Protecting Innovation in the AI Infrastructure Industry

Justin Pierce: 00:13

If you're a business leader or general counsel, you already know AI isn't just another tech trend. It's the next frontier reimagining and reshaping how companies operate and compete. With that transformation comes complexity around intellectual property, data rights, regulatory oversight, litigation exposure, and brand integrity. This podcast is designed to give you clarity. I'm Justin Pierce, cochair of Venable's intellectual property division.

Justin Pierce: 00:43

In this season, I'll talk with colleagues and industry leaders about how AI is reshaping business, ecommerce platforms, toy companies, industrial automation, luxury brands, beyond, and the legal strategies companies need to protect innovation while moving fast. Our goal is simple, to help you turn our legal insight into your competitive advantage. Welcome to season two of AI and IP: The Legal Frontier. This week, we're talking about AI and the semiconductor industry. Hi.

Justin Pierce: 01:24

I'm Justin Pierce, a partner at Venable and co chair of the firm's intellectual property division. As artificial intelligence accelerates innovation in the AI infrastructure and semiconductor industries, companies are rethinking how they design, develop and protect increasingly complex technologies. From AI enabled chip design and data centers to faster time to market pressures, the intersection of hardware, software, and data is reshaping both engineering workflows and intellectual property strategy. At the same time, legal and research and development teams must navigate evolving questions around inventorship, confidentiality, and the balance between patent protection and trade secrets. In this episode, we explore how AI is transforming semiconductor innovation, how in house legal teams are adapting their IP strategies, and what it takes to stay competitive in a rapidly evolving technological landscape.

Justin Pierce: 02:27

I'm talking today to Kelvin Vivian, associate vice president and Deputy General Counsel of Intellectual Property at Marvell Semiconductor. Welcome Kelvin.

Kelvin Vivan: 02:38

Yeah. Great to be here, Justin.

Justin Pierce: 02:39

Awesome. Kelvin, I'm looking forward to a good discussion. I think you've got an interesting background that our listeners and audience would want to hear. Can you tell us a little bit about how you got started in IP law?

Kelvin Vivan: 02:52

Yeah. I started off basically planning to be an engineer. So I studied electrical engineering and computer science at, UC Berkeley. And then I was able to get an internship while I was still an undergrad at a aerospace engineering company and met a patent attorney while I was interning, and he had an electrical engineering background and planted some seeds. So then went to law school. So that's how it started.

Justin Pierce: 03:16

That's great. So engineering before law school, that's a path that I've heard from some others where you learn in really the first stage of your first job before becoming an attorney. That's great to hear.

Kelvin Vivan: 03:28

Yeah. The longer you spend in engineering should be okay. I spent one year. Some spend ten years or more, but it that's fine too. You can always transition later.

Justin Pierce: 03:41

Right. So you went on to law school. What'd you do after graduating from law school?

Kelvin Vivan: 03:48

So after law school, I started with a law firm, Fish & Richardson. I was there for a number of years and then transitioned to another smaller patent boutique, Sawyer Law Group. And then from there, I went in house.

Justin Pierce: 04:04

So a couple stints as outside counsel then in house and you've been in house since at Marvell the whole time?

Kelvin Vivan: 04:11

That's right. Eighteen years now.

Justin Pierce: 04:15

By many standards these days, that's a long time at one company. Given that in your current role, do I understand it that you are the VP or lead executive over the IP function?

Kelvin Vivan: 04:29

That is correct. I'm associate vice president at Marvell Semiconductor, managing the legal function, the intellectual property legal function under our general counsel.

Justin Pierce: 04:38

Okay. Great. And what do you do in that role?

Kelvin Vivan: 04:41

Yeah. Basically, I oversee all IP related functions that includes patents, trademarks, the IP related litigations and licensing. Open source is also included and any other random intellectual property matters that may come across. And we also interface with our legal colleagues, counterparts with respect to commercial contracts, mergers and acquisitions and other aspects.

Justin Pierce: 05:09

So at the top of the session or right before we began recording, we were talking a little bit about just how things in your industry have changed in recent years. I assume in your current role, you've had a front seat to all the disruption and excitement caused by AI.

Kelvin Vivan: 05:28

Definitely. We're seeing it not only within the legal department, but our engineering teams as well because, yeah, there's just semiconductor industry is seeing unprecedented design complexity changes and and shorter development cycles. So you definitely need to have a tool to help you just stay competitive.

Justin Pierce: 05:50

Sure. And thinking about the business of Marvell, you being there eighteen years, you've probably seen change at the company and industry level. I know traditionally it's been thought of as a company that's a player in the semiconductor industry. What do you make of this more recent use of the term of AI infrastructure and what role does Marvell play in that?

Kelvin Vivan: 06:13

So AI infrastructure is the term that's generally used to define semiconductor products and and other technologies that are required to just deploy AI solutions. And that could include this full stack computing environments, like you have your GPUs, you have your connectivity. Those are important aspects for connecting a lot of these and enabling the AI technologies to do inference and scale.

Justin Pierce: 06:45

So in your role and given the vantage point you have, I'd love to hear some more on how you've seen AI be adopted just in the industry overall, then maybe we'll talk more about the company. But focusing in on the industry, what are some of the biggest changes you've seen happen as a result of the popularity and demand surrounding AI products and services?

Kelvin Vivan: 07:10

So in general, I'll just say with respect to the semiconductor industry, just talking in general, I'd say you'd probably see AI tools being more utilized in design and verification workflows. So designing of the chips and in general, just analytics. You know, when whatever job or your task you're doing it's nice to have information and data processed and move down into a form where you can easily compress and work with that data. And I'd say it increases your productivity.

Kelvin Vivan: 07:46

So that's how I'm seeing it being used on seeing the engineering front. And on the legal front, there are tasks that we do. Like, say, for example, in the context of mergers and acquisitions, there's just a lot of contracts that need to be reviewed.

Justin Pierce: 08:00

Right.

Kelvin Vivan: 08:01

So as a force multiplier, you can you can enable tools to help you summarize agreements. But, obviously, there needs to be a human in the loop. We cannot just rely on the output of these tools. There's a been a fear of, hey, will these AI tools replace us? And I'd say, no. That would be one of my main takeaways. These AI tools will not replace us. Maybe some functions, but I see the use of these tools as being force multipliers.

Justin Pierce: 08:33

Got it. So somewhat amplifying, not replacing. It's going to help augment and amplify what you are already doing or what you need to do, for instance, in your legal capacity.

Kelvin Vivan: 08:44

That is correct.

Justin Pierce: 08:45

Okay. So thinking more about your company and what Marvell is doing, what are some of the business segments or business areas that the company is producing products and services for?

Kelvin Vivan: 08:58

So I'd say for data centers, hyperscalers, for sure. But basically, it's all areas of interconnect. So for example, GPUs. You need to be able to connect GPUs together. So there's what we call scale up networking. That's where you're connecting GPUs together. And then there's scale out networking. What is that? That's connecting racks of GPUs to other racks of GPUs. So that's scale out.

Kelvin Vivan: 09:25

And then there's scale across. Connecting different parts of a data center together within a farther reaches of a data center. And then there's scale out. You need to connect data centers together. You know, that could be across different regions.

Justin Pierce: 09:44

Right.

Kelvin Vivan: 09:45

And so those are the different aspects of connectivity that are needed to support what is called AI infrastructure. And that's what Marvell provides. We provide all the different connectivity points as well as the accelerators, data processing accelerators to enable these GPUs to manage and accelerate their processing.

Justin Pierce: 10:08

Right. So clearly, your products and services are integral for data centers and connectivity on any sort of wide scale. It's good to hear. With that, I'm wondering if you have any success stories you can share in terms of the use of AI. Maybe thinking of it in two ways, maybe any success stories you could share in terms of how the technology has been used either in what you do or your team does, and then maybe we can think about larger, you know, company or industry.

Kelvin Vivan: 10:39

Yeah. One small simple example that I'll use, just a personal example. Dealing with an Excel sheet. I had a large number volume of data where the information, like, say, names, was in a format: first, last. And I needed that information, a large volume of data. I needed it to be last, first.

Justin Pierce: 11:05

Got it.

Kelvin Vivan: 11:05

Just prompted the tool. Hey. Can you do this? Just a few prompts and within a minute it's done. These are simple, easy tasks but would require a lot of manual typing or human level work. But having a tool just program it, code it, implement it, done in minutes.

Justin Pierce: 11:28

Right.

Kelvin Vivan: 11:29

That's just a simple exercise. But in terms of just taking a large amount of volume and data and just condensing it into a manner that's digestible quickly is great.

Justin Pierce: 11:42

Any others at more of a departmental or group function level that you can think of?

Kelvin Vivan: 11:49

Yeah. I'd say, well, in general, you just contract drafting and such. Just coming up with just different thoughts that you angles and perspectives that you may not have considered when you're reviewing an agreement, looking for any holes or gaps. Again, it's a tool assist. So there are tools out there that can do that.

Kelvin Vivan: 12:10

And to the extent that you're able to leverage those, it may lessen your need to rely on outside counsel. That's another thing. But again, I'd say it's not a replacement. It's more of just increasing the speed at which we're able to get to a reasonable resolution. Because I may do something and come up with a solution and then provide it to outside counsel and say, hey, what do you think of this?

Justin Pierce: 12:39

Have you seen examples where AI tools have already reduced reliance on outside counsel or perhaps helped you frame issues better to give to outside counsel?

Kelvin Vivan: 12:53

I'd say it's helped frame issues so that we can quickly arrive at a solution when discussing an issue with outside counsel. So typically, sometimes you may just lob the ball over the fence. Outside counsel, what do you think about this? But now, we can get a good idea of what an outside counsel may reply.

Justin Pierce: 13:23

Right.

Kelvin Vivan: 13:25

And we would like a human in the loop to just verify that. So I wouldn't just not rely on assistance of outside counsel when needed.

Justin Pierce: 13:35

Certainly.

Narrator: 13:37

Building a beauty or wellness brand takes more than great products and marketing. On Venable's Beauty Law Glow-Up podcast, partners Claudia Lewis and Kristen Ruisi explore the trends, challenges, and opportunities shaping the industry today. From influencer marketing and product safety to privacy, trademarks, and brand growth, each episode delivers practical insights to help companies build, protect, and grow their brand with confidence. Search for The Beauty Law Glow-Up wherever you listen.

Justin Pierce: 14:16

We've talked success stories there in your department inside the company. I'm curious on a bigger picture within your industry, maybe in talks or your dealings with other in house counsel in the infrastructure space or semiconductor space. Any other examples at an industry level where AI is help being helpful?

Kelvin Vivan: 14:39

I'd say it's in the reducing the time to market. Again, design complexity is increasing and the time to market is key. Anything that you can do to shorten your time to market and shorten design cycles, verify. Verification is another aspect of trying to verify whether these chips work according to your design.

Justin Pierce: 15:07

Right.

Kelvin Vivan: 15:08

But verification plans are becoming largely more complex as these chips are becoming more complex. So using these tools allows us to shorten that time. And I'd say just it's important to everybody's generally using these tools to do so. So if you're not, then that puts you at a disadvantage.

Justin Pierce: 15:29

Your focus on time to market or time to produce something that's getting increasingly complex is an interesting thread. I want to explore that a little bit. One area that makes me think, just given that we're both in the IP area and we've worked with teams who are in R&D or production specific, how much have you seen guidance being given? And this is just in general around issues like inventorship. If you are in R&D, or let's say you're part of that creative team that's coming up with a new product or a new semiconductor, how much guidance are those teams being given nowadays in this world where case law is relatively recent, but with this challenge legally that if something is created by AI and not a human, you wouldn't be able to claim patent protection or ownership as a company for that.

Kelvin Vivan: 16:25

I think just talking generally, you need to understand what AI tools are being used and when. So that's the first order of things. Once you have that, then you can develop a strategy for how you want, if you can, how you want to protect the information that's generated using an AI tool.

Justin Pierce: 16:46

Right.

Kelvin Vivan: 16:47

There certainly is guidance that's required to teams and a lot of companies now have generative AI policies that they share with their communities, their engineering communities. Even with respect to patent innovations and patent development, we certainly need to know if an AI tool was used and how. And there are disclosure requirements with respect to the USPTO to the extent that if how is AI used utilized in the development of this invention.

Justin Pierce: 17:20

Another question on that. Given the speed of development, you even referenced earlier how important it's becoming to get something from conception or idea stage to market. Given the speed of that seems to be outpacing the time it takes to get a granted patent, particularly in the software and AI spaces. How much do you see companies relying on trade secrets as opposed to patenting, for instance, given the rise of AI?

Kelvin Vivan: 17:49

Yeah. That's a good point. So again, there are a lot of different factors on in deciding whether you wanna keep something as a trade secret or decide to protect it through a patent. The fundamentals of however companies decided to keep something trade secret or not. I don't really see it changing too much.

Justin Pierce: 18:06

Okay.

Kelvin Vivan: 18:06

Actually, I don't. I don't see it changing too much, your core analysis that you would implement.

Justin Pierce: 18:13

And you think that's because the regardless of the timing, same factors apply. I definitely get that. What do you think about just the timing though of the fact that it takes a long time to get certain patents issued. Is it better in that case to push more into what I call the trade secret budget or trade secret effort as opposed to patent or not really?

Kelvin Vivan: 18:40

If you're going to keep something as a trade secret, you're going for a longer time of protection. And hopefully it's something that will not be readily discovered.

Justin Pierce: 18:53

Right.

Kelvin Vivan: 18:54

So potentially, if some things in the software space are taking a longer time to patent, but eventually you're successful in getting a patent, then I think that's fine so long as you're not wanting to enforce that patent in a time frame that's sooner. That's where I'd come down on that.

Justin Pierce: 19:16

Got it. So enforcement is one of the key considerations there.

Kelvin Vivan: 19:21

Yes.

Justin Pierce: 19:22

Okay. Interesting. So we've talked a good bit about some of the ways AI is used, talked about the industry in a good sense. Stepping back, just given your views and your experience, I do want to talk a little bit about opportunities and risks. Are there any particular opportunities or risks that really stand out for you when it comes to AI or where AI is going?

Kelvin Vivan: 19:51

The opportunities are great in terms of increasing your workflow, being able to do more in the day, regarding your everyday work. But the risks are also out there. And there are stories we've heard of situations where confidential information was lost out into the public domain. Or there's been a reliance on case law and briefs submitted to courts that were completely hallucinations.

Justin Pierce: 20:24

Right.

Kelvin Vivan: 20:26

So there are risks out there. With respect to loss of confidential information, you just need to kind of understand what tool you're using, understanding all the safeguards, if any, that are being utilized with respect to this particular tool you're using. And in the second case, dealing with hallucinations, you definitely need a human in the loop.

Justin Pierce: 20:52

I agree. So we've covered a number of different topics today. And one thing I like to do towards the end of our discussions, Kelvin, is think about or at least leave our audience with a takeaway or insight. One that I see as I look at our discussion and think through some of the things that we talked about today, is really twofold. On the one hand, particularly with the industry that Marvell is in, where things are fast moving and time to execution, time to market is important.

Justin Pierce: 21:23

That consideration in the world of AI between what you patent versus trade secret seems very important to me. And then I think also in an industry like this where creativity moves fast, but it's important to understand where that creativity was born from, whether man or machine, some of the different things you spoke about, like having good records as to who the inventor was, the efforts behind creativity are important to have as we move forward. How about you? A takeaway or insight for the audience?

Kelvin Vivan: 22:01

I'd say AI tools are becoming more predominantly used. It's the now, not just the future. It's the now. And it's important to understand how these tools are being used by team members across an enterprise, your company, so that you can then understand and give the best guidance that you can for how innovation can be protected and set your company up for success.

Justin Pierce: 22:29

Well, thank you, Kelvin. Great discussion.

Kelvin Vivan: 22:32

Thank you, Justin.

Justin Pierce: 22:33

Okay. That's all we have time for today. I want to thank Kelvin Vivian for helping us understand how AI is transforming AI infrastructure and semiconductor innovation, how in house teams are adapting their IP strategies, and why strong governance and human oversight remain essential in protecting innovation. You can read more about how Venable is helping businesses navigate the AI frontier by visiting venable.com/ai.

Justin Pierce: 23:03

I'm Justin Pierce. Thanks for listening to AI and IP: The Legal Frontier.