I mean, like Austin, tell me this. Whenever there's any technology comes out, why do you think everybody thinks it's going to destroy HBM? New inference accelerator on SRAM, HBM. New algorithm comes out, HBM. It's always the first the memory companies that go down.
Hello everyone, and welcome to another Semi Doped podcast. I'm Austin Lyons of Chip Strat, and with me is Vik Sekar from Vik's Newsletter. So, Vik, like 10 minutes ago, there was a press release I saw just popped up on my phone. GlobalFoundries files patent infringement lawsuits against Tower Semiconductor to protect high-performance American chip innovation. What's your initial reaction?
Yeah, what? Totally. What did they infringe? I don't know. Like optic stuff or... Well, so that's a good question, because my initial reaction was like, 'Oh, fascinating. This must be about optics or lasers or something, right?' Because that's kind of hot, and that's how everyone is talking about those two companies. Let me read some of it, and then we'll see what we can make of it. And I also think the language is very interesting. GlobalFoundries, a leading American semiconductor manufacturer, today announced that it has filed multiple lawsuits in the US against Tower Semiconductor alleging that it has infringed GlobalFoundries patents by free-riding on decades of GlobalFoundries innovation with an intent to unlawfully take business away from the American chip maker. The lawsuits were filed today in the US International Trade Commission and the United States District Court for the Western District of Texas. So, my first two thoughts: one, free-riding, you know, fighting words right there. And then two, Western District of Texas, you know, obviously there's not much semiconductor innovation happening in West Texas. It's probably a podcast for another time, but I think that's just where kind of patent trolls go to patent things. Do you know anything about this space?
No, I don't. I don't know what is like... What else is interesting to me? It says multiple infringements of patent. Like what are we talking about here? Like what... So one thing I can think of is like that is common to both these companies is RF-SOI technology. And I'm not sure like why they're fighting over this stuff. Like have you seen the mobile space right now? Like nobody's able to sell phones because of these memory prices. Why is everybody stressed out? So, I don't know. Maybe that's one of the multiple infringements they're talking about. I've no idea.
So, it continues to say there's the actions assert infringement of 11 GlobalFoundries US patents protecting high-performance technologies critical to smart mobile, automotive, aerospace, and communications infrastructure. So, you might be right. Yeah. So, it's basically all of it. What is left? Yeah, totally. So, this will be interesting to track. Of course, to your point, I'd love to know like what are these 11 patents? You know, what are they talking about here? 11 patents, okay. Okay, we'll see. Yeah, exactly. I guess last thing I'll leave. Let's see, there's one other quote from the GlobalFoundries CTO. 'There is no shortcut to real innovation. Companies that attempt to extract value from patented process technologies without authorization or investment undermine fair competition and the integrity of the semiconductor ecosystem.' So, again, fighting words of just like, 'Hey, they're cheating.'
I have a question. What do you think would happen if Tower were found in violation of GF patents? What happens to Tower? Because you know what? Tower is a very, you know, favorite company among investors these days. Everybody's, you know, putting money down in that stock, and it's been going up a lot. What does it mean for them?
That's a good question. I mean, at the end of the day, surely there'll just be some sort of financial repercussions. Small if they say, 'No, nothing happened here,' or bigger if it's like, 'Oh yeah, they infringed, and now they need to pay you,' or something. And so from an investor point of view, I just think of like, 'Oh, I guess you have to update your valuation in your model for Tower.' What if it's you they have to, you know, pay a billion dollars or something, and how does that affect their future cash flows? Because I would assume that ultimately they'll just find ways to work around this. Like in other companies I've worked with before, you know, and you might be able to speak to this, maybe not, but you kind of get a lay of the landscape, and then it's like just added constraints. As you're creating a solution, you kind of know where you can innovate and where you have to stay away from. And so, then you think about how to work through that patent landscape essentially. So, I'm assuming it doesn't mean that like, 'Oh, Tower's screwed, and they're going to be dead in the water on anything,' but I just think of it as like a financial slap on the wrist. But again, we have no idea, so we don't know who's at fault or anything like that. But yeah, what's your reaction on how to think about like the outcomes?
Well, I just looked up the stock price. It's down like 5% as of right now. So, yeah, everybody's obviously spooked about this stuff, and immediately the news is like in all these like, you know, outlets are like, 'Yeah, GlobalFoundries files patent lawsuits.' Yeah, so I don't know. You know how these things roll out, right? Like we've seen these patent battles go on for a long time, and then ultimately somebody comes out the winner, whereas in reality neither company wins. Only the lawyers do.
True, totally. And so, the so-called, you know, winning company doesn't always win. Then the fortunes, you know, change, and the winds blow the other way. This is what this industry is about. So, anyway, it is just a part of the industry. Well, we should have an IP law semiconductor expert on sometime and have them teach us more about this, because it's clearly just like a necessary game that companies play, and everyone's obviously always trying to protect their own IP and all their own investments. But what I think interesting is the strategy behind when do you decide to try to enforce your patents, and what is the cost, the opportunity cost, the distraction, the financial cost, and what do you stand to gain by enforcing it? And what you stand to lose by enforcing it, right? Like, you know, so yes. We'll carry on, but we should totally have an episode on this, because it's super interesting.
Yeah, let's see how the drama plays out first. Maybe there's nothing to talk about. I don't know.
True. Yeah, yeah, exactly. Okay, so let's move on to TurboQuant, which was the big news, I don't know, yesterday or the day before. Tell us about TurboQuant. What is it, and why should we be concerned or not, and are people thinking about it correctly? TurboQuant. Yeah, this thing like lit up on X, and everybody got scared that TurboQuant is going to destroy memory companies. I mean, like Austin, tell me this. Whenever there's any technology comes out, why do you think everybody thinks it's going to destroy HBM? Like HBM is like the first thing people think it's going to destroy. New inference accelerator on SRAM, HBM. New algorithm comes out, HBM. It's always the first the memory companies that go down. Why nobody sells off like, I don't know, optics companies if like an algorithm change comes in? I don't know. Maybe it's all connected, right?
Totally. Totally. That's such an interesting point. And I also think like, 'Okay, if we just took stock prices totally out of this, how would people react?' Like when I was in undergrad, anytime I heard of an innovation, I got way excited, because you're like, 'Yes, this is what we learn about in college.' Like the wheels were going. Like things are always improving, and that means when I get into industry, I'll get to help continue improve things, and our lives will continue to get better, and we'll, you know, we'll continue to get more prosperous. And so, it's this funny like, 'Innovation is a good thing, and everyone should celebrate it.' And then you get into a different group of people, and then it's like, 'Oh no, innovation's going to blow up something, and I'm invested in it, and innovation's bad, and HBM is dead.' Meanwhile, every single GPU is all using tons of HBM right now, and it's always more than was being used yesterday.
Exactly. Yeah, exactly. So, this is what happened. Like when we were younger, we'd be like, 'Oh wow, look at this cool invention. Wow, the Google guys are so smart. I want to be just like them. I want to work hard and become a Google researcher myself. It's so inspirational.' This is how we think. But once you get into the industry, and you got a little bit of money, and you start buying stocks and stuff, and you hear like, 'Innovation.' Like, 'Oh my god, my memory stocks are going to tank.' Like, 'Panic. Panic.' So, it's always like any innovation is met with panic. Okay, but first, I think we should explain what is this thing, okay? TurboQuant is a compression algorithm, okay? And when I heard there's a compression algorithm, as did a lot of people on the internet, the first thing was like, 'Oh, this is like the Pied Piper of Silicon Valley.' It's like hilarious. So, there's so many Pied Piper memes. Yeah, the TV show. That's I did use some myself, because I can, and because it's hilarious. So, it's funny that we're living through like actually, there's so many of the scenes from Silicon Valley have like played out in real life. It's so amazing. Anybody has never seen it, like you should go watch it. By the way, that's exactly how people in Silicon Valley like even act, okay? There's like accurate depiction. Sorry, you know, anybody listening to this from Silicon Valley. I have never lived there personally. I've visited many times, although I have lived in California. It's always been San Diego. So, my perception of the, you know, Silicon Valley is that, and it perfectly matches my experience. Okay. Now, onto the real stuff. So, this TurboQuant is again became big news as a compression algorithm for KV cache compression. So, in the past we've spoken about KV cache, which is, you know, the key value cache that is continuously built up in the autoregressive nature of inference. So, what that means is like every token that comes, the KV cache is calculated and then it grows. So, every token that comes in has information of all the previous tokens that came before it and all that is stored in the key value cache, a very important part of the whole AI process and attention calculation. So, because it's such a key component, all of this stuff needs to be stored in HBM because it has to be accessed quickly. Because you have to pull a key value cache, process the token, and then, you know, send it back, wait for the next token to come back, pull the key value cache, and go on and on. So, you need it to be in a high memory tier that can do this quickly. So, key value cache needs to be in a fast memory tier. So, many times what happens is if it can't hold it, it's offloaded into DRAM or, you know, we even spoke about on this podcast with the Weka chief AI officer, Val Berkovici, who explained the whole context storage thing, which means that you can store a lot of the key value cache in actually SSDs in a separate storage rack and connected with high network bandwidth and access it that way. It's not as good as HBM. Like now, we want the key value cache to be as small as possible. That's the thing. Right? And so, TurboQuant is a blog article that Google Research put out like a day or two ago. Now, here's the funny thing. This is not like a day or two old. It is over a year old. Like, if you go read the arXiv paper, it's dated like April 2025. So, it's been lying there in the dust unseen, unloved for all this time.
Well, again, to the point that we keep making, which is it's seen and loved by the researchers and the people who are studying arXiv and what Google's publishing, but it was unseen by investors. So, it took the Google blog to I think to put it in front of them and let them react. But carry on.
Yeah, no, I don't blame anybody because had I seen this paper on arXiv, I'm no ML researcher, okay? So, my understanding of this is sketchy, too. So, I might say stuff that isn't entirely accurate. Please correct me if I'm wrong. I'm totally happy to learn. But yeah, I just had to kind of find out what it's all about. So, I sat down and read a bunch of like papers on it. But yeah, if I had come across TurboQuant on arXiv, honestly, like, you know, so many amazing papers are out there. I would have not thought too much about it. Like, yeah. I think the whole point came up because the Google Research blog makes a really nice case of explaining why it is so good. And they do a really good job with like nice pictures and like layman explanation, layman-ish. I mean, it's still hard for me to understand, okay? So, it's much better than the original paper because the original paper is like full of math. It's just... So, yes, yes. So, the insights are there, but it's kind of like a foreign language and they did a good job of making it easily accessible, easily understandable. Yep. Exactly. So, what this algorithm does is it reduces the KV cache storage needed for, I don't know, compared to like using a floating point 16 FP16 KV cache, using this method compresses the cache by like six times, which is a big thing. And then they tested it on an H100 GPU and it like sped up inference by like eight times, right? That is like amazing because you're like, wait, what? That's like is that like SRAM speeds? You know, memory bandwidth, if you talk about memory bandwidth, HBM has like 8 TB per second. And then if you look at the SRAM LPU like implementation, the Groq thing is like 80 TB per second. So, like that's 10 times faster, right? So, inferencing is 10 times faster by just KV cache compression using this TurboQuant? That's amazing.
Yeah, I didn't catch that part. I knew that it reduced the amount of storage that was needed, but you're saying that it speeds up inference. And is that because now whatever's being done with the actual math during attention or whatever is using like fewer bits? So, just like every calculation has fewer bits or why is it so fast? Or you just have to retrieve less data?
So, I think both. So, yeah, you'll have to retrieve lesser data. Also, the way the compression algorithm works is that it eliminates this layer called normalization that happens in attention calculation. Okay, this is a good lead-up to what this algorithm actually does. Okay, it has actually two parts to it. Okay, the first one is an even older paper from 2024. It uses a method called PolarQuant. Okay, so this is the first part of TurboQuant. The second part of this is called QJL. And it stands for quantized Johnson-Lindenstrauss transform. Okay, like this is like way too much ML for anybody to process, but I'll try to explain it in the easiest way possible, okay? Now, let me start with the attention calculation I spoke about, where before the calculation of attention even, you need to renormalize numbers. And before you do the feedforward network, you need to renormalize numbers, right? So, all of this is in the decoder architecture. So, this is the important thing. So, the normalization has to happen. Now, why this has to happen is because if you don't normalize it during the process of attention and feedforward network, is that there are these really large numbers that show up in the matrix and it like blows up, okay? These numbers get too large, it blows up. So, you have to kind of normalize it in every step you do. You're kind of bounding it, essentially. Yeah, you keep bounding it. Yeah, that's the normalization process. That takes memory overhead. Okay, let's just leave it at that. So, PolarQuant is like a way that it does something slightly different. So, what it does is if anybody has like worked with engineering stuff, this PolarQuant, what it does is it takes a vector, which means it's a bunch of numbers, okay? Let's simplify it. Let's just take two numbers, X and Y. Think of it like a point on a graph, right? The one thing you can do with it is you can transform it from X Y, which is what is called Cartesian coordinates, into polar coordinates, which means you can make it R theta. So, it's now in a circular graph, where you have a radius and an angle. So, R and theta. Quite a common transformation, like rectangular to polar coordinate transformation is very well known all across engineering in any kind of engineering and math. A lot of people run into it all the time. So, that's the whole idea here. So, but before they do that, what they actually do in this thing is they do a random preconditioning step in TurboQuant, especially in the PolarQuant part of the calculation. So, what they do is they take a random matrix, okay? And they multiply it with the KV cache vector. And what that does is two things, right? It kind of preserves the relationship between all these vectors. Okay, like the relative distances in terms of radius and angle, it preserves it. And the other thing it does is it takes these, you know, I told you these outliers with these really large numbers that require renormalization, it kind of takes that and spreads it out amongst everything. It makes the data flatter and more uniform. And then they do the rectangular to polar conversion. So, I know this is getting a little bit machine learning-like, but essentially what happens is very simple. It makes the radius completely break down into a single bit representation. So, the radius as part of this QJL transform that follows PolarQuant is represented by one bit, plus one or minus one in polar form or in bit form you can just think of zero or one or whatever. So, you only need one bit to represent the radius. And by all these complicated mathematical things I was trying to explain, I don't know how well I did, but what it does is it changes all the angles into just a few numbers. The angles are not all over the place. The angles become very clustered around a few numbers. So, think of it like this. If the angles went all the way from, you know, zero to 128, you know, you need like seven bits to represent, let's say, 128 angles. Now, if you're... So, almost like discretizes it. Yeah, yeah, it makes it narrower. So, all of like imagine if it now had only like eight angles it could represent all of this stuff in, you only need three bits to represent eight angles. Yep. In rather than 2 to the power of seven. Yeah. Yeah, 2 to the power of three is eight bits or 2 to the power of seven is like 128. So, if you can cluster the angles into a narrow range, you can use fewer bits to represent them. This is the fundamental idea, that the radius can be discretized into a single bit representation and the angles can be collapsed into just a few numbers that takes only three bits or some two or three bits to represent the angles. Now, you can think of what happens to the KV cache. The KV cache number vector now can be represented with just three bits. So, this drops the memory usage significantly compared to using something like 16 bits. So, this is TurboQuant, okay? I know people might have slept off or dozed off, but this is TurboQuant, yeah.
Fascinating. So, at the end of the day, it's just a clever way to ask, 'How can we use less bits to store this KV cache information?' And some brilliant folks said, 'Oh, let's convert to polar coordinates, and that will solve our problem.' Yeah, really, really smart folks. Totally. So, then of course, the implications are interesting because now we're saying, 'Oh, you use less bits to store your KV cache.' So, of course, the one sort of skeptical response is like, 'Oh, no, we don't need as much memory.' But, of course, actually, like if you think of it like a product manager, if an engineer comes to me and says, 'Hey Austin, guess what? I reduced our KV cache size by 50%, let's say.' Then I'll say, 'Great, shipping tomorrow, instead of 1 million token context, can I have 2 million tokens of context?' And they're going to say, 'Oh, yeah, maybe not quite 2 million, but maybe 1.5 million or something.' I'm going to ship it. Let's go, right? Yes, precisely the right interpretation to make. Totally. It's just I mean, literally, what are neural nets? It's a compression of information into like the world like what is an LLM trained on? It's like the entire corpus of the internet's text, and it compresses it all down, and it finds all the relationships. What's been happening over the last several years? Oh, we find out if you go from FP32, FP16, FP8, FP6 and 4, you can actually continue to compress down and use less bits to store all of these representations, which means you can start to do things like, you know, run it on your maybe you can actually run some of these on your laptop or you can, you know, what's been happening is like, 'No, no, keep running this stuff in the data center.' And how can we have even more memory, even more weights, even more context, right? And so, I think that even more has what has continued to unlock like what's so useful about Claude? Like, well, half of it is it's like a really good model, but the other half is like, 'Dude, I give it like tons and tons and tons of it, you know, earnings and just be like, Look through the last eight quarters of earnings, right?' And before I could do that, it wasn't so useful. So, yes, as more like an eternal optimist and kind of having been a product manager, I hear compressed KV cache, and I think, 'Sweet, we can do more. We can do more in the web.' But, also, you know, I think everyone is always like, 'Hey, this stuff will never run client-side. Maybe it'll be on a workstation on your desktop,' which I totally believe in. But, like memory is always going to be a problem, but it's like, 'I don't know, dude. Now we're storing stuff with three or four bits. Maybe there is a world where with continued algorithmic innovation, there will still be some frontier, today's frontier maybe, but some level of very interesting and helpful, you know, AI generative AI that actually can run locally, thanks to innovations like this.'
Yeah, so I agree. So, there is this aspect that yes, some of this might make it into client devices and edge inferencing because actually you're reducing the problem down with clever vector quantization methods, which actually does not lose any performance. Like the original you can just see the Google Research blog, too. They show some benchmarks of what is called a needle in a haystack benchmark, where you give it a lot of information and try to find one specific thing from it, and that it has, you know, has a good score. It does really well compared to things that don't have this kind of quantization. Normally, quantization loses information. So, but this seems to do really well. But, yeah, so that's really good, right? Because we can now serve longer context. You can serve it on maybe smaller devices, or maybe if you continue to have your cloud data center, you can be like, 'Wow, I can serve maybe more users.' Yes. That's a good unlock, too, for people to think through. Like, for hyperscalers, it's more concurrent users with the same infrastructure. Yeah, yeah, yeah, all of this stuff. So, that's why this is actually a positive thing. It unlocks more use cases, more bang for the buck with existing hardware that we could not have done before. Totally. Oh, and here's the... Yeah, you go ahead. And those Hopper H100s are even more valuable now, even though they have HBM3 and not as much as what Vera Rubin has, right? So, counteracting the whole like, 'Oh, yeah, these things depreciate and get worthless.' This kind of continues to show what Jensen has always said, which is like, 'No, we continue to come up with better software and can squeeze more out of even older hardware.' Yeah, so, you know, this is a kind of the thing that keeps happening, and we cannot I mean, we're going to continue to do this and like get scared and sell off at every occasion, but I think that kind of lesson to learn here is yes, initially we had methods to reduce KV cache size. You can think of, you know, grouped query attention, multi-headed latent attention, which is like DeepSeek, right? Those were big movements in the history of how we've been doing these things. And the moment DeepSeek came, yeah, there was a DeepSeek moment, everybody was crazy about it. I think it was early 2025, if I was right. So, almost it's just it seemed like it's just a year ago, yes. Seems forever in AI years. But, ever since the DeepSeek moment, which reduced the KV cache usage significantly, we still haven't run out of like HBM requirement, right? I mean, we're still buying and trying to get more HBM. What happened after DeepSeek? Nothing. We need more memory. Totally. Yes, totally. So, those were methods that try to reduce the size of the KV cache. Then there's been a lot of research into people who'd like researched on what is the best strategy
to, you know, evict KV cache from like HBM to DRAM, DRAM to context storage on SSDs. So, there's a lot of research on that kind of stuff. This is another kind of research where normalization, how do you, you know, quantize and normalize differently so that you can get more, you know, bang for the hardware buck by numerical and mathematical methods. So, all these things are always happening, and it's not like this is something that erupted out of the blue. This stuff has been around for a year. There have been many methods before it that didn't get all the attention, but instead of, you know, 6X compression, it probably did 4X compression, you know, compared to FP16. Those literature reviews are out there. So, nothing makes news until like Google puts it up on their blog. But, these things are going to continue to happen, right? This is not the last innovation. This will continue to happen. So, always, it only means that we're going to get more out of our hardware. And in a time where chips are the fundamental limitation to growing compute, it is very important that these innovations continue and will continue to unlock things that were not possible before. Because we just can't keep making more chips.
Totally. Totally. Yes, to your point, there will always be a memory hierarchy tier, SRAM, HBM, DRAM, SSDs. And as we continue to get really interesting innovations, we might, of course, be able to store more and more in those lower tiers and access it faster and faster or for cheaper and cheaper, but there will always be a desire to have that which we hold most precious in the very expensive SRAM, in the very expensive HBM, right? It might just be like, 'Oh, now it's more about like, hey, we're going to store, I don't know, some like really precious memories that we access all the time in this stuff, and everything else can be stored elsewhere, right?' But, we're never going to say, 'Good, we don't need SRAM or HBM anymore, you know.'
Yeah, yeah, totally. All right, so, let's carry on. How about we talk Arm AGI CPUs?
Yeah, let's do it. You were at the Arm Anywhere event, right?
No, what do you call it? Anywhere?
Everywhere, everywhere. Yes, yes. Although, anywhere could potentially work, too, you know.
No, no, rolls off the tongue. Arm Anywhere is the...
Yeah, yeah. Yeah, anywhere. Totally, totally. Um, yes, so I was in sunny San Francisco. It was very sunny, very nice, just beautiful weather. And Arm launched, you know, the big announcement is they're making silicon. And so, now I thought the framing was really interesting as to why Arm, an IP licensing company, would now change their business model, or I should say add a new line of revenue to their business model, which would be actually selling silicon instead of selling IP. Um, I'll kind of walk through the framing again, in case anyone didn't see it. And of course, you can push back, you can give your takes, whatever. Obviously, so first of all, the timing is really good. So, when I talked to some of the Arm executives, I was like, 'I don't know if you guys, I don't know when you guys started this, but you obviously timed it very well with Open Claw it showing like, oh, agentic AI super useful, super powerful.' And then of course, Claude Code. And then of course, just whatever, a week ago, Jensen is saying, 'Hey, agentic AI is placing such demand on CPUs, we Nvidia feel it's so important for our system performance that we have racks of CPUs right there, because the CPUs are the bottleneck.' And so, this is the environment in which Arm gets to come in and basically say carry it further and say, 'Yes, agentic AI is creating this bottleneck. You've got this CPU orchestration that needs to happen with all this agentic tool calling, web fetching, all that stuff. It can't just run on the head node CPUs. We need more CPUs, but they need to be right there to keep the GPUs fed.' Therefore, yes, there is this need for agentic AI CPUs, but then the interesting framing that they took was they said, 'Hey, before agentic AI, we think that you needed something like 30 million cores of CPUs per gigawatt, and our calculations are in an agentic AI world, you need 120 million cores per gigawatt.' And so then the framing that they really wanted to hammer home was like, 'We think that there is now an inflection where you need 4x CPU cores in the same power envelope per the same gigawatt, and therefore then the question becomes okay, if I need all these CPU cores and they need to be very power efficient, who should I buy the CPUs from?' Now, I would add my own little wrinkle in there, which I think part of the interesting framing is you should ask, 'Okay, well, where are these CPU workloads running today?' Like, what are the head node CPUs that are running but becoming the bottleneck where you just need a bunch more of them? And the interesting thing that I've been paying attention to is like, in the early generation of Hopper, those head node CPUs were often Intel Xeon or maybe AMD x86. Then the next generation, it was all Grace Blackwell for the most part, not all, but like the vast majority. So like, all of those AI workloads are already running on Arm. So the whole like x86 legacy, all that stuff, like that's kind of different. That makes sense for like the whole web conversation, but for the AI conversation, AI workloads very quickly went from running on x86 head nodes to running on Arm head nodes. So then it might make sense to ask like, 'Well, can I just buy Arm CPUs to sort of expand I'm already running stuff on like Grace Blackwells or Vera Rubins? Can I just get more?' And of course, yes, Jensen has said you can get more Nvidia CPUs. You can get more Veras, but you might ask at what cost and are there any alternatives either from a cost savings or a supply chain diversity? So then the next natural question is like, 'Okay, well, who else sells Arm CPUs that I could buy that are made for agentic AI?' Of course, there's Ampere or Qualcomm who are trying to make like web server CPUs and there's all the cloud CSPs who have made their own cloud native CPUs, but has anyone designed a CPU for agentic AI? And into that, Arm has an opportunity to come and raise their hand and say, 'Hey, we already did this CPU IP. We've already done the CSS, which is like a fuller compute subsystem that's hardened. Like, we're just going to take that and make a full chip and sell it. So hey, we're Arm, we're here, we've got a rack of CPUs that you can buy.' So I'll pause there. What is your reaction?
Yeah, I have a lot of things to say. So first of all, as an AI architecture, AI CPU architecture, I don't think it can really be decoupled from what it used to do in the web era. And the reason is that whatever we do agentically now is a complex mix of what CPUs used to do in the web era and what it does now with GPUs, working with GPUs. So I think there is still some compatibility layer issues that exist, but I don't think they're that big of a deal, okay? That doesn't mean like, 'Oh, it's Arm.' Just because it's Arm, you can't use it and it's like dead because of that. No, I don't think so. But yeah, you know, at some level there may be some advantages to x86 and its legacy depending on the task at hand. That's one thing.
