U.S. manufactured locations. And that will become an important part of the government policy to make all of this stuff happen. Yeah, yeah, I agree. So just one last question, the CHIPS Act. Do you think that was worthwhile? Do you think there should be a CHIPS Act too? What's your opinion? I think certainly it was worthwhile. I mean, it certainly galvanized the industry to, it was a carrot that everybody ran towards.

You know, it has driven a lot of things in U.S. semiconductor manufacturing. I would say we are still in early stages. I mean, the news that we just saw where, you know, Intel CEO announced retirement is not a very positive sign for the industry in terms of bringing manufacturing back to the US. So more work needs to be done on that. And so one could argue that the CHIPS Act was a little slow in its execution. And I expect, I, you know, personally, I'm supportive of the CHIPS Act and of bringing manufacturing here to the US in meaningful quantities. Whether there should be a CHIPS Act too of bringing manufacturing here to the US in meaningful quantities. Whether there should be a CHIPS Act too, frankly, I don't know. I'm not an expert in that.

I personally am not a believer of very large government myself. And if there are ways where private investors can be incentivized to make it happen, that would be an excellent way to do it because it's just more efficient way of doing things. Great. How about we close with just a 30-second elevator pitch for your company, Auradine? What is it you guys do? So Auradine, we provide semiconductors, systems, and software for two areas, blockchain infrastructure and the AI infrastructure. And we build our products with three things in mind. The first is innovative capabilities.

The second is sustainability. And the third is scalability. We want to build products that are super energy efficient, that drive the planet's objective towards lower power and drives the planet's objectives towards greater productivity and intelligence available to all. Within those two, we are building products that do data center scale Bitcoin mining. And that's a Teraflux product line that is shipping at high volume, and we're shipping three nanometer products in production right now. And then the AI area is we are building an AI network fabric solution based on open standards, which, as I described to you earlier, is very important for innovation in the long term. And our goal is to drive greater intelligence,

lower power consumption in a very, very cost effective and open fashion. And so that's our links initiative. We are part of open industry consortiums as part of that, the UA Link and the UEC, which are exciting new consortiums that a lot of the industry has coalesced towards. And we are super excited to build a great company driving the infrastructure for the next generation technologies.

Great. I'll put a link to your company in the header. And it's great meeting you. And thank you for the the time and hopefully we can speak again next year. Absolutely. Thank you so much, Daniel. That concludes our podcast. Thank you all for listening and have a great day.