In The Arena by TechArena - Ken Sun of AMI on Firmware, Fragmentation & AI Infrastructure

Episode Date: August 25, 2026

In this episode of The Control Plane, host Allyson Klein is joined by Ken Sun, Corporate Vice President at AMI, a Lattice company, for a wide-ranging conversation about firmware's growing strategic importance across the AI infrastructure stack. Ken brings a perspective shaped by years across service providers and cloud infrastructure, and frames firmware as the connective layer binding together the five layers of the modern tech stack: energy, silicon, cloud, models, and applications.

Welcome to the arena. We are on a control clean episode and we are here with Ken's son of AMI. Welcome back to the program. Ken, how are you doing? I'm doing great, Allison. Nice to join you. So Ken, you've come to AMI recently after roles spanning across service providers and cloud infrastructure. What drew you to AMI at this particular moment in the industry? And what does the role look like from your vantage point? Yeah. So it's just a super-examination. exciting time to be in this space, right? And the era of agentic AI has just led to a massive acceleration of the AI infrastructure buildout. And, you know, we kind of refer to as the most consequential engineering undertaking in modern history. Right. And the impact of this

engineering effort is just enormous. For example, we're really dealing with the amount of investment flow that's like in the trillions, right? These numbers are just getting so large. Right. And And also just from a power perspective, which is obviously another huge challenge and opportunities, are equivalent to the size of cities now. And also the pace of innovation that collectively we're all working at is just incredible. It's at light speed. So for me, the opportunity I saw was just a really incredible time to enable the entire ecosystem, ranging from silicon to OEM, ODM, and hypers to transform.

and collaborate and co-innovate in new ways, right? And I feel that it's the right time because people are looking to collaborate differently because these engineering challenges are just never seen before, right? So again, it's just been super exciting to be part of this journey. And I look forward to helping AMI to really provide the go-to-market strategies and also from an ecosystem enablement view and also ensuring customer success. on this effort. Now, Ken, it's interesting.

firmware has always been known as kind of a quiet part of the industry where people don't really talk about it until something goes wrong. But firmware has been very talkative lately about the importance of the technology to the foundation of AI data centers. Tell me what's changed and what's driven this value chain focus on firmware. Yeah, you know, we used to joke you only know about firmware when something is wrong. But obviously, I think now it's really changed and often refer to what's talked about today as the five layers of the tech stack, right? You know we talk about how it starts with energy and then silicon layer and then cloud models and then the apps.

What I always say, it's these are the five layers, but what's not talked about, it's firmware, right? That's the glue in bringing these five layers together to work as one. I really see firmware, you know, and also platform software to some extent as critical pieces to help secure scale, manage and also enable sustainability of your entire AI infrastructure state. So again, I think these are the five layers that we're working with and firmware works with all five, right? And really ensure that there's a coherent, you know, connection across the layers and across the ecosystem to make sure we deliver on the expectations and outcomes of the. the end customer. I love this layer model because it kind of gets me into my next question, which is how AMI also sits at the intersect of all the value chain.

When you think about powering cooling vendors and the integration of IT and OT infrastructure, the silicon providers, OEMs and ODMs, hyperscalers and neoclods, and then even open source communities and industry standards efforts, AMI is really engaged across that value chain. When you look at that space and you have such a unique, purview on what's going on in the industry around data center infrastructure. What are you most excited about either from critical players that are coming forward or how companies are in collaboration around delivering new capability?

What do you see pushing you yourself to build things that you couldn't have built alone? Yeah, I mean, when you think about it across this industry, everyone is essentially solving for the same problem, right? And that is how do we deliver together on this massive demand that we talked about, right? And the questions are for sure around how do we innovate fast enough? Innovation alone is not enough. How do we scale the operation? And then how do we deploy at the pace that's required?

And I think in the past, and I know you've spoken to Sandra about the five-headed hydra. I really like the frame, by the way. It's really around you can't tackle any one of these challenges. on its own. And it seems like whenever you have control over one of them, another problem comes up. Right. And, you know, the problems are around fragmentation, around security, around power and thermal, around fleet management, and also just how to ensure this mission critical infrastructure is working right. And at AMI, we believe, again, the answer lies in firmware and also in transparency and open standards. I think that's one area where we've talked about a lot, where we do see open source as a key contributor to allow people to innovate on a common stack, right, and also to iterate faster on the stack.

But at the same time, we know it's not easy, right? Because it is open, often it actually leads to more fragmentation, right, and also more forking in this code base. So this is where we see AMI plays a role, right? Where we not only part of contribution, and we are one of the leading contributors to open source in the open compute project for whether that's open BMC or other aspects, but at the same time, AMI offers the services and support that's needed for enterprises and support them throughout the design and deployment and maintenance lifecycle, right?

So I think this is super exciting for me because one, we are leveraging open source and building on it and actually promoting it together with our ecosystem partners. But at the same time, we are part of the solution in helping address a lot of the fragmentation and security issues that we do see that are very much top of mind for our end customers. Again, to answer your question, I think this is super exciting, energizing. And just it really, it's a collective ecosystem effort, right, to get a handle around this challenge ahead. You know, I know that you were recently at Computex in Taipei with the team looking at what the ODMs and OEMs were featuring in terms of next generation designs and reference

architectures. What did you see in terms of trends and infrastructure innovation that made you lean in and say, hey, this is somewhere that AMI can really help? I'll say the first takeaway from Computex is just seeing the pace of innovation in action. I know we talked about it, but again, it's very different when you're there and you see these systems and just you compare it to what we saw last year. It is just incredible to see how things are becoming modularized and how things are becoming much easier to build and operate and deploy. And also how things are just becoming more compact, right?

you're definitely squeezing more computers in smaller real estate and we're finding ways to cool and drive thermal efficiencies. So again, I think that's probably one of the big takeaways, right? Again, to see this piece of innovation in action and seeing all the partners come together to deliver on this. And I'll say the other takeaway is just how there is commonality around the challenges that we're all solving for. So besides speed that, you know, I probably talked about quite a bit, I think that's the other

There's definitely themes around secure management and rack boots, how basically now you're booting and managing hundreds of nodes in Iraq or more across a pod, right? How do you do it efficiently? How do you do at scale and the innovation around that? I think, again, it's very exciting and to see how partners are coming together, brainstorming around solutions for that. And also just flexible control of power and cooling. again, it's not a surprise that cooling and power is no longer static, but very dynamic, right?

And actually, it's becoming probably one of the toughest parts of the equation that we're all solving for. So I think that part, again, the innovation and also just solving for that problem across hardware, firmware and software is super interesting in the collective discussions and how do we really have more of a solutions approach has been very refreshing coming out of computer. text. I would say two other topics that came up, you know, as top of mind is just around predictive maintenance or now predictive failure. I think the ultimate vision is now people are talking about data centers in space. I think it's actually by time late this week that we're talking about that. But again, the vision is around how do you now manage these systems

without human intervention? How do you really have self-awareness, self-healing? And also in some ways, how do you retrofit these systems, right? Without human intervention, how to retrofit, you know, old fleets that you want to keep upgrading and keep them in operations. It's a large topic, but again, people are zooming in on just different approaches. And am I being the key firmware provider, we're in all these conversations to ensure that we can help address this health of mind. And also just the last one, I'll close on, is for sure they continue to talk around accelerating time to market. I do see more discussions around just modular plug and play,

even from a hardware slash firmware perspective. Again, we're seeing now people continue to have more modularity around parts of the system level design. We know whether that's for security purposes or reuse and also just plug and play capabilities. Yeah. So I think just around these four areas of innovations are having. been super refreshing coming out of Computex

and innovating with our partners around that. You know, one interesting thing about your career trajectory in particular is that you've sat at different places in the value chain from systems to services and know into firmware. What lessons did you learn along the way in those other spots that help guide you in terms of your partner strategy within AMI?

The lessons, I would say, it goes back to collaboration, right, in order to scale. In the beginning, I talked about we are solving for engineering challenges that are probably the most consequential in history. So I don't think any one part of us have a complete solution or understanding. So I think the fact that the ecosystem is very much open and collaborative is really critical, right? Because especially when these problem statements continue to be,

reshaped. So yeah, I think it's not only IT and OT coming together, but coming together and just being able to build on each other's perspectives and keep a very open mind in how we approach these problems and challenges in different ways. Right. So again, I think it's being able to lean on each other, build on common platforms, building our open platforms, and then moving fast together. So I think it's something that I bring in, you know, from the different backgrounds and ecosystems I've worked with. But I think now it's really being put in action with this AI infrastructure build out. I've been in the data center arena for 25 years. I think that we haven't seen this pace of change and pace of innovation before.

And the workloads themselves are getting more and more complex. How does that impact the way that those traditional forms of collaboration change? And what is the pressure on the industry and the entire value chain around the pace of innovation to drive to this AI objective? In many ways, right? And I would say, first of all, customers want to know how firmware is integrated with their fleet tools, right? And how quickly they can patch thousands of nodes and how to verify trust across the entire platform. So I think one, it starts with telemetry, right? And, you know, I think many of the operators talk about that's where everything starts, right?

how do you really get the signals and telemetries? So you can build this self-awareness and self-healing aspects of their operations. And that's really foundational, right, to their AI workloads and their infrastructure operations. At the same time, these systems need to be mission critical, right? Which, again, the expectations for uptime is expectations for security are at the highest levels. And they continue to be higher. as the data and privacy becomes even more important as we, you know, feed more data into these AI systems.

I'll say the second thing I call out in working on these workloads is I think the pain point for customers really around power and cooling is unprecedented, right? And really, this is now the new equation to, I'll say, the industry, right? And these are now multi-kilot boards in tens of thousands of kilowatts, you know, per rack. So liquid cooling become table stakes. And also, again, the telemetry and density are just exploded, right? And that's to ensure that we can operate these more complex systems in real time. And also drive the sustainability and efficiencies that we expect from these systems. And also, in addition, to support this innovation, just the innovation around new accelerators and cooling technologies

continue to speed up, right? These cycles continue to speed up. And for whether it's an operator or a system designer, it's very challenging, right, to spin up new designs and new control models for each new generation. So, yeah, I mean, again, this is where AMI, as a partner, we play a role. We position ourselves as part of the solution

where we can innovate together so they can keep up with the pace of innovation that's required. And I'll say one final area where we're really leaning in to help address some of these challenges is around regulatory and complies demands, which again adds to the complexity of the problem, right, as if it's not already tough enough. We're seeing obviously in regions of the world where not only the requirement for security is there, but there's regulatory acts, for example, with CRA, with Cyber Resiliency Act, where it's no longer just about writing secure code. But it's really a required processes for vulnerability management, for firmware system updates, for documentation, and most importantly, incident response, right? With security incidents do have throughout the product lifecycle. Again, I think this is just an example of areas where AMI plays a role in helping to contribute to the ecosystem, whether that's just with firmware and tools or around cooling thermal challenges or around regulatory and compliance needs. So I'm glad you brought up security.

You know, it's moving farther down the stack every year. How do you look at security and the responsibility of AMI's contributions is a platform foundation for security? When you look at things like secure firmware baselines or out of station or compliance, how do you balance your own innovation with open standards and what your customers are demanding of you? Well, yeah, I mean, first of all, security has shifted from being a design checkbox. to just board level mandate. And also shows up in every aspect of security, whether that's hardware, that's up to your system stack,

and also to the models, how to ensure that it's tested and verified. And from our view, we think every component now needs hardware route of trust, attestation, increasingly even crypto agility to address post quantum requirements. The security pressure is rising because the attack surface is rising with it, especially when you're not addressing the fragmentation challenge, right? Again, I think the two come hand in hand, right, because of the innovation around silicon, around the different operating models are so fast. We do see a fragmentation of the approach.

And when that's happening, the security risk becomes increased, right? I mean, that's just what happens when you have an inconsistent and unverified approach across a mixed fleet. So here, again, here's where AMI plays a role in, really taking a holistic approach with our security offering. And at the same time, I mentioned before, having a unified and open source code base with our Megarack community edition approach, right, where we're taking these different pieces of innovation in open source and streamlining it and upstreaming it, right?

So I think that's really the beginning of the journey, right? Where we take a more holistic and unified approach. So we become the solution to the fragmentation. In the same time, we're building our security IP and capabilities all the way up the security staff. Right. So wrapping this up, I would say from a customer conversation view, we're seeing that when we lead with this approach, the conversation is no longer about, okay, which technology or which route of trust to pick. But it's really become around, okay, do they have a trusted partner to work with, and not just a supplier? Right.

And ultimately, this is the role that as AMI, we want to play. Again, there's many solutions out there, right? And people can pick and choose. But we think in the end, that doesn't really help solve the challenge, right? It doesn't help solve the fraction issue. It doesn't help solve the approach to the security challenges. And as AMI, we want to beat that trusted partner for our customers. Ken, it's been an awesome conversation.

I just have one final question for you. We have so much innovation happening in this space. where can folks find out more about AMI's own trajectory in this space and engage in collaboration with the team? Yeah, so we're actively working with all of our ecosystem partners. And we've had that approach over the last 40 years, whether that's working with our Silicon partners upstream or downstream with OEMODMs, also with operators. In our mind, it goes back to really being focused on addressing. the end challenges of the customer, right? Because again, I think it's very aligned with what I mentioned before.

It's a very dynamic and large ecosystem of partners that we work with. But we're all solving for the same challenge. So being razor focused on just the top pain points and also providing solutions that are most impactful for our customers in addressing those point points are super important to us as AMI. And we really see the control plane as the strategic layer. of the modern data center of addressing power, thermal security

and regular demands that are changing faster than ever. Right. So here's where AMI really keeps a very open approach in terms of co-innovating with all of our partners. And I say that in both a technical and a commercial perspective, so that we can provide a well-integrated solution.

And at the same time, I would say a very integrated and adaptable approach to the different parts of the, the value chain. So here's where kind of my role is to really combine our work on the product side and solution side and also with the go to market to ensure that we are translating our innovation to customer value in delivering it to our end customers together. So yeah, I'm super excited to be working with the entire ecosystem on this. And we have programs in place where we can do this at scale. I am very much, you know, welcome.

on the participation from our ecosystem partners. Thank you so much, Ken. It's been a pleasure talking to you. Can't wait to have you back on the show in the future. Yeah, thank you so much. Alison, for your time and look forward to it. Thank you.