On this episode of The Six Five Webcast, hosts Patrick Moorhead and Daniel Newman discuss the tech news stories that made headlines this week. The handpicked topics for this week are:
- Intel Vision 2024 Event
- Google Cloud Next 2024 Event
- Apple M4 Chips & New Macs
- TSMC Gets $6.6 billion CHIPS Act Funding
- Marvell Accelerated Infrastructure for the AI Era Event
- U.S. Inflation Data
For a deeper dive into each topic, please click on the links above. Be sure to subscribe to The Six Five Webcast so you never miss an episode.
Watch the episode here:
Listen to the episode on your favorite streaming platform:
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we ask that you do not treat us as such.
Transcript:
Patrick Moorhead: The Six Five is live on a Friday morning. It’s the middle of April. Dan, we almost made it through the month. It’s great to be back here at home. We were on the road for two events in three days. We were,-
Daniel Newman: Pretty normal.
Patrick Moorhead: Yeah.
Daniel Newman: Pretty normal, right?
Patrick Moorhead: Pretty normal. We jetted out on Sunday, went out to Intel Vision in the middle of the Arizona desert and then we made its way back to another part of the desert, which was our second home Las Vegas. It was really good. Dan, how are you doing this morning? We made it 212 episodes.
Daniel Newman: 212 episodes. Another jam packed action packed week. Feeling very chippy this week. There’s lots and lots of chippy announcements, new instances, new silicon, hyperscalers building more of their own. There’s more AI, there’s more stuff. I mean Pat, I think you were like the what, 73% of the media was Pat Moorhead this week. You were on fire, CNBC. I got three pictures of you yesterday sent to me by friends and I’m like, yeah, you know how to rub it in guys. Now I’ve been reminded that I’m supposed to just be happy for you, so good job. No, but and seriously you were all over the tube this week. Google Cloud, Intel, Facebook, Apple. I mean Apple’s got an AI PC. Nope, didn’t say it. We don’t want to tease the topic.
Patrick Moorhead: It’s okay. But hey, we do have an incredible show for you. Come on back in, Dan, we missed you. We’re going to be talking Intel Vision event, Google Cloud next, Apple M4 with new Macs coming up. That was the leak to Bloomberg. TSMC got some money for CHIPS Act. We’re going to break that down. We’re going to talk about Marvell’s accelerated infrastructure for AI event that was held yesterday and we’re going to follow, we’re going to end here with a non-controversial topic like US inflation data. Real, fake, does doing a quarter year over year, every sense or should we be looking at kind of a three to maybe five years? So Dan, I always mess this up, but I need to talk to our producers about this. I’m going to call my own number here and that is the, I always do this, man.
Daniel Newman: Our producer sucks.
Patrick Moorhead: No, really he does. I’m going to fire him so hard. Anyways, lets dive in here. So the Intel Vision event was a 50, 50, I’ll call it NDA versus public event. I always need to watch very carefully what I say, publicly but in addition to analysts, Intel had its biggest customers, its biggest partners here at the event and a ton of Intel salespeople and of course executive, Dan, you and I got the chance to chat with the big guy, big Pat, Pat Gelsinger in addition to the leader of the PC group.
Daniel Newman: Michelle Johnson, Justin Hotard.
Patrick Moorhead: Justin Hotard, just a lot of great stuff. And also the leader of Edge as well. So the executive access was incredible. So two big announcements here. One was Gaudi three and there wasn’t a tremendous amount of new information about Gaudi three, but it is sampling and for certain workloads it does operate quicker than NVIDIA for training and inference. And you might be saying, “Hey, a 50% improvement doesn’t matter,” but it does. This is classic accelerator versus GPU. If your accelerator can’t be faster and more efficient, you’re doing something very wrong because GPUs on the whole are more easily programmable but less efficient. And the big story that Intel came with is its programmability across its multiple variants of silicon.
Now, one API doesn’t hit it all yet. I think Intel will sell, I would say single digit billions of this product before we get to what I think is going to be more of a transformative opportunity for Intel. And that’s 2025 when you have the combination of CPUs with onboard inference, when you have Falcon Shore GPU sprinkled with Gaudi three ASIC juice inside of it and then connected with an end-to-end software platform. And Matt is my AI hardware guy. He’s piecing together the TCO story kind of on a generational basis because the one thing we know with GPUs is you’re going to get three to four generations of training or inference out of it, whether you’re doing ML, DL, generative AI, mixture of experts, whatever’s going to be next here.
And I think that’s a critical piece when you’re a hyperscaler or even an OEM. I did get some confidence with Michael Dell showing up with a really cool video and listen, he was the only executive called out in Jensen’s piece. It’s not like you talk about AI accurately being a zero-sum game here, but Dell is clearly excited about the Gaudi three. I need to do the double click to see exactly what that means. The final thing, I don’t want to steal everything here is on AI PC, we got clarity on the Lunar Lake roadmap. Lunar Lake systems could literally ship in October. Right. In my head I was always thinking Lunar Lake was December 32nd, right? Lunar Lake has 45 NPU TOPS, 100 TOPS system-wide. But I was always expecting it to miss the holidays and I think this is a positive thing overall in the market.
Qualcomm will still be first with big NPU and I do believe that from a power PPW, I think Qualcomm will likely still be on top, but for Intel to be pulling this in is monumental. Listen, I’m not Babe Ruthing it, right? They could find something late in testing that requires a metal spin, but boy are we getting closer and closer. So I think what we have here is a three-horse race between Qualcomm, Intel, and AMD. Qualcomm I believe will have the operating system first. AMD and Intel will likely get it in the first part of 2025, but it is on baby. The super cycle is strong and I talked about this on CNBC and Yahoo Finance.
Daniel Newman: Anywhere else? The Six Five dude. You talked about it on The Six Five. Don’t worry. Everybody needs to hear this once again. Pat, the Intel show to me kind of pointed in two directions. One is in the PC space in particular, you just can’t rule the company out. I know so many people, the Twitterati all want to basically say it’s over, right? Everything’s going to be an AIPC. Intel’s late, they lost. Whether it’s been the articles saying that Qualcomm’s big NPU was going to dominate the market, whether it’s been AMD first to market shipping the, what was it, 17 TOP AI PC. Pat flashed once again, what is it? 40 million quote-unquote AI PCs.
Now again, what an AI PC is really hasn’t been starkly defined as one thing or staunchly is the right word. Staunchly defined as one thing. I mean, we know that 10 TOPS people have questioned that, whether or not that’s real AI PC, but Intel has made the claim it’s sticking in the market. People get it, but the thing about Intel is you just can’t underestimate the provenance of the relationships that it has in the go-to market. It has had the retail relationships, the distribution relationships for the longest time. It has the OEMs, the designs, the scale, the MDF program, the support, the ecosystem, the advertising, and the marketing. Disrupting that is going to be complicated. It is not a, Oh my gosh, we come up with a better mousetrap, we have a better NPU, we have better integration potentially with a Windows on Arm and everyone’s just going to flock away. It’s not how behavior works.
Like you said, we’ve had better Android designs than Apple phones for long times and people won’t leave it because of one stinking feature. We are going to see a slow distribution. I do think Qualcomm’s doing some great things. I do think AMD has some strength in its roadmap, but I think Intel proved and by pulling this forward, I think it was almost like a little card trick up the sleeve like, hey, here we go. You thought you were gaining some market advantage, but we know how to play the game too and like you said, having this out by Christmas or holidays, that’s a huge unit volume opportunity. That’s going to change the calculus as people like to say. On the data center side, the Gaudi thing, Xeon 6 is obviously pretty important too, and I mean we didn’t spend a lot of time talking about it, but Pat, you and I have had this kind of repeat cycle of trying to explain to the world, like a lot of inference is going to be done on CPUs.
All this existing cloud data centers and these data centers that have been built around general purpose computing are still going to be there and they’re still going to be doing things to inference and AI against data. Our ERP systems, our supply chain management tools, our CRM systems, a lot of the types of inferencing we need to do does not require a big GPU. And also obviously new architectures that might include extensions plus Gaudi plus Xeon six will be able to offer some really powerful optionality for customers. We’re seeing an entire market coming together for the first time in history to have an answer to NVIDIA. I call it the NVIDIA is Apple and everyone else is Android.
And so there’s a desire in the marketplace to have an Android, to have another alternative. This doesn’t mean that NVIDIA doesn’t have a tremendous success path in its future. It just really means AI is really big and there’s going to be a really significant distribution of how AI gets deployed. It’s not one way, it’s not one thing, it won’t be one company. Intel has a long road ahead to try to play in AI. They do need to get their GPU to market. They need one API at scale. That abstraction to be able to handle all the software and make it very, very easy to compile and put on Intel hardware will help them gain momentum. But the TLDR, as I like to say, is you can’t rule them out, not on PCs, not on data center. And of course we didn’t even talk about Foundry, but we talked a lot about that last week. We’ll come back to that another time.
Patrick Moorhead: Good stuff, Dan. Let’s move to the next event and take a deep breath. Dan, get a gulp of coffee because I’m calling your number dude. Google Cloud Next 2024. What happened? I think it was a whole lot of AI.
Daniel Newman: Gosh, so much AI. Look, I want to talk about the macro of Google. So you and I have both said this. Look, Google’s had this weird stop start across their early launches and they’ve sort of had a couple fumble the ball moments. Now I want to give them a lot of credit because they had to fumble the ball moment in almost all of their first three launches in some way, shape, or form. And yet at the same time at Google Cloud Next, they absolutely came out and reiterated that their provenance in AI is so deep. I mean, the world’s most prolific and important data set by volume is with Google.
What does that create? It creates a massive opportunity to build AI capabilities in their cloud and of course across their entire ecosystem. Let’s be very, very clear what Google is able to do because it has itself as customer zero is build some pretty incredible AI capabilities in its cloud to be able to support its solutions, whether that’s going to be YouTube, whether that’s going to be search, the next generation of search, generative search. And so having the chance to go there, listen to customers, listen to executives, listen to Thomas Kurian, what blew me away. So first general compute offering Axion of course was this star of our show because Pat, we get more media inquiry I think, than anyone on the planet when it comes to chips.
But again, chips is the foundation. Silicon is what enables all this cool software. So all this generative capability, all the stuff we saw with workspace, with vids, with meet capabilities, of course Vertex for enterprise, for applications, what these companies are doing, it all starts with chips. Google has basically finally, or they’re moving away from kind of homegrown silicon just for their own purposes to now democratizing and offering homegrown silicon at scale. Who’s done this before? Well, they’re following the footsteps of AWS. They’re following the footsteps of Microsoft, which on the whole compute thing, the TPU though, of course, which they offered a new, their next generation clusters and of course, what is it v5, Pat? I’m trying to remember just because I’m talking so fast.
Patrick Moorhead: V5p.
Daniel Newman: V5p. You and I tweeted this. Probably our most successful tweet of the whole event. Dude, they trained their entire Gemini Pro model, their Gemini stack, all on their own hardware. Now, I don’t know about you, but I’ve been told many times by that same Twitterati group that also told me Intel’s going away, that the only way to train an LLM is on NVIDIA. And it turns out that Google has built full stack from hardware to software for LLMs all on their own engine, on their own silicon. Now Pat, this is really interesting and important because two things. One is they’re doing it for themselves, but now they’re democratizing it for customers. So customers are going to be able to use Vertex, democratize all the models. They by the way, don’t have use Google. This was a really important reiteration that they made is you can use them, but you can use Anthropic, you can use Hugging Face.
It’s a use your own adventure when it comes to LLM, but you can run it on TPU. I think this also reiterates my point about Intel and kind of the, everyone in the world is collaborating with NVIDIA like Google, but also has a very strong hedge to NVIDIA because they know they need it. But Google’s done an amazing job. They’ve built a very powerful accelerator, XPU, and of course they’re offering all the merchant’s silicon. We can talk about this all day. My last thing is Workspace. Pretty cool stuff coming out of Workspace Pat. I saw some of the capabilities with vids. I saw some of the next generation capabilities with Meet. I saw some of the next things in terms of document management and file management and generative content creation. Pretty blown away by what it can do.
Now, I’m going to tell you what I told Google’s executives about this is there’s still this little bit of a gap between these technologies and reality for me. Sometimes you see it like a vids Pat, like you and I, like all the hard work we do to put together a video for an interview or a customer, a reel, and they made it look like this can all be done in two seconds by uploading some stuff to photos and some stuff in G Drive and Docs. I’m really eager to see how well this stuff works. But the next like too long, don’t read, just read me, is that it’s changing quickly. This probably doesn’t work exactly as I need it to today, Pat, but in the next year or two years, so much of the sludge of work that we do with stuff, if we use Workspace, it’s becoming really competitive.
And by the way, when you see what they’re doing with Workspace and how much it could help an enterprise potentially like a co-pilot can, you also go, I get why they might want to try to buy HubSpot. And I know that’s kind of a rumor. We didn’t have a segment on that here, but that rumor has continued to persist throughout this week. Google has the best data. Google has the full stack from your enterprise data now. They can take your customer data, your advertising, your behavioral data, and then they can put it into a CRM. Is Google going to be the next big enterprise software play? It gets kind of interesting, but they’re doing some really cool things with generative AI. Great week.
Patrick Moorhead: Yeah, you pretty much talked about everything, Dan.
Daniel Newman: Did I?
Patrick Moorhead: I’m going to try to do some gap fills here. No, we all do that. We go back and forth. So one of my big takeaways, and again, I’m not always the sharpest knife in the drawer here, is this ability for agents or co-pilots to rewrite the enterprise SaaS market. If you have access to the data, and again, Apple didn’t say, sorry, Google didn’t say it out loud, but when you look at employee agent, well that could be a human resource agent. It could be a marketing demand gen agent, it could be a marketing brand agent. Right. Why do I need these CRM tools? Again, this is why it’s imperative that all the SaaS companies jump on this and integrate it quickly. I thought that was pretty interesting.
I continue to be impressed by Vertex AI. It seems easy, and again, I’m not a data scientist or a computer science person, but when I look at, I think communication is half the battle and it looks like the simplest end-to-end platform that I’ve seen, and them adding RAG capabilities to that and getting a little bit more respect for how enterprises use some of their data to do that. And by the way, even leveraging Google search in addition I think is pretty powerful. Like you, I was impressed that TPU v5 did all of the Gemini training that was out there. I got a lot of good shame out there on X saying,-
Daniel Newman: Your tweet was epic.
Patrick Moorhead: It said, “Hey, idiot. They already told people that.” Well, maybe I missed it. And you know what? Based on the hundreds of thousands of people who saw interest in that, I wasn’t the only one who missed that.
Daniel Newman: I did.
Patrick Moorhead: Yeah. Workspace, I got to tell you the RAG capabilities that they put in there, the ability to plop in 100 resumes, do compare and contrast, put in four proposals for, I don’t know, video production companies and to be able to create a grid and compare and contrast them. I couldn’t help but to think, gosh, is this a risk to Office? I went in and played with Copilot. Heck, Dan, I’ve got the highest grade Copilot I can get for Microsoft and I can’t even dump in files in like, I can even a GPT-4, or GPT-4 Turbo and competition is good baby.
Daniel Newman: Vertex is solid though, the Vertex workspace, like they’re off to a good start.
Patrick Moorhead: Yeah, for sure. I got to talk about Axion. Right. Three to four internal organizations using this general purpose CPU already. Google has the largest data state on the planet. They don’t get a lot of credit and Google Cloud folks don’t like to talk about that, which always kind of stymies me. The one thing I still think Google missed big and a lot of hyperscalers are whiffing on this is how do I leverage on-prem data? And the answers seem to be, oh, just take it to the cloud. I’ve talked to a total of zero enterprises that are interested in wholesaling, moving their data up to the cloud. So come on folks, let’s do this. I want to see a Google distributed cloud solution with infrastructure from the players who are in there to be able to do that, right?
And AWS is doing kind of a local zone strategy more than they are on their distributed cloud, but it’s still very frustrating to me. The final comment that I think Google needs to work on, and then you brought this out there, is the stumbles that they had with three of their launches, and I brought this on Yahoo Finance. That bleeds over to the commercial market. And I think that Google needs to do some hard work on proving that their AI is industrial strength, not necessarily coming up with odd images and some odd results. Anyways, I’ll leave it at that. Woo. Big event, baby.
Daniel Newman: There was some oxygen there because I kind of just breezed past Axion, I mean with Neoverse two and how it’s designed and how is it differentiated, and I think we need to come back and do a segment on this one because we should do something Pat in the future that kind of looks at Graviton Azure’s offering and of course what Google’s building to kind of talk a differentiation and what’s really the value prop because everyone got this kind of more performant, more sustainable, but I feel like it’s the same slide. What really is different and when do companies pick these things versus the merchant stuff that is still so pervasive?
Patrick Moorhead: Yeah. Let’s go into, by the way, we should do that, but we have to do it quick because this thing changes about once a month, right?
Daniel Newman: Yeah. But it keeps us busy, doesn’t it?
Patrick Moorhead: I know. No, no. Pandemonium loves analysts.
Daniel Newman: For guys that know a little bit about chips.
Patrick Moorhead: And analysts like pandemonium. Hey, let’s jump into Apple M4 chips and New Macs. Apple leaked to Bloomberg that they were,-
Daniel Newman: Remember when Apple didn’t leak dude? Remember when that wasn’t a thing? It feels like everything leaks now.
Patrick Moorhead: Well, the only time Apple goes early on anything is, is when they know that they have heavy duty competition or something like that. And I think investors needed some positive news after they’re getting dumped on for China, getting out of automotive and kind of the lukewarm response to a pretty good product, which is Vision Pro. So yeah, again, congrats to Bloomberg on this. So I’m going to put this in perspective. So last year I talked about a PC super cycle that would start in mid 2024 led by on-device capabilities from Qualcomm, Intel, and AMD. And Qualcomm led the way with their Snapdragon Elite X announcement last year with giant on-chip AI and ARM-based custom CPU architecture that showed, and they’re still showing benchmarks, better performance per watt than Apple’s M3. And then Intel, right, announced at CES in January, Meteor Lake that had I think around 10 TOPS in there and an improved distributed architecture.
At the event last week, like we talked in our earlier segment, Intel did what I consider a pull-in of Lunar Lake, right has, I don’t know, six X, the NPU capability is and 100 TOPS and it would be available for the holiday selling season. And obviously AMD is in the game too. I don’t want to leave them out. They were technically first with an on-chip TPU, but I don’t think that has got them much because of the software enablement is weak. So yeah, I mean I got to give Apple credit. Apple was running the tables from a performance per watt and a battery life. Now it lacked desktop overall performance because it couldn’t connect with NVIDIA, but they were really focused on PPW. So Qualcomm, Intel, AMD are challenging Apple’s notebook supremacy.
So what do I expect from these chips? I expect a big NPU baby and potentially some improvements in the way it scales its architecture. While we’ve seen some interesting thing with electron microscopes on how M3 architecture changed, I am not expecting a radical IPC improvement. I don’t think that that’s what we are going to see. And then on the desktop side, I think Dell, HP and Lenovo have an advantage with NVIDIA AI and all the RAG capabilities that NVIDIA has put into their latest and greatest.
On the plus side, again, I think Apple’s entry into the AI PC market will accelerate the category and overall this is a good thing, right? They’re completely vertically integrated, which is interesting and I’m interested to see how Microsoft Surface and it’s integrated with its new operating system, AI enabled with the applications and new chips first out with Qualcomm, Intel, and AMD are going to mean. But the great part is Apple jumping in gives credibility to AI PCs.
Daniel Newman: Yeah, listen, this is all rumor right now, but it’s not. Right. This is like it’s rumor, but it’s not rumor.
Patrick Moorhead: Right.
Daniel Newman: Yeah. And Apple one, doesn’t have its same capabilities to secrecy as it once did, but this isn’t an accidental leak anyway. This is a, hey, we need everybody to know we’re not basically doing nothing here. Now again, there was a lot of prognostication that Worldwide Developer Conference would be a moment and you would be hearing about what we’re hearing about, but I think waiting multiple weeks, months to get this out in the world we live in is too long. Now Apple’s,-
Patrick Moorhead: Clearly responsive and Intel was the clear driver.
Daniel Newman: To its credit, I mean, Intel is putting the fear of markets into Apple and the fact is, is that Apple, they haven’t been on the front end of anything for a long time. I mean, let’s be very, very clear. They’re late with VR and AR. This isn’t new. They’re out with something cooler, but it’s much later and that’s kind of become their bit of their MO, is they’re kind of a laggard in a lot of things, but then they do it a little bit better, a little bit more friendly experience wise and that’s been their thing for a long time now. It’s not Steve Jobs’ Apple anymore. Let’s just be very clear about that.
But Apple’s also suffering Pat, like let’s be really candid. They’re getting crushed in China right now. Their iPhones are down by mid-double-digit percent sales. They’re moving sales and manufacturing to India, which I think China’s somewhat punishing them for amongst also trying to punish any US-based company in general for what we’re doing from a chip control standpoint and they’re getting more nationalistic. That hurts a company like Apple that has a huge market there. Now again, this isn’t about phones, this is about Macs, but Pat, their sales dropped 27% for Mac in the last year, so that’s not good. That’s a big fall-off. So now you’ve got phones falling in their biggest market, you’ve got Macs falling off and then they didn’t have anything innovative to offer. So of course they’re going to come out and say, hold on everybody. We’re not going to go without a response here. We’re going to come up with something and it’s going to be good.
Now, I did some evaluation of the ReALM or ReALM models that they’re doing, Pat, and the stuff that Apple is doing with Gen AI on-device is actually somewhat interesting. If you looked at the stats versus I believe it was some of Mistral and they did it against GPT-3.5 Turbo, I believe, and again, I’m doing this from memory, Apple was able to show some pretty high performance on models with sub-billion parameters, some as low as like 250 million getting very, very good performance. This is going to be related to future versions of Siri and generative capabilities on-device, but Apple needs an answer for on-device AI. This is going to be the answer. Will it drive, go ahead. I can see you trying to get something in. Let me take a breath.
Patrick Moorhead: No, no, I thought you were done.
Daniel Newman: No, I’m never done. I just talk forever. I just keep going. Pat’s like, yeah, I know. Jesus, this is terrible. You know what? Let’s just move on. I’m done with this.
Patrick Moorhead: You sure?
Daniel Newman: Yeah, yeah, yeah. I said my piece. I said my piece. Apple, you got work to do. There.
Patrick Moorhead: Yeah, credit to Apple for stirring up PPW and resetting what you could do. It woke the entire industry up and their,-
Daniel Newman: And by the way, it’s more TV we can do. It’s more stuff for us to tweet about.
Patrick Moorhead: Exactly. I get it. Okay. Let’s move to TSMC. All the American companies getting money. Dan, what’s going on here?
Daniel Newman: I don’t think the T stands for US. By the way, I don’t think the U and UMC stands for United States either, but I don’t know who the next company that’s going to get money that is not American. So the CHIPS Act was really all about building resiliency for the supply chain, maintaining technology leadership amongst the kind of alliance countries in the Western world, meaning everyone that’s not China, Russia more or less, but really China, Russia’s not doing anything with chips. If you want heating oil, you can go to Russia. The bottom line of it is though is so last week or two weeks ago Pat, Intel got its award. It was like 11 billion plus, no, it was 8 billion plus another 11 billion in loans. And who else is getting money? We’ve seen others getting money, we’re hearing Samsung is getting money. GlobalFoundries has gotten some money.
But TSMC got a big grant award from US 6.6 billion, almost as much as Intel, which to me was kind of wild. And they got another what, 5 billion more or less in loans. And this is three chip factories that they’re going to build in Phoenix, right, or Arizona. They’ve had a hard time building these factories, these fabs. It’s taken them a long time. And so will this money help them accelerate that? Now again, I believe Pat, this is five nanometer is what they’re talking about this being, their five nanometer process.
Patrick Moorhead: Two nanometer.
Daniel Newman: It’s two. I’m going to go back. I believe this is their two nanometer process and this is a pretty considerable loan and support for TSMC to get there because beforehand they were really kind of only coming down to what would be kind of an older leading nodes like seven, and they hadn’t really executed or built anything here. Here’s my take on what happened though Pat, is I think the fab less companies, while they are willing to consider Intel and starting to work with Intel at the foundry level, I think they are very happy with TSMC and I think they are generally putting some level of pressure on US regulators and policymakers and those that are doling out the money to basically make sure TSMC is here for two reasons. One is we want competition in the US for leading edge foundry, and two, they want to continue working with the company they’ve spent, invested heavily into be able to build their chips.
I also think Intel, while I do feel optimistic about their foundry prospects, hasn’t really been able to prove yet that it can do the foundry thing at scale. So if we actually want to execute this leading edge manufacturer here in the US, we need diversification. So that’s why I think it was so much money, it ended up being a lot more than expected. It kind of went against the grain of real nationalism. Now we’re having more investment in Samsung and more investment in TSMC and even global foundries is Middle Eastern sovereign money. I mean, but at the same time it’s what they needed to do. So I know we’re running quick and we got to keep moving. So my take here is good win for TSMC. Overall, probably good for US, that we’re going to have a diversification of manufacturing. Kind of a kick in the teeth for Intel. But look, Intel executes, I think Intel’s foundry business has a good opportunity.
Patrick Moorhead: So CHIPS Act, US wants to get 20% of leading edge in the US by 2030, and this is a good step in the right direction. Listen, more competition is good. It increases innovation, it lowers prices, which in the end is what makes the world go round here. So I’ll be honest, it’s a little murky at exactly when we’re going to see two nanometer in Arizona. I believe that there was some sort of discussion to get this money that would be okay, you can’t be N minus one and if it’s truly 2028 before it’ll be online, they will be very much behind Intel on that and they’ll have two nanometer in Taiwan earlier. So I’m piecing the dates and the technology together, but one thing that I do know for certain is TSMC hasn’t broken ground on their new facility yet Intel has in Columbus,
And listen, Intel has had fabs in the United States forever. They know how to bring that ecosystem together in the United States. TSMC is less experienced at that in the United States, and quite frankly, there has been some friction with TSMC and the unions. I’m going to paraphrase, TSMC essentially calling US workers lazy. And when you’re trying to go for multiple billions of dollars, that just creates a political mess. And I will tell you, having talked to senior executives at every one of their US based design customers, they cringe and TSMC needs to get its act together, quite frankly.
Daniel Newman: Domestically.
Patrick Moorhead: Yeah, domestically. Their public posture is very much a monopolist and when you’re on top, and yes, they are on top as a foundry, but in the end, more competition is better. And with that said, I’m going to move to our next topic here. Marvell had an event called Accelerated Infrastructure for the AI Era Event. It was held in New York City for investors and from a top line basis, I thought it was a very positive call, sorry, video that they had.
And a couple of things to pull out here is that their AI revenue is growing incredibly well, right? Let me make sure I get the numbers right. $550 million in fiscal 24 going up to 1.5 billion in fiscal 25, that’s 272%. That growth is pretty much NVIDIA. Right. So them and NVIDIA in terms of growth rates, not whole dollars here, they’re dramatically different. So I think people kind of miss that. The second thing is we saw two new AI accelerators announced, which is super exciting. And I have to think with Hock on the board of meta, right, we’re looking at Microsoft, AWS, as Google between AI accelerators and this new Arm CPU. I have to think that this Arm CPU that really was never announced and didn’t have new on it was to detract attention from Axion launch the day before. I don’t know. We don’t do rumors here, but piecing this stuff together is pretty exciting.
The TAM opportunity, right, growing from 21 billion, and by the way, these aren’t our numbers, these aren’t my numbers, and I need to do the double click on it. Going from 21 billion last year to 75 billion in 2028, I think that’s a 28% CAGR. And then the objective there, and they were very clear and very few people do this, they said they were going to double market share for the long term, going from 10% to 20% of that TAM. If they do that, and if you buy into that 75 billion TAM, that is a gigantic number and about zero of it is integrated into their financial. One other piece of news, just to end this out, I know I’m bouncing all over here, my apologies. They said they were going to sample a 1.6 terabit coherent DSP in the data center, not across the data center, in the data center. And a few more details on silicon photonics. It looks like those acquisitions in silicon photonics and of course not that copper’s going away, it increased innovation on that as well.
Daniel Newman: Woo. Good stuff there, Pat. By the way, there were some really great presentations. I have took so many screenshots of them sort of doing the front end back-end network configuration and kind of walking through how all the connectivity that goes from in the rack out the rack. Pat, it’s really interesting because they have so much depth across obviously silicon photonics and light. And then of course ethernet. The show for me was stolen by their announcement of their XPU wins. I mean, three of the four domestic large cloud providers are going to be building custom chips with Marvell. That’s pretty exciting.
They’ve talked a lot about that business. That’s starting to come to light. You can see also how much of the market and the TAM they expect over the next five years, substantial CAGR rates. They also talked about, obviously they gave some data points about the overall growth of that particular space. There’s some slowing growth in what I would kind of call traditional compute networking infrastructure and all the growth is going to AI. That’s where all the growth is going right now. And what works really well for Marvell is that’s where they’re set to play. I mean, they’re set up very, very well to play in the connectivity and that’s where their wins need to come.
Now, the market sort of responded in a somewhat unexpected way. I thought after they kind of announced the bigger TAM, the bigger opportunity, they would’ve gotten a bit more of a positive reaction from the market where a lot of the positive reaction shifted over and went to Broadcom, which was sort of interesting. But I guess they’re seen as the one-two with Marvell being kind of the two in this particular space. And so Marvell got a big jump. Sorry, Broadcom got a big jump because I think Marvell did a really good job of explaining how big the opportunity is for Broadcom. It also goes to show how, right now I think the market, the risk on off with sort of the, we’ll talk about a little bit more on the market and the overall economy next topic, but how that’s sort of creating, it’s pouring into what the market considers to be the safest bet.
But I actually think Marvell’s a really safe bet. I think the fact is it does trade at extremely high multiple. I think it’s got the highest forward multiple out of all of the chip stocks right now, I believe. Again, don’t quote me on this. And so I do think some people obviously think maybe it’s a little bit out in front of itself, but for good reason, it’s got compute, it’s got switch, it’s got interconnect, and it’s got big opportunities in both switch and compute to grow.
And that’s where it’s going to come from, Pat. I mean, I wish they could talk a little bit more about who they’re building these for, not because you and I can’t figure it out, Pat, because we can, but because we can’t tell and they can’t tell. And then ultimately people don’t know who these companies are for sure, and then they don’t necessarily get as much credit as I think they might deserve for the things that they’re helping do. But you know what? In the end,-
Patrick Moorhead: Who does that Dan? Who does that? Broadcom doesn’t do it and,-
Daniel Newman: No. No. I just mean, I think for a company that’s not as big and doesn’t have the density, that’s a good thing that they could share that. In some cases, everyone knows Qualcomm is the only reason an Apple phone works. You need their modem.
Patrick Moorhead: Well, you can also do, you can de-lit it and you can see the Qualcomm modem and the RF function on the board with, I mean to me it’s pretty simple. Octane’s on the board of Meta. Okay. And they do ship a ton of networking silicon into Meta, but, and it’s three out of four, then it’s AWS, it’s Microsoft, and it’s Google so mystery solved.
Daniel Newman: I’m just giving my personal feelings, which are very important, Pat.
Patrick Moorhead: They are important. Feelings are important,-
Daniel Newman: I am not a robot. I think this could help. Listen, I’m just saying that Marvell’s doing a lot of really important work for a lot of important companies and sometimes I don’t know that they’re getting credit, but when you do have this high forward multiple, maybe they are getting enough credit. And so fact is, is that the connectivity, we have an entire world of brand new data centers that are going to be built from the ground up for AI. Is not going to be, while we talk about using CPUs, this whole future XBU GPU architecture is going to create a whole new world of data centers. They’re all going to need to be connected. That’s a big opportunity for a company like Marvell. Let’s talk about the economy. What do you want to talk about next?
Patrick Moorhead: No, let’s go to the economy. One thing I wanted to point out though on Marvell is they do InfiniBand as well. They can benefit and monetize that. So they benefit both.
Daniel Newman: All the optics.
Patrick Moorhead: I wanted to throw that in there. And interestingly enough, they had a really interesting take on co-packaged optics with silicon photonics, which I need to double click into and compare that to what Broadcom said. But hey, let’s move on to something that is pretty controversial here, and that’s the US inflation data. Dan, you consider yourself somewhat of a junior economist. You like this stuff, so.
Daniel Newman: Look, everything’s inextricably linked. Okay. The economy,-
Patrick Moorhead: Okay. You got me there. I’m listening.
Daniel Newman: The economy and the performance, things like inflation drive the valuation and the innovation that comes out of tech companies, okay? When the dollar’s inflated, that means it gets you less. It takes you less of the way, and it means that goods cost more and it ends up trickling through the economy and it causes, I think it causes stagflation over time. It causes a slowdown of investment and growth. And here’s the thing is I think a lot of people just don’t understand what’s going on. This doesn’t take an economist, but I think taking a moment to just explain like, we have not actually tamed inflation at all since the peak of inflation. We’ve slowed inflation, meaning inflation has not gone down at all in any area. It’s gone up slower. So quarter after quarter, when you hear these numbers at 3.5, it’s still inflating after these huge inflation numbers, another 3% it’s going up, another 3.4, another…
And Pat this to me, and I think I shared this and I think you glommed onto it a little bit too, was that things that we actually spend money on and buy, most of us as the average citizen over three years have gone up an amazing amount, absorbent, like transportation is up 35% in the last three years. Oil, gas 31%, homes 31%, utilities 24%, gasoline 23%, food away from home, restaurants are up 21%. So I could do the whole list. Wages though, 15%. So it means that things are going up in cost at twice the rate in which people are getting paid. And so right now, people are hurting. I think like, I kind of said this, Pat, it’s been the quietest recession that nobody’s heard about, and it’s been backed by a couple of areas of strength. One, strength in jobs, which is all kind of BS right now.
We have tons and tons of jobs that are part-time and low paying that are being created, and we’re somehow continuing to float this really strong jobs number, which is total horse crap. And then the other thing about it is, is that we aren’t acknowledging the fact that with all these things going up by this much, that three and a half percent isn’t tamed, and you feel it when you go out. Pat, I go to the grocery store, my wife’s like, it was $450 for groceries this week. I’m like, what do we get? It’s like nothing. We got nothing. And this, by the way, is worse in other parts of the world. So we’re kind of delivering to the markets and the markets are reacting to this lower number, but in reality, the average household is feeling this. People aren’t being paid enough. And it’s all just this kind of house of cards that we’re built up on right now, Pat.
And in my take in the tech world, the AI moment has been the fuel of keeping markets alive. So this is probably the most interesting thing I’ll say here is while inflation is up, the one reason the markets continue to roar is that full self-drive or the future of AI and autonomy, you’ve got NVIDIA and you’ve got the AI boom there. I mean, how much has NVIDIA’s valuation brought the market up single-handedly? Microsoft, Google, Amazon, all up. Google just hit record or all time highs again. You’ve got a small number of companies that are cashing in on this AI moment. People are pouring money into these small number of things, Pat, and all the while it’s covering for the fact that jobs aren’t paying enough and everything we buy is too expensive, and we’re just all kind of like, eh, something going on? Nah, I’m just going to turn the ballgame on.
Patrick Moorhead: Well, I’ll tell you what, a big degree of the population is frustrated, shown by all of the data out there, but it doesn’t get covered a lot in the press, and it’s super disappointing. It’s kind of like a dangling chat, right, and something that doesn’t, that listen, I grew up in the 70s where we had a thing called stagflation, which was inflation in prices and no growth. That’s thankfully what we’re not here. But the fact that, think of somebody who makes an average of $50,000 a year, their homes where they live, apartments, houses, cars, food, gas, electricity has gone up 20 to 30%.
And I just don’t know how we sustain that. And we’re not even willing to have the real conversation about what we’re spending as a government. And everybody in DC I think is just failing us there, and we need to get that figured out. I mean, economically, there are no free lunches. There’s a train of school of thought that says, hey, if everybody is doing this globally, there really isn’t any impact. Right. I think the interest on our debt is essentially nearing in on GDP. Correct me if I’m wrong.
Daniel Newman: A trillion in deficit every three months, Pat. I mean, how the heck do we think we’re going to lower rates? And by the way, I want us to lower rates because I want my assets to explode in value, but how do you lower rates when we’re actually not slowing inflation?
Patrick Moorhead: I don’t know, buddy. I don’t know.
Daniel Newman: That’s why we’re not economists. We don’t have the answer. But modern monetary policy, can you imagine if we could run our businesses this way? Let’s just borrow $2 billion now of extra money and we’ll just run our companies. We’ll just run them to a deficit of billions,-
Patrick Moorhead: You know what? I don’t know if I’ll ever pay it back.
Daniel Newman: I’m never paying that back. By the way, when we run out of this 2 billion, we’ll borrow three.
Patrick Moorhead: No, that’s good. Yeah.
Daniel Newman: I don’t need to go to college for that, by the way, Pat. I can do that without any real formal education.
Patrick Moorhead: Maybe AI is the fix here. It’s the grand efficiency adder that’s going to get us out of this hole. And generative AI robotics is going to bring manufacturing back to the United States. By the way, you saw the how many robots China has versus the United States meme? It’s like 20 times the United States. We ended this show on such a,-
Daniel Newman: We build a bridge, right? We’re going to rebuild that bridge over a decade. They would’ve had that built over the weekend. That thing would’ve been done by now.
Patrick Moorhead: I know. I had of people come back and said, “Well, the quality’s not the same.” I’m like, you know what though? We can meet in the middle between 10 years and maybe a week, but hey, gosh, what came out of my mouth there? I appreciate everybody showing up, Dan, and thanks for moving us forward here. Have a great weekend. Let us know what you think. I hope this was a spicy show, informative, educational, and fun. Thanks for tuning in. Get ready for The Six Five Summit coming to a theater near you in June. Have a great one. Hit that S button.
Author Information
Daniel is the CEO of The Futurum Group. Living his life at the intersection of people and technology, Daniel works with the world’s largest technology brands exploring Digital Transformation and how it is influencing the enterprise.
From the leading edge of AI to global technology policy, Daniel makes the connections between business, people and tech that are required for companies to benefit most from their technology investments. Daniel is a top 5 globally ranked industry analyst and his ideas are regularly cited or shared in television appearances by CNBC, Bloomberg, Wall Street Journal and hundreds of other sites around the world.
A 7x Best-Selling Author including his most recent book “Human/Machine.” Daniel is also a Forbes and MarketWatch (Dow Jones) contributor.
An MBA and Former Graduate Adjunct Faculty, Daniel is an Austin Texas transplant after 40 years in Chicago. His speaking takes him around the world each year as he shares his vision of the role technology will play in our future.
