Advancing AI 2026 | Sagi Paz, AMD

Clips
More from Advancing AI 2026

Sagi Paz

Head of AMD Ventures

AMD

play_circle_outline AMD Ventures Thesis: Invest Where AMD Adds Unique Ecosystem Value — Strong Teams, Differentiated Products, Large Markets

play_circle_outline Stage-agnostic investing with checks from ~$1M to tens of millions

play_circle_outline Preference for software investments (~80%) over potentially competitive hardware

Info

Sagi Paz, AMD

Sagi Paz

Head of AMD Ventures AMD

search

(INTRO)

John Furrier

>> Welcome back to theCUBE's live coverage here in San Francisco of AMD Advancing AI. This is going to be the show that sets the agenda as the AI infrastructure continues to build out, of course, it's bounded by energy and capital. Our next guest runs AMD Ventures, Sagi Paz is here. Thanks for coming on. you're writing a lot of checks enabling ecosystem partners, startups, big companies, to align with the major shift.

Sagi Paz

>> Yeah.

John Furrier

>> Whole other generation of system architecture coming online with AMD and AI infrastructure, large scale deployments of clouds, neo -clouds, data center modernization, and a massive growth in open source, which is developers, physical AI, the portfolio laying out here. What's your investment thesis? Everything that moves in AI infrastructure? What is the focus of AMD Ventures? because this is a critical enablement for you guys. you have some successes. Take us through what your approach is and what your focus is.

Sagi Paz

>> Yeah, first of all, AI is a team sport. So there is many different layers across the stack. So it starts with the compute layer that AMD really can deliver on. But in order to provide value to customers, you actually need to have the AI model layer, you need to have the application layer, you need to have the software. And where we focus on is around the AMD ecosystem. So whenever we have companies that AMD is uniquely positioned to add value to, that's where we're getting excited about making investments. But it's not just around the strategic alignment. We definitely want to focus on the best companies out there. So whenever we do make an investment, we start with the fundamentals. So each investment needs to stand on its own. It needs to be a very credible company, great team, differentiated product, big market. But then having the angle in which AMD can provide some unique value to the company is where it gets us really, really excited, and that's where we make investments.

Dave Vellante

>> So staying with the investment thesis for a moment. We John and I did a lot of research this week with our team, and we laid out this three -layer architecture. Organic R &D, innovation, critical, where you have gaps. Gordon does his M &A thing, obviously. you guys have been very acquisitive. And then the third is the ecosystem, what you just mentioned. So what are the salient aspects of a company within that ecosystem? What are you looking for? Are you looking for differentiation? Are you looking for market share, expansion of the go -to -market? What are the things that you look for?

Sagi Paz

>> it all starts with the end customer experience. So AMD has customers and we want to make it easier for them to consume AMD products. So whenever you have other companies within our ecosystem that can make it easier for companies to use AMD, that's where we get excited.

Dave Vellante

>> So enterprises or OEMs or both?

Sagi Paz

>> It can be both, it can be end consumers as well. it can be whatever companies or individuals that are looking to engage with AMD. So whenever you have companies that can add value, it can be companies that are optimizing on top of our GPUs, it can be companies that are using our GPUs in order to develop their own models, it can be companies that are using our embedded products in their physical AI offering. It can be many different things, but at the end of the day once they start using AMD or being part of our ecosystem, that's when AMD can actually add some value. So the key rationale here is to help companies that can push the broader AMD vision scale faster. And we are doing that by providing them with many different things. So it can be access to compute, it can be technology collaboration, it can be access to customers in our ecosystem. But at the end of the day, it's about helping them scale faster.

John Furrier

>> So just to get the stats out there, you guys are writing checks across all phases of the rounds, early stage through late stage. Yeah, so going into - Equity, deals, is it for profit, is it strategic? How would you classify that?

Sagi Paz

>> Yeah, yeah, good.

John Furrier

>> Put you in a bucket.

Sagi Paz

>> Yeah, yeah, good question. So we view ourselves as stage agnostic, so we can start investing early stage and we can invest later stage. Check sizes range a lot, so we can write small checks, $1 million. We can write very, very big checks, tens of millions and beyond. The key rationale for the check size is the level of engagement that we have with the company, how much value we can provide, how important it is to AMD, and also how much money they are raising. So when a company is raising billions of dollars, then potentially we can invest more versus companies that are raising less. But I would say that, again, the investment size is two orders of magnitude that we can play in, and the stage is stage agnostic. It all starts with the strategic vision.

Dave Vellante

>> Do you try to avoid taking board seats, or do you insist on taking board seats? More the former, right?

Sagi Paz

>> We want to have board observer seats. It's harder to have director seats because there are some fiduciary conflicts between AMD and the company. But we do want to have close dialogue with companies, and sometimes having board observer seats is a way to maintain very good engagement with the company, so we optimize for trying to take it. We can't take it in every single situation, but probably half of our portfolio, we have board observer seats.

Dave Vellante

>> And you do that often times personally?

Sagi Paz

>> Yeah,

Dave Vellante

>> yeah. What's the max that you can do? It's like direct reports. How many can you actually have?

Sagi Paz

>> The jury's still out on that. I think that it's easier when you have observer seats versus director seats. But probably around 20, I think is the max. And some companies require more time than others. some companies, especially the ones that are doing very, very well, they don't need a lot of advice from me, the ones that are a bit more struggling usually require more time.

John Furrier

>> Yeah, I think it's important, I think people need to understand that with the ecosystem success that AMD's had, having a venture arm really allows you to, one, fund the potential next use case. One of them is agentic, and agentic will lead into physical AI, which is a lot of what we had multiple segments on earlier. One company I interviewed in Paris two weeks ago at the RAISE Summit was Fireworks AI with Lin, the founder. And I like this company because she validated our thesis when OpenAI launched. We predicted a power law of models. And we were contrarian at that time to say there'll be specialized models. One of the things that they hit on, they have all the top AI native companies doing inference. They nailed the specialized models that work well with the bigger model, so model fusion if you will. So I bring this up because you never know what's going to come out of this because one of the hot topics today is you can have compute routing, model routing, so there's a whole level of intelligence around that. That's just an example of what could come out of an investment. So you're looking at a thousand flowers blooming. So as a venture, how do you look at the patch? Do you have a criteria? Are there things that are maybe off roadmap that you might take a chance on or is there an intentional strategy to throw a lot of seed on the soil and let it grow?

Sagi Paz

>> I think that it depends on the stage of the company. you mentioned Fireworks and we invested in Fireworks a few years ago and the bet was really on the team. When you have a very, very strong team in a market that is moving extremely fast and you know that things are going to change within a year or two, you want to bet on the best team that can make sure that they tackle the right opportunities and can create value for their customers. That's exactly what Fireworks has been doing. So especially when you're looking at earlier stage investments, a lot of it is just having conviction on the team, making sure that they have very, very relevant experience and can tackle the right opportunities.

John Furrier

>> When should someone call you or ping you on DM or whatever? What's the criteria? What's the bar? How tall do they have to be to ride the roller coaster?

Sagi Paz

>> I would say that the first question is whether AMD is uniquely positioned to add value. I think that if the answer is no, and there is nothing unique that we can provide, then usually it's going to be a very shallow conversation. We are not a financial investor. We want to make sure that we can harness the value of AMD and make the right company scale. So that's probably the first criteria. The second criteria is that the company that is looking to raise money, right? We are investing alongside other great investors. We are not looking to necessarily lead rounds. We much prefer to participate alongside very strong investors. So the company definitely needs to be in fundraising mode when they talk to us.

Dave Vellante

>> When you think about your portfolio, do you even worry about mix hardware versus software? What type of software? And then I have a follow up on that.

Sagi Paz

>> Yeah, we definitely care about that. We want to be more focused on areas that we believe are really going to grow. So right now, most of the focus is actually on the software side. And that's not because we feel that hardware is not going to grow. It's because hardware is AMD's core business. It's actually sometimes harder for us to invest in other hardware companies and add value to them because they can be viewed as more competitive with what AMD is doing with its core offering. Sometimes it's easier to invest in the software, which is not competitive with AMD, and we can just have a very synergetic relationship with them. So probably 80 % of what we do is on the software side. And within the software space, we have how much we want to invest in AI model companies, how much we want to invest in specific verticals, and we have very high level understanding of the portfolio construction, but at the same time, a lot of it is driven by the opportunities that you see. So if you're going to see a great company raising a great round, and there is a lot of value that you can bring to the table, you're not going to pass on it, because it's going to maybe over index on that particular area. So you have overall guardrails, but at the same time you're very driven by the deal flow that you're seeing.

Dave Vellante

>> My follow up is how far up the stack you go. you've got some software experience, you're at Microsoft, so you understand software. The entire software stack is completely transforming. you've got a $50 billion BI business that is under attack, you've got LLMs attacking traditional SaaS, you've got this whole new end user experience going on, you've got this, what we call the system of intelligence, or Alex Karp's ontology, which everybody wants a piece of that action. How do you think about, how important are the changes in that overall software stack, and how far up will you go?

Sagi Paz

>> the changes are extremely important to know which companies are going to scale. We would go as far up as where AMD can add value. So if you're just talking about application companies that are just using AI model companies' inference platforms, but don't really touch the hardware or leverage the hardware directly, then it's really hard for a company like AMD to add value. But when you're talking about application companies that are a bit more vertically integrated and maybe do run their own GPUs and their own workloads, that's where AMD can actually help with optimization. So the key question is whether there is something unique that AMD can bring to the table. If the answer is yes, we can potentially make an investment. If it's a great company, then we're also going to be very, very excited.

John Furrier

>> Sagi, great to have you on theCUBE and sharing your approach and the work you've done. Got a lot of companies. Putting my startup founder hat on, I have to ask the question for other founders out there and teams, where's the white space? Because you guys see a lot too at AMD. you see your roadmap. Part of the service that you obviously bring to the table is you might recognize some white spaces. And startups love white space. They can get a position, maybe an opportunity. What white spaces are out there that you think people should focus on? Is it robotics, are there blockers in the enterprise, on agentic, what are some nice areas that are open for business right now?

Sagi Paz

>> Definitely physical AI is something that is still out there, right? you don't have any company that is actually deploying at scale, but that is going to be the next inflection point. So I do believe that there's going to be a ChatGPT moment for physical AI and that everyone's going to recognize the value that you can drive from that type of innovation. So definitely bringing AI to the physical world is something that there is a lot of opportunity there.

Dave Vellante

>> China's got a big lead there too. Especially on the hardware side. What about energy? Are you investing in energy?

Sagi Paz

>> we're not making direct investments there but AMD as a company, obviously we care about access to power, that's a very big limitation right now for deployments but we haven't made any direct investments, but again, we are expanding our investment horizons. So, indirectly.

Dave Vellante

>> A lot of vertical integration on these neoclouds. Well, I was just going to say, neoclouds, that's an indirect way of playing the energy.

Sagi Paz

>> So we definitely invest in neoclouds, and we invested in TensorWave and Vultr, so yeah.

John Furrier

>> Sagi, we'll keep in touch. If theCUBE needs some funding, we will tap some free credits. Actually, the GPU cloud's appealing to us. So thank you for coming on. Congratulations on a great event. Love the portfolio and we'll be watching your deals and we'll keep in touch.

Sagi Paz

>> Thank you.

John Furrier

>> Alright, AMD out there seeding the marketplace because the next wave of innovation is just one entrepreneur away. There's a ton of white space. Agentic is going full AI infrastructure. That's a direct lead into physical AI which is a lot of advancements in robotics and hardware specifically, form factor, embedded systems, a lot of action. Of course, AMD's got the market covered with capital. We're doing our part here at theCUBE. I'm John Furrier with Dave Vellante. We'll be right back.

Sagi Paz, AMD

search

(INTRO)

John Furrier

>> Welcome back to theCUBE's live coverage here in San Francisco of AMD Advancing AI. This is going to be the show that sets the agenda as the AI infrastructure continues to build out, of course, it's bounded by energy and capital. Our next guest runs AMD Ventures, Sagi Paz is here. Thanks for coming on. you're writing a lot of checks enabling ecosystem partners, startups, big companies, to align with the major shift.

Sagi Paz

>> Yeah.

John Furrier

>> Whole other generation of system architecture coming online with AMD and AI infrastructure, large scale deployments of clouds, neo -clouds, data center modernization, and a massive growth in open source, which is developers, physical AI, the portfolio laying out here. What's your investment thesis? Everything that moves in AI infrastructure? What is the focus of AMD Ventures? because this is a critical enablement for you guys. you have some successes. Take us through what your approach is and what your focus is.

Sagi Paz

>> Yeah, first of all, AI is a team sport. So there is many different layers across the stack. So it starts with the compute layer that AMD really can deliver on. But in order to provide value to customers, you actually need to have the AI model layer, you need to have the application layer, you need to have the software. And where we focus on is around the AMD ecosystem. So whenever we have companies that AMD is uniquely positioned to add value to, that's where we're getting excited about making investments. But it's not just around the strategic alignment. We definitely want to focus on the best companies out there. So whenever we do make an investment, we start with the fundamentals. So each investment needs to stand on its own. It needs to be a very credible company, great team, differentiated product, big market. But then having the angle in which AMD can provide some unique value to the company is where it gets us really, really excited, and that's where we make investments.

Dave Vellante

>> So staying with the investment thesis for a moment. We John and I did a lot of research this week with our team, and we laid out this three -layer architecture. Organic R &D, innovation, critical, where you have gaps. Gordon does his M &A thing, obviously. you guys have been very acquisitive. And then the third is the ecosystem, what you just mentioned. So what are the salient aspects of a company within that ecosystem? What are you looking for? Are you looking for differentiation? Are you looking for market share, expansion of the go -to -market? What are the things that you look for?

Sagi Paz

>> it all starts with the end customer experience. So AMD has customers and we want to make it easier for them to consume AMD products. So whenever you have other companies within our ecosystem that can make it easier for companies to use AMD, that's where we get excited.

Dave Vellante

>> So enterprises or OEMs or both?

Sagi Paz

>> It can be both, it can be end consumers as well. it can be whatever companies or individuals that are looking to engage with AMD. So whenever you have companies that can add value, it can be companies that are optimizing on top of our GPUs, it can be companies that are using our GPUs in order to develop their own models, it can be companies that are using our embedded products in their physical AI offering. It can be many different things, but at the end of the day once they start using AMD or being part of our ecosystem, that's when AMD can actually add some value. So the key rationale here is to help companies that can push the broader AMD vision scale faster. And we are doing that by providing them with many different things. So it can be access to compute, it can be technology collaboration, it can be access to customers in our ecosystem. But at the end of the day, it's about helping them scale faster.

John Furrier

>> So just to get the stats out there, you guys are writing checks across all phases of the rounds, early stage through late stage. Yeah, so going into - Equity, deals, is it for profit, is it strategic? How would you classify that?

Sagi Paz

>> Yeah, yeah, good.

John Furrier

>> Put you in a bucket.

Sagi Paz

>> Yeah, yeah, good question. So we view ourselves as stage agnostic, so we can start investing early stage and we can invest later stage. Check sizes range a lot, so we can write small checks, $1 million. We can write very, very big checks, tens of millions and beyond. The key rationale for the check size is the level of engagement that we have with the company, how much value we can provide, how important it is to AMD, and also how much money they are raising. So when a company is raising billions of dollars, then potentially we can invest more versus companies that are raising less. But I would say that, again, the investment size is two orders of magnitude that we can play in, and the stage is stage agnostic. It all starts with the strategic vision.

Dave Vellante

>> Do you try to avoid taking board seats, or do you insist on taking board seats? More the former, right?

Sagi Paz

>> We want to have board observer seats. It's harder to have director seats because there are some fiduciary conflicts between AMD and the company. But we do want to have close dialogue with companies, and sometimes having board observer seats is a way to maintain very good engagement with the company, so we optimize for trying to take it. We can't take it in every single situation, but probably half of our portfolio, we have board observer seats.

Dave Vellante

>> And you do that often times personally?

Sagi Paz

>> Yeah,

Dave Vellante

>> yeah. What's the max that you can do? It's like direct reports. How many can you actually have?

Sagi Paz

>> The jury's still out on that. I think that it's easier when you have observer seats versus director seats. But probably around 20, I think is the max. And some companies require more time than others. some companies, especially the ones that are doing very, very well, they don't need a lot of advice from me, the ones that are a bit more struggling usually require more time.

John Furrier

>> Yeah, I think it's important, I think people need to understand that with the ecosystem success that AMD's had, having a venture arm really allows you to, one, fund the potential next use case. One of them is agentic, and agentic will lead into physical AI, which is a lot of what we had multiple segments on earlier. One company I interviewed in Paris two weeks ago at the RAISE Summit was Fireworks AI with Lin, the founder. And I like this company because she validated our thesis when OpenAI launched. We predicted a power law of models. And we were contrarian at that time to say there'll be specialized models. One of the things that they hit on, they have all the top AI native companies doing inference. They nailed the specialized models that work well with the bigger model, so model fusion if you will. So I bring this up because you never know what's going to come out of this because one of the hot topics today is you can have compute routing, model routing, so there's a whole level of intelligence around that. That's just an example of what could come out of an investment. So you're looking at a thousand flowers blooming. So as a venture, how do you look at the patch? Do you have a criteria? Are there things that are maybe off roadmap that you might take a chance on or is there an intentional strategy to throw a lot of seed on the soil and let it grow?

Sagi Paz

>> I think that it depends on the stage of the company. you mentioned Fireworks and we invested in Fireworks a few years ago and the bet was really on the team. When you have a very, very strong team in a market that is moving extremely fast and you know that things are going to change within a year or two, you want to bet on the best team that can make sure that they tackle the right opportunities and can create value for their customers. That's exactly what Fireworks has been doing. So especially when you're looking at earlier stage investments, a lot of it is just having conviction on the team, making sure that they have very, very relevant experience and can tackle the right opportunities.

John Furrier

>> When should someone call you or ping you on DM or whatever? What's the criteria? What's the bar? How tall do they have to be to ride the roller coaster?

Sagi Paz

>> I would say that the first question is whether AMD is uniquely positioned to add value. I think that if the answer is no, and there is nothing unique that we can provide, then usually it's going to be a very shallow conversation. We are not a financial investor. We want to make sure that we can harness the value of AMD and make the right company scale. So that's probably the first criteria. The second criteria is that the company that is looking to raise money, right? We are investing alongside other great investors. We are not looking to necessarily lead rounds. We much prefer to participate alongside very strong investors. So the company definitely needs to be in fundraising mode when they talk to us.

Dave Vellante

>> When you think about your portfolio, do you even worry about mix hardware versus software? What type of software? And then I have a follow up on that.

Sagi Paz

>> Yeah, we definitely care about that. We want to be more focused on areas that we believe are really going to grow. So right now, most of the focus is actually on the software side. And that's not because we feel that hardware is not going to grow. It's because hardware is AMD's core business. It's actually sometimes harder for us to invest in other hardware companies and add value to them because they can be viewed as more competitive with what AMD is doing with its core offering. Sometimes it's easier to invest in the software, which is not competitive with AMD, and we can just have a very synergetic relationship with them. So probably 80 % of what we do is on the software side. And within the software space, we have how much we want to invest in AI model companies, how much we want to invest in specific verticals, and we have very high level understanding of the portfolio construction, but at the same time, a lot of it is driven by the opportunities that you see. So if you're going to see a great company raising a great round, and there is a lot of value that you can bring to the table, you're not going to pass on it, because it's going to maybe over index on that particular area. So you have overall guardrails, but at the same time you're very driven by the deal flow that you're seeing.

Dave Vellante

>> My follow up is how far up the stack you go. you've got some software experience, you're at Microsoft, so you understand software. The entire software stack is completely transforming. you've got a $50 billion BI business that is under attack, you've got LLMs attacking traditional SaaS, you've got this whole new end user experience going on, you've got this, what we call the system of intelligence, or Alex Karp's ontology, which everybody wants a piece of that action. How do you think about, how important are the changes in that overall software stack, and how far up will you go?

Sagi Paz

>> the changes are extremely important to know which companies are going to scale. We would go as far up as where AMD can add value. So if you're just talking about application companies that are just using AI model companies' inference platforms, but don't really touch the hardware or leverage the hardware directly, then it's really hard for a company like AMD to add value. But when you're talking about application companies that are a bit more vertically integrated and maybe do run their own GPUs and their own workloads, that's where AMD can actually help with optimization. So the key question is whether there is something unique that AMD can bring to the table. If the answer is yes, we can potentially make an investment. If it's a great company, then we're also going to be very, very excited.

John Furrier

>> Sagi, great to have you on theCUBE and sharing your approach and the work you've done. Got a lot of companies. Putting my startup founder hat on, I have to ask the question for other founders out there and teams, where's the white space? Because you guys see a lot too at AMD. you see your roadmap. Part of the service that you obviously bring to the table is you might recognize some white spaces. And startups love white space. They can get a position, maybe an opportunity. What white spaces are out there that you think people should focus on? Is it robotics, are there blockers in the enterprise, on agentic, what are some nice areas that are open for business right now?

Sagi Paz

>> Definitely physical AI is something that is still out there, right? you don't have any company that is actually deploying at scale, but that is going to be the next inflection point. So I do believe that there's going to be a ChatGPT moment for physical AI and that everyone's going to recognize the value that you can drive from that type of innovation. So definitely bringing AI to the physical world is something that there is a lot of opportunity there.

Dave Vellante

>> China's got a big lead there too. Especially on the hardware side. What about energy? Are you investing in energy?

Sagi Paz

>> we're not making direct investments there but AMD as a company, obviously we care about access to power, that's a very big limitation right now for deployments but we haven't made any direct investments, but again, we are expanding our investment horizons. So, indirectly.

Dave Vellante

>> A lot of vertical integration on these neoclouds. Well, I was just going to say, neoclouds, that's an indirect way of playing the energy.

Sagi Paz

>> So we definitely invest in neoclouds, and we invested in TensorWave and Vultr, so yeah.

John Furrier

>> Sagi, we'll keep in touch. If theCUBE needs some funding, we will tap some free credits. Actually, the GPU cloud's appealing to us. So thank you for coming on. Congratulations on a great event. Love the portfolio and we'll be watching your deals and we'll keep in touch.

Sagi Paz

>> Thank you.

John Furrier

>> Alright, AMD out there seeding the marketplace because the next wave of innovation is just one entrepreneur away. There's a ton of white space. Agentic is going full AI infrastructure. That's a direct lead into physical AI which is a lot of advancements in robotics and hardware specifically, form factor, embedded systems, a lot of action. Of course, AMD's got the market covered with capital. We're doing our part here at theCUBE. I'm John Furrier with Dave Vellante. We'll be right back.