SiFive and the RISC-V Revolution

00:09 

John: Welcome to circuit talk funders and founders. I'm John Cole, Senior Manager on the semiconductor team at MITRE Engenuity. We are a nonprofit dedicated to solving problems for a safer world. Our semiconductor team is hard at work meeting the nation's challenges around semiconductor breakthrough technologies and the CHIPS Act. Circuit Talk: Funders and Founders is part of MITRE’s Circuit Talk podcast and video series, and it elevates the revolutionary disruptive work being done by semiconductor entrepreneurs and investors. This is an exciting time to be working with semiconductor startups. The nation is waking up to just how critical they are to our national and economic security. We're joined today by Tom Leahy from SiFive, where he's Global Head of aerospace and business development. SiFive is a startup providing high high performance RISC-V IP portfolio, and the startup was founded in 2015 by the inventors of RISC-V, the folks who originally developed that open standard at UC Berkeley. Tom is an industry veteran, and before joining Syfy, he worked at Lattice Semiconductor, Xilinx, LSI, and Genesis Associates. So he's bringing sort of a broad perspective on on the entire industry. Tom earned his BS in electrical engineering from University of Massachusetts Amherst, and his master's in electrical engineering from Cornell. Tom, it's great to have you on the show today. Thanks for coming and telling us about SiFive. 

  

01:29 

Tom: John, thanks very much. I appreciate being here. And the opportunity to speak on SiFive as we have 

  

01:34 

John: Great, well, maybe you could start just give us a little bit of background, where it's SiFive come from and what's what's the vision of the future through SiFive. So 

  

01:41 

Tom: as you stated earlier, you know, it started coming out of UC Berkeley. So DARPA, you know, part of the US government, DOD, actually funded Berkeley to come up with the open standard instruction set. So to give a little bit more, you know, overview on that, you know, it being a open ISA, very similar in nature to Linux, in the software world, which was an open standard, lot of people can partake in that. So fundamentally, what was done was, the other team created the ISA, which is the instruction set architecture. And that's what people take and can download, and create their own company with. So as a, for instance, you can start up a RISC-V, instruction set company, if you will. And with that being said, you would pull that down, and really proliferates the ability for multiple vendors to get involved in this and provide, you know, a multitude of offerings to people. And one of the reasons that the government was really interested in this is, you know, if you look at the X86 world, you know, you've got Intel and AMD, it's a proprietary ISA. So only they will create X86 based processors, the same holds true for really, what's our largest competitor is ARM, so ARM has been around for about 30 years. Now, again, they have, you know, a proprietary ISA. In fact, they've had multiple ISAs and sometimes going from one ISA to another is not the easiest thing, sometimes it's a little lift, sometimes it's a highly, you know, high lift. And with that with the, you know, starting out with a fresh sheet of paper, we get something that starts out you know, way down in the microcontroller 32 bit level, going all the way up into the high performance compute. And you know, with that, that ISA extends across all of that, to where you actually have the ability from a software perspective to go from one processor to another. So that's a little bit of the the differences there, there are extensions that are made to that, that add certain features such as vectors, etc. So when you are creating something from the RISC-V ISA, you can determine as a as a company, you know, whether you'd like to take advantage of all those instructions above and beyond, and really put things in to what your exact requirements are. We are a little bit different in the sense that we are a US based company, and we have a complete portfolio. And I think that's important. If you look at ARM having the low end up to the high end, we are really looking at being an IP company, you've got a lot of people that are jumping into the RISC-V arena, that will you know, go off and create AI processors, right. There's a lot of companies that are doing that. And one of the reasons for that is the RISC-V instruction set really lends itself well to providing the best PPA, so power performance area, as you're looking at a lot of different architectures. And you know, x86 is probably the easiest one. There's a lot of baggage that they have to bring forward because it's been around so long, and it's very robust, but they've got to bring that forward for all the new iterations. And us starting out with that fresh sheet of paper really allowed people to create this new ISA and they also created a RISC-V, where at that time, which is a consortium that has now turned at RISC-V International to where there's over 3200 members across 70 countries. So you have people giving back to that. The open standard, which reviews things, such as in our case, we just released our world guide, security solution, back to RISC-V International, so that has its own CEO and structure etc. And when you're pushing things back into that, it has to get approved gets, you know, reviewed, etc, and goes through that whole process. So that's one of the things that's a benefit for people, you know, we providing back into the RISC-V International. 

  

05:42 

John: So if I understand correctly. RISC-V makes the ISA decides on that standard. And then companies like SiFive will come in and sort of develop an IP layer on top of that, is that right? 

  

05:52 

Tom: Yeah, they essentially create a, an IP. So in our case, we have a, what we call an RTL deliverable. So it's a Verilog model that is delivered to the the end customer. And with that, everything, if you look at it, it's kind of our secret sauce is what you would consider the micro architecture. So it's the implementation associated with all of those instructions, all of the features, etc, to where we are paying close attention to that, you know, we've got a good jumpstart, because, again, the founders here invented, you know, RISC-V. So we've been able to take that and bring that forward and bring forward again, PPA, the power performance area, that really not only provides an alternative to, you know, the competition, but something that adds value above and beyond, we've got a lot of things that are going on now, for small form factor applications, low power applications, you know, that have really come to bear in the industry itself. 

  

06:54 

JOHN: So why are companies coming to SiFive, like if I'm, if I am deciding between, say, RISC-V and arm and then I'm finally narrow in on psi five, what are you offering that sort of brings a company into where you are. 

  

07:09 

TOM: So one of the things is, if you look at it is really is A) the robust portfolio, because you can start out, there's a lot of people doing, you know, different architectures, where they'll use multiple processes on the same design, and that that allows you that benefit, the same codebase, etc. If you look at where things are coming from, from our customer base, they are looking to have something that has a strong ecosystem, has something that is a longevity, you know, if you look at it, there was a couple of major announcements that have been made recently, like NASA JPL has made an announcement that the next generation, high performance space compute is going to be based on our solutions. And we're partnered with microchip on that where they are providing the actual device, we provide the the architecture and the RTL. And you know, they take that to become a product add, you know, additional things on top of that to meet the requirements of NASA. So that's a real big opportunity. And it's there. And that was something that was based on longevity. You've got other folks that are...Google had made an announcement, that is going to be using some of our stuff as well, for the next generation TensorFlow processing unit. So if you look at it, you know, Google is very well known in the AI world. And with that, they've got their accelerators. And that's really their, their crown jewels, if you will. So they're using us to be able to have something that's tightly coupled and has an interface to be able to talk directly to that AI, you know, hardware accelerator, and provide the best solution. So there's a great example of, you know, a very large company that could go off and create something on their own. But they figured, you know, we are leading the pack, we've got something that's very, you know, from our micro architecture, provides a lot of value has the features that are there. So we're really trying to take advantage of that, you know, on on all facets, you know, throughout the different segments. 

  

09:04 

JOHN: So, so Google is using your technology, NASA is using your technology, SiFive, in the DARPA toolbox, I saw defense, that sort of opens up defense, which is another big customer. We think about these larger customers, typically, they're slower customers that are sort of harder for startups to interact with. Why are you thinking about defense, maybe defense in aerospace, and you're you're doing a lot of business development with defense and aerospace in particular, why is that important to SiFive? 

  

09:31 

TOM: So I think there's there's a couple of reasons there. You know, first one thing that I will mention too is presently a we we have SiFive being used in and engage with eight of the top 10 semiconductor companies. So that really just shows how pervasive RISC-V has been and you know what you're seeing now, that you know, is a great state we're very proud of that. But if you if you take that back a step for you look at you know into the aerospace and defense sector, you know, I was brought on specifically to help out in that arena because of the background that I do have. With that being said, it's providing the opportunity to have that open standard. That's something that the government has been pushing on for quite some time, they're really jumping into supporting Linux as well. And the fact that we are having advanced discussions and dialogues with many of the folks within the Department of Commerce, the Department of Defense, the Department of Energy, because you've got a lot of different things that fall through that, then you've got the other layer of the agencies, the AFRL’s of the world the Navy cranes, the Army research, labs, etc. And then you've got the whole complete DIP. So the defense industrial base, there is just a large hole right now, relative to this type of architecture, especially given the fact that we're providing PPA advantages. That's one of the biggest things that are there. 

  

10:54 

JOHN: Can you unpack the so I started NASA is using the SiFive technology at JPL for high performance space compute, I think, right? It's pretty gold plated customer. And that's a very, that's a very demanding mission, right there. Sort of rigorous testing before you put something up in a satellite and sort of let it go. First, I guess what, why does space need a high-performance computer? They're not mining Bitcoin, right? What are they? 

  

11:20 

Yeah, so I mean, if you look at it, the the amount of the missions in space now are really having more and more demands relative to compute, there's a ton of imaging that can be you know, do print processing on, etc. And, you know, that's one of the biggest things that's up there is, I was just out at the space center last week, and they were talking about the straw, that is feeding everything back from all the satellites that are there, there is more compute and information that is out there, that needs to be brought back. So a lot of this, the capabilities to do some of the pre-filtering on what's being brought down. So you can have, you know, live data, that's, that's streaming down, then there's other applications that they need all that data. So it just takes a significant amount of time to take a snapshot and bring all that down. 

  

12:09 

JOHN: When you're talking about this draw, you're actually talking about the idea that you're taking in a lot of data on the satellite, it's an edge computing problem. And this process is much there, because you have a very limited sort of down up link, 

  

12:21 

TOM: That is 100% true. Yeah. Yeah. So you know, with that we've been, you know, again, trying to match up with that, then the requirements that are moving forward, you know, what's being brought forward today, you know, it's something that's going to be able to provide 100x times the performance of what was used before. So, again, your point of getting into the space domain is something that is held very near and dear to the space community, you've got a lot of hurdles to overcome. But you know, they're using a product that is based on, you know, a PowerPC, that's 20 years old. So they're really looking to up the game there, you want to start being able to support some of, you know, the activities for a lot of vector processing a lot of AI etc., and really offloading as you stated at the edge, so that that downlink stream can provide as much information that's been pre filtered as possible. 

  

13:16 

JOHN: Certainly, the challenges that have the opportunity to like sort of be up there are big, but getting everything into space just got to be the most exacting standards for reliability and testing, sort of any software, let alone any hardware that you put up, there's got to work for 10, 20 years, right. So that's a pretty hard market to penetrate. Why is and that seems a little bit, say a different direction that I've seen other startups kind of come in, right, you look for an easier even almost almost like a consumer based market is an easier one to initially penetrate. And eventually, after years of that you get to the space space market or the harder to penetrate markets, right. 

  

13:53 

TOM: I think, you know, to answer that one, you know, if you look at it, we've had some some great penetration, you know throw out as you mentioned, on the commercial sector, a direct transactional engagement is a little bit easier for, you know, new startups to come to, but, you know, we are definitely growing, we grew, you know, 85% last year in our engineering alone, we are able to support more and more and with that, there's been a large concentration on automotive. So that was a sector that we spent a lot of time on for reliability, quality, ISO 26262 etc. I was brought on specifically because of what was going on with NASA, you know, that was a year and a half engagement and just evaluation, and now to really promote that and work with the government. And, you know, the the mantra that I bring forward to every one of my meetings is the fact we are looking at becoming the de facto architecture, you know, within the DOD and DIB and we want to bring a differentiated product forward. So that's something that's very important to us and I think there's you know, there's definitely a bit of patriotism there as well. Patrick Little, who is our CEO, and our complete executive team is behind that. And one of the reasons they brought me on. So we've been spending a lot of time really trying to get in front of a lot of the activities, you know, not look at just what we have in our portfolio, our next generation, but start having some of those discussions as to, you know, what can we do to differentiate what's going on in that space, and in a lot of cases in the DOD and Department of Commerce, wants to take something like that, and also have the viability of going into the commercial sector. So we weigh all that stuff quite well. And a lot of the things that we've done, or the HPSC, where it's a 280 based solution. But that's where we brought forward a vector processor in conjunction with a scalar processor. So it's really the best of both worlds. And it's something that no one else is doing right now. And it's providing a huge differentiation. That portfolio, which we're calling our intelligence series, is growing faster than anything that we've had to date. And a lot of that has to do with because we're listening to the customer base to help drive that it plays not only, you know, with supporting the DOD, but also has great commercial aspects to it as well, as you see, you know, Google making a statement that they're using us as a very big one. And that's just continuing to proliferate. 

  

16:26 

JOHN: Well, that does you know, that in the dual use market, I guess, folks talk about where you're sort of partly serving the government, oftentimes defense, right, but also serving the commercial sector. That playbook can work out really well to I'm curious, like, if you can, it must be easier to go to the automotive manufacturers and say, you can trust this in your airbag, because we've sent it to space, and it sort of has met those exacting requirements. And so there's probably some proof there for for RISC-V and SiFive, 

  

16:54 

TOM: Yeah, there's been a lot of, you know, investments from the from the government sector to have programs that have been proof of concept, if you will, from a from a space perspective, you know, we're working closely with all of the the onshore fabs as well, just relative to, again, some of the CHIPS act Microelectronics Commons, and the onshoring, you know, capabilities. But, you know, to that point in the commercial sector, we are also heavily invested with TSMC, many of our customers are using them, and being supported, you know, within Taiwan. So, you're right on the money relative to, we could take a, an easier path and just stay 100% commercial, if you will. But I think there's some real good benefits, if you look at the technology, and a lot of people don't necessarily appreciate that I've been doing it for a while now. They really do, they are on the front edge of technology, what's needed and really drive things to where you put people in a position saying, “Okay, well, we need something even better than that.” And, you know, it just drives us well, in addition, because we don't go, so we've got some competition, but we are really driving towards making sure that we are providing the most robust portfolio and feature set to stay ahead of the curve. Because quite frankly, and that's the one thing that's beautiful about the open standard. You know, if we start to slack off, somebody else will be there, and you have the ability to go off and, you know, use the same instruction set and port to another architecture. So there's a beauty there. But from our perspective, you know, we've got to stay on our game and continue to execute and make sure that we're listening to our customer base. 

  

18:32 

JOHN: Yeah, that's well put. Everybody wants to see that in a business. That's great. You know, just kind of back to,  you, talked a little bit or you you mentioned briefly, the DOD microelectronics Commons program, talked a little bit about some of the new government programs that are emerging. Switching gears, CHIPS Act is being implemented right now. And DOD micro electronics comment program is like the first sort of out of the gates program in terms of timeline, is SiFive planning on participating or anticipating, you know, participation anticipate. 

  

19:03 

TOM: So we've been, we've definitely been fully engaged there have a lot of, as mentioned, you know, working with the, the DOD, again, the Department of Commerce. Department of Commerce, you know, for this type of spend, it's something that's relatively new for them, right, if you look at the size and scope of the program, so they're bringing on a great, great amount of folks relative to the chips.gov team, the DOD, microelectronics commons, you know, that's more of the standard fare that they're used to. That's a $2 billion spend. But, you know, to both both sides of that, we've been very active with regards to, you know, folks that are putting forward proposals for the hubs. They're indicating that that should be chosen by the end of August, and they're looking at probably between six and nine from the latest statements being made from the teams. And then you've got the whole CHIPS Act, which is even larger. Where you have the NSTC which is the National Semiconductor Technology Center. So there's there's definitely a lot of activity there. And there's a lot of activities relative to speaking to the government, but also speaking to many of the states. And then you get into for the ME commons, that has a lot of academia that is involved, and they're bringing things forward and really trying to couple and meet the six different objectives out of the ME Commons. So we've, we've partnered, and I'll say, you know, spread ourselves and having conversations with ten of the different proposals that are being brought forward. And we're making sure that there's some of these activities, the fact again, that we're a US based entity. You know, we're just saying, you know, really take advantage of that, because we we are looking to, again, bring out that differentiated solution, that we can work with some of the other players, whether they be the fabs whether they be some of the other architect teams, and you know, in the DIB, and we've been talking at that level, to say, Okay, what is the application that you're looking to go off and solve, because that's one of the things that are being done, and really just drive that across all the different vectors that are there, it's, it is a tough thing to follow. There's a lot of activities between the state side things the Department of Commerce, and then you get the Department of Defense. But thankfully, I actually had the opportunity probably about a month ago, to present to the DMCFT, which is the defense microelectronics cross functional team. So they have members of the Department of Defense Department of Commerce, as well as Department of Energy. And what they're trying to do is, you know, not recreate the wheel, that's something that's been known, and this whole siloed approach. So they're trying to break that down a little bit. And we're just saying, you know, again, being US based and a RISC-V, let's get in front of us this, have some of these conversations, get get us involved with the architects, whether they be from the DIB, or there's some brilliant people within AFRL, etc. so that we understand what those requirements are. And, you know, potentially, if it if it aligns, you know, add some features there. So that's something that's been very well received overall. And we're trying to to assist as much as possible. 

  

22:21 

JOHN: Yeah, that's often one of those overlooked advantages of just the CHIPS Act, it costs almost no money, right, but just the, you mentioned, being able to talk to your customers or talk to other players in the ecosystem and work together through the convening function is something that happens at almost no cost of just saying let’s all get together and like talk about these problems. And that cross kind of cross pollination is some I'd say, generally nets something pretty nice. But zooming out, you know, looking at the whole ecosystem for a second taking off your SiFive that like, just you've worked at Xilinx and Lattic ebefore this. So semiconductors are now getting some much needed attention, right. It's kind of dark there for a while, getting some funding from the CHIPS Act. And meanwhile, fabless is thriving in the US, I think we've got a number of startups, especially in specialty chips that are focused on that they're getting funded. But how do you see SiFive interacting with the CHIPS Act? And when you sort of look beyond SiFive, right, what is the CHIPS Act or need to do to seed more semiconductor startups in the US? 

  

23:25 

TOM: I think, you know, one of the things that they are trying to do, and they've got it, they've had a couple of different, you know, plays, even with the ME commons, where we have been working with some partners, relative to having a common infrastructure. So you want to have something and we've kind of brought this forward for some of the performers, you know, we're willing to work with them on the front end of things, get engaged with academia, get engaged with some of the startups. And that's fundamentally what they're trying to go off and do. And with that, we've been talking with individual hubs, as well as you can have a kind of a generalized infrastructure that has also been put forward, that the hubs instead of creating their own, they could have this and it includes essentially things in the cloud, so and all the EDA manufacturers, for us, we kind of push for that, because we're just saying, you know, a process is used on just about everything. So with that, we're trying to enable that group of folks to be able to bring things forward some new fresh ideas, and lower the bar of entry, if you will. And that's the same thing that the DARPA toolbox was doing. This is just at a grander scale. There's also an activity going on right now which is called D2TA so it is designed to technology acceleration. So this is taking it outside the scope of the CHIPS Act. They really liked the thought of doing that. How can they enable all of the, and this is something that they're doing for the defense industry, if you will, for aerospace and defense. And that's something that that trying to solve and create an environment that takes a lot of this stuff off the plate and makes it easier for some of these players to be brought forward. But then you have the capabilities of all of the defense industrial base that might have all their own server farms, etc. But now you can also have something that's done at the cloud, as a real good, good, good 

  

25:25 

JOHN: You’re kind of talking about two kind of here two separate ideas here. One is, you know, focus on cloud-based EDA, so you have more tools and more people's hands, right. And then also focus on sort of shared infrastructure. So players up and down the value, up and down the site, players of different size up and down the ecosystem of the value chain, have sort of access to plant and materials and machinery and all the other kind of expensive, very, very expensive parts of the manufacturing process is that is that kind of what you're describing? 

  

25:57 

TOM: That’s correct. As an example the, there was a program called Ramp, which you may have heard of. And Ramp fundamentally had three different major programs. One was with Raytheon, and two with BAE. And what was created, there, was an infrastructure that was done out in the cloud had all the EDA vendors coming in, hadsome of the IP, not not all of it, because it was just too much to do in a short period of time. But they were able to, you know, get three programs done and out within 10 months. And one of the biggest things that I saw there is Raytheon, and VAE, you know, created a tight hole in the firewall to be able to work with this, and had all the structures and security metrics in hand. But were able to do that. And then, you know, you look at all the EDA players getting in the same sandbox. And really working things forward. Once the legal portion of things were done, people are stepping up and really trying to, you know, look at the look at the geopolitical situation overall. And, you know, just try and help out. And, you know, that's where that's where things are going. And, you know, it's exciting to be part of it, because there's a lot of activity. 

  

27:07 

JOHN: Yeah, yeah, no, it is an exciting time. And hopefully, somewhere between the convening function and the actual cash dollars, results in some real better systems for both startups and for larger players to make new innovative designs and new innovative technology. Coming back. And one more question about RISC-V, if you don't mind. You know, when I think about it, in my mind is their, and it ties back into the CHIPS Act. But there seems like there's an equivalence to the open source software movement. I know, there's the open ISO standard, right? There's the movement sort of occurred 20 years ago, where you had this emergence of open source software. And I'm just kind of thinking, in my mind that maybe RISC-V is sort of like the Linux and SiFive is the Red Hat. Is that a fair analogy? 

  

27:50 

TOM: That's a that's a perfect example. Yeah. Because that's, you know, there's, there's the ability to make and the features that are being brought forward. And, again, how robust and functional things are from a Linux Red Hat, people are very happy with that it's, it's gone through the rigor of being able to be brought forward. And that's one of the biggest areas even that you can go off and get things and we've actually pushed some stuff up onto open source as well. There's a rocket, it's called Rocket, it's a core that was a 32 bit machine, if you will, that's been used on multiple programs within academia and the government. But if you look at some of the higher end things, that's really where, again, that implementation from a micro architecture perspective really comes into play. And that's our true value add. So it has the ability to offer both, you know, where you can go down the full open system, or you can go off and get something that's gone through a lot more rigor that's kind of proven in the industry, and brought forward and you know, that's that's kind of where we're at right now that there was a perfect comparison. 

  

28:59 

JOHN: I think you kind of alluded to it, but just understands of you, there's a, an open...the open architecture, like SiFive will make suggestions or pushes back to that, that are open to the rest of the community. Right. So each sort of each player in the community will make, to some degree, open source contributions, so to speak back to the to the standard is that...? 

  

29:22 

TOM: Yeah, and that’s something that again, gets reviewed, goes through a whole through a whole process, which is pretty rigid. It's not a simple laissez faire, you know, wishy washy. It's got a large review process before anything gets turned on and brought forward so there's different revisions if you wanted to kind of go down that path. Again, everyone can bring things forward. Some are some are better ideas than others, but that's got the capability of that, that whole open standard, but you know, it does have and is addressed by RVI which is the RISC-V International As far as the process and what's what's properly committed to, from any changes moving forward, and then if you want to be compliant to that, you know, you can add that to your, your next generation product or possibly even retrofitted into an older one. 

  

30:15 

JOHN: Yeah, that's a great, that's a great system, right? In a sense, a lot of contributions kind of spread among the whole ecosystem, but there are some strong, strong winners, and everybody kind of benefits from from a strong ecosystem of, of interplay there. 

  

30:31 

TOM: It provides an equal footing, for the most part, to be able to have that where everyone can bring forward their ideas. And there's, there's a lot of very smart people out there, you know, a one man band might be able to, you know, offer something up that people do a review on and say boy, that's actually a great idea. And somehow that gets, gets pulled forward, or even maybe augmented and, and brought, you know, brought forward 

  

30:53 

JOHN: Is there, and I guess one of the vendors is SiFive is the equivalent of Linux, Torvalds, the founder of Linux, they are kind of working at SiFive right now. Is that right? 

  

31:03 

TOM: Yeah, I mean, that's, that's one of the, you know, Kirsta. And that's where our, you know, our original founders are all are all on board. And, you know, with that being said, that's, that's really been great in and of itself, because we have that, that mind meld between those individuals, and we've just continued to expand that. And our executive team has really, you know, been a situation of been there, done that, and having the growth that we're seeing, you need to have that in place, in order to stay the course and make sure that you continue to execute, you know, don't get, don't continuously follow the shiny penny. Because that's, I've been down a few different roadmaps with, with some startups that I've worked at. And I think we've been doing a good job trying to make sure that we're staying focused on, you know, what we committed to and who we've committed to with, and, you know, just continue to grow our portfolio. 

  

31:59 

JOHN: Well, Tom, so that's fantastic. Putting on your wide angle glasses one more time, and you know, all your experience at Xilinx and Lattice and sort of looking across the whole semiconductor industry, who else's work right now, in the ecosystem, especially startups, are you most excited about seeing how it lands unrelated to what SiFive is doing? 

  

32:16 

TOM: I think a lot of the things, you know, if I look at RISC-V based applications, etc, you know, we've been getting, you know, great support relative to, you know, the EDA vendors, there's, there's a ton of things going on right now, with regards to say, the CHIPS Act ecosystem, the chiplet ecosystem, so you could take our RTL, you could go off and create a chiplet that you can go off and create, you know, customized products bring that forward. But the EDA vendors, you know, they've got a lot on their plate as far as to what they need to deliver, you know, heterogeneous compute is not an easy solution, the packaging, all the rest of that. So, we've been working very closely with the major players in that space. And you know continue to push and want to work again, you know, with the government on the CHIPS Act side of things, but working outside those bounds as well for a lot of the automotive product that we're working on. And a lot of it tends to once once the traction is there, everybody's there to try and help out and just continue to push their deliverables as well, but EDA vendors have, that's what's real interesting to me, as far as this overall solution and the CHIPS Act. 

  

33:26 

JOHN: And other than sort of the major players. Are there any other any sort of emerging folks, and that's just in the software space around EDA that are doing some things, something interesting, or something new that you can't wait to see what happens. 

  

33:39 

TOM: I think, you know, a lot of it has to do in my eyes, is if you look at and this is my own personal opinion, you know, a lot of the packaging, right, you've got all these activities surrounding, you know, the fabs, etc. A lot of stuff coming on Sure, you've got the likes of, you know, Global Foundries extended, IFS, Samsung, TSMC, SkyWater for RAD hide stuff, Honeywell, etc. And I'm sure I'm missing a couple people there. But, you know, that's, that's pretty well understood of what they're looking to try and do it help the supply chain. But now if you look at say, what they are trying to do, everybody's going down the chiplet path, the activities for the packaging is something where there's an even larger gap, and some of the activities relative to some of these high-end applications, that's where we can add some value. There's a lot of investment that's being made there as well. And I think that's going to, you know, just as far as from an industry perspective, kind of step things up to the next level, we're always going to rely on all of the major things that are overseas for you know, a lot of the packaging, etc. Our value add, you know, isn't isn't quite there for that. Whereas in some of the higher end stuff, you know, some of this high performance compute a lot of the stuff going on within this in space arena, and really some of those tools that allow that to, to be a little bit more of an ease of implementation, that's where, you know, these large customers have the armies of folks to be able to throw at it. But how can you take that and get the next wave of folks, you know, the startups, academia, to add that value and start being able to, you know, take advantage of the chiplet ecosystem, there's a lot of stuff that they want to do there, you know, overall, within the chips act, and it's very interesting to see, but I think there's, there's a lot of activity there don't necessarily want to name any companies. But you know, there's, there's real good things going on there. And I'm very interested to see how it all pans out. 

  

35:44 

JOHN: We've talked to more than a few startups about chiplets, and everybody is almost ready to talk to is bullish about it. Right. So some of the advantages are advantages on supply chain, or it enables, even from a startups perspective enables them to simplify their design or to integrate with other folks design, right? Is that the I mean, what's the big, big deal? I guess, from your perspective of chiplets? What does that do? What does that for America? 

  

36:11 

TOM: One of the biggest things is you can start integrating in different process technologies, right? So you can, you can start engraving analog in, and you can start saying, Okay, I've got a seven nanometer fab chip in with a 60 nanometer, or maybe even a three nanometer, however, that all all works out. And one of the big benefits is, you know, just the interconnect philosophy and saving of power. We are getting driven by so many different applications, which is great, because the technology is going gangbusters. But a lot of things operating on batteries now. So that low power, and again, you know, some of the small, small sizes of different applications are really kind of coming into play. So that's going to going to lend itself well, on the high high end, because you've got folks like Nvidia that are, you know, doing this to really bring forward the most compute within a single chip as possible. But then you've got it on the other end as well, to where if I can cut my communication, instead of going off chip, for some of the DDR, I can use something some of the technologies that are available today, it's all a power game for on the lower end. So there's a wide variety of applications that can be be utilized there, I think we're really just hitting the tip of the iceberg. You know, years ago, if you look at it from, I go back too far in the DOD side of things where if you ever mentioned the word MCM, which is a multi chip module, so it was chiplets, before that time, right, where you needed to have known good die and all the rest of that, that's something that has come full circle now. And it's almost a requirement to be able to merge some of these devices, instead of doing monolithically. So now you can, you can take a chiplet that's a processor, start adding A to Ds D to A's on it and get the best of that technology, you start integrating all of that into into one device. And you're really, you know, kind of shoehorn yourself into saying, okay, I can only hit these applications. But if I have something that's got the interface that's already set, and you go off, and you look at an ADI or a TI, or whomever it might be and say, Okay, our interfaces match. Now, I need to put together a chiplet that does this. That's a perfect thing. And then one of the other things that is that is going on as well, is they want to have a trusted repository for some of these chiplets. Because, you know 

  

38:38 

JOHN: Just IC in general? 

  

38:42 

TOM: Exactly. If you look at the aerospace and defense market, it's only 1% of the overall TAM of microelectronics. So to your point early on, it's not always the best thing to go after. But there's a lot of benefits from that. So that the learnings that you can take, bring those into the commercial sector, it's so it's really a win win for everyone. 

  

39:06 

JOHN: Yeah, that's fantastic. So what's holding up chiplets? In your, I don't want you to name names or get in trouble here. But what's what's the what's the big hurdle? And how does the CHIPS Act sort of get us over that hurdle? In your mind? 

  

39:18 

TOM: Well I think you know, with that there's a lot of investment going on, in the EDA tool sets. That's probably one of the biggest areas, you know, everybody does have, you know, all the major players are actively involved in that and you know, some of the service companies have the folks on that can go off and, and support some of the system level, you're really solving something a little bit different. You're coming from the system level on down versus a chip on up. And with that, having those in place having the the knowledge at these companies to be able to support that. And the easier the EDA vendors make it to be able to incorporate some of this versus having to have the expertise in house yourself. So that's where I think like that ecosystem is really just getting the ball rolling right now. But you've got the major players that are going to provide that enablement. And then you've got all these different incarnations of, I'll say, different packaging capabilities, and how you can go off and solve that problem. You know, there's not just one way, there's multiple. So I think that's, that's really where we're at right now. You know, again, some of the major players have the armies of folks to be able to go off, and they've been doing that, playing with it doing a lot of test chips, etc. But now it's the matter of, okay, how can I go off and enable some of these mid-tier players that don't have all that much experience to go off and do that. And I think that's really where the driving towards, and the chips act is trying to promote that whole ecosystem. And that goes from the design IP all the way through to the packaging, testing, etc. 

  

40:56 

JOHN: Yeah, certainly, that makes sense. It's sort of kickstarting it will be you really didn't say this, but it's kind of implied that you got to have both a customer and a supplier at the same time coming together to make like an ecosystem work, right. Or in this case, many also have to agree and sort of just, you know, at some point, the initial activation energy 

  

41:16 

TOM: If you look at it, you know, there's what's called a ship program, which that's the heterogeneous compute, you know, Darren Crum was kind of leading that effort. But you know, you've got Intel, you know, working along with that, and they've provided those devices, and now they're looking to take those and have applications driven onto them. So you know, it's the proof point. But it's also really looking at that process, how can we streamline that in a better fashion? And I think there's a lot of people stepping up to put that together. And one of the things again, you know, the chips.gov team, I think, is fantastic in the sense that they're bringing on a lot of seasoned veterans that know, you know, that know, the semiconductor space, and I think that's very important in order to be a successful migration forward. And, you know, hopefully, that that's going to we'll have a CHIPS 2, because I think, you know, if you look at the geopolitical situation, we're a little late to the game of having this, you're starting to see, I was just reading earlier this morning, you know, Germany's putting up $20 billion from an fab perspective. And, you know, it's, I think it's, it's good to see this, it gives a long awaited effort in the microelectronics, where there's a huge buzz, I think there's great things to come out of it, you know, moving forward and, you know, SiFive, we want to, we want to definitely be part of it. And really, just bring our wares forward and see how much we can help out. 

  

42:48 

JOHN: Tom it was great to have you on the show. Thanks for this sort of long view of the semiconductor industry, it's great to kind of hear the perspective of both the way things have been and the way they're going. And then, you know, a broader view of how everybody's playing together and how the government and startups can work together. And thanks for the introduction, again, about ISA and the open standard. Really appreciate it. Thanks. Great to have you on. 

  

43:11 

TOM: No, thank you very much, John, this has been great and I greatly appreciate the opportunity and it's an exciting time and being in the in the semiconductor world in the sales side and business development side. You know, for the last 25 years, this is probably the most vibrant that I've had my career in quite some time. So it's good, good stuff. Appreciate it.