OpenAI's Equity Play, Anthropic's Access Reset, and the Billion-Dollar Race to Own AI Deployment

The handpicked topics for this week are:

1. OpenAI's Equity Offer to Washington: OpenAI is reportedly considering offering the federal government a 5-10% stake in exchange for a two-year runway on Department of Defense contracts. Pat and Daniel read it as a distribution and IPO hedge rather than a governance concession, especially with a trillion-dollar valuation still the target. (The Decode)
2. Anthropic's Phased Return After Export Controls: Fable 5 is back online, and Mythos remains partially restored to a limited set of vetted institutions following Commerce Department approval, a rollout Pat treats as the first real test case for how future frontier models clear review. Daniel adds that Alibaba's claimed block on Claude in China carries more optics than substance given how easily a VPN routes around it. (The Decode)
3. NVIDIA and Palantir's Open Source Enterprise Play: Alex Karp's viral CNBC interview argued that enterprises pairing Palantir with NVIDIA's Nemotron models can match frontier-level output without a frontier lab in the loop, a clip that pulled nearly 200,000 views once Pat posted it. Both hosts agree the priority for open source models has moved from leading capability benchmarks to closing that gap fast enough for slower-moving enterprises to adopt them. (The Decode)
4. Forward-Deployed Engineering Becomes a Billion-Dollar Line Item: AWS invested $1 billion, followed just two days later by Microsoft's $2.5 billion commitment to forward-deployed engineering teams, signaling a shift in where enterprise AI value is being created. Daniel argues the biggest opportunity now lies in implementation talent rather than model development, while Patrick sees the investments as a strategic hedge, noting that frontier AI labs have already been building their own forward-deployed engineering organizations to help customers put AI into production. (The Decode)
5. Meta's Cloud Ambitions Meet a Trust Problem: Meta Cloud can likely match neocloud-level GPU pricing and performance, but Pat argues the company's record outside advertising makes enterprise-grade services a much harder sell. He expects the offering to function as tactical overflow capacity for labs needing compute, with a shelf life tied to demand rather than a durable AWS rival. (The Decode)
6. The Flip: Has Enterprise AI Value Already Left the Model Layer? Daniel argues that Microsoft and AWS writing nine- and ten-figure checks for forward-deployed engineering in the same week, alongside an essay from Microsoft’s Satya Nadella declaring models replaceable, confirms that implementation talent now captures value model providers used to keep. Pat counters that today's models remain far from AGI, and that once frontier labs get there they will spin off cheap narrow models fast enough to keep pricing power on their side.
7. SpaceX Joins the Nasdaq 100 Within Weeks of Its IPO: SpaceX entered the index just 15 days after going public, the fastest addition on record, which Pat says will pull in passive buying regardless of the underlying valuation. Both hosts flag the speed itself as a signal of broader market froth.
8. SK Hynix and Samsung Signal Memory's Return to the Center of the AI Trade: SK Hynix is pursuing a $28 billion US listing that could value the company near a trillion dollars, timed just ahead of Samsung earnings both hosts expect to show a sharp profit jump. Daniel ties the moves directly to NVIDIA's memory demand, and Pat notes most investors had barely heard of SK Hynix before this year.
9. NVIDIA and Microsoft Split Paths in the Magnificent Seven: NVIDIA gained 7% for the half while Microsoft shed 23%, a divergence Pat ties to Microsoft's exposure to OpenAI's fading momentum and its perception as a bundled SaaS play rather than a leading infrastructure provider. Daniel calls the selloff overdone and floats it as a generational buying opportunity.
10. NVIDIA's Roadmap Rebuttal Tests the Limits of Corporate Denial: NVIDIA issued a statement calling its roadmap intact after a SemiAnalysis report raised delay concerns tied to its Kyber rack architecture, and Pat reads the denial as legally meaningful given the liability regulated companies take on when refuting analyst claims. Both hosts land on the same read, that first customer shipment may hold while full volume ramp slips.
11. Palo Alto Networks and CrowdStrike Undercut the SaaSpocalypse Narrative: CrowdStrike climbed 95% and Palo Alto Networks gained 113% from April to June, each posting record quarters that Daniel says quiet the theory that frontier models would simply replace dedicated cyber tools. Pat adds that security was always the wrong category for that narrative, since it ranks among AI's biggest risks rather than one of its casual replacement targets.

New episodes of the Six Five Pod land every week. Watch the full video at sixfivemedia.com, and be sure to subscribe to our YouTube channel so you never miss an episode.

Disclaimer: Six Five Media 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.

Daniel Newman:
Yo, yo, yo, yo. What is going on, everybody? We are back. Another week in the books. What is this, 311? Remember 311, the band?

Patrick Moorhead: 

Yeah, I still listen to at least one of their tunes.

Daniel Newman: 

Down Down? We can only be down down, right? Exactly, Dan.

Patrick Moorhead: 

Daniel, your breadth of skills amazes me. What we haven't talked about is that you used to be Jam Master Dan.

Daniel Newman: 

I may or may not have produced a few electronic albums, LPs that were spun on records at raves all over the world in my early days. Infamous. Anyways, who cares? That was so long ago. I was looking back at something like, I was thinking, man, it's been like longer that I've been out of college than I've been in, than I was in when I was in it. I'm old now. I am an old man. I'm not old like you, but I'm getting there. So good, you know, we're a little late on this one. But it was a holiday. I was out on the boat, setting fireworks off that I was holding in my bare hands. Well, you know, lifting, I brought a barbell on the boat with 245. You see that video of me? I was pumping iron on the boat.

Patrick Moorhead: 

I did a fireworks are going off. I mean, I was worried for you the first time you didn't get your didn't get all the way up. And I thought you blew out your shoulder or something.

Daniel Newman: 

So yeah, no, no, I just, uh, it was just an awkward, you know, I just, I just picked it up awkwardly because I actually, I put like three 15 on the bar because I was going light. So I didn't want to go heavy, but I still somehow made it awkward. I was going to one hand it like Anatoly. You've seen that guy and, and, and a tall or whatever his name is, the janitor guy that goes around the gyms and. Yeah. Here's with people like he has like the 40 kilogram mop that he walks around with.

Patrick Moorhead: 

Yeah.

Daniel Newman: 

And then he like one arm deadlifts 500 pounds and makes all these big dudes. Cool. So, so I'm, I'm, I'm one of those big guys that can't really lift that much weight. Um, so busy week though. I mean, you know, there is no off weeks. It was a weird one, like not doing it going into the 4th of July because, you know, God, I hope you weren't listening to podcasts and you were actually like out doing something fun, blowing fingers off, celebrating America. I love this country. Um, watching the world cup. Watching USA tonight, in case you want to know what day it is, it's the sixth. Anyways, how was your fourth? You really didn't talk about it. I know you asked me to work out today, but I was too busy doing legs.

Patrick Moorhead: 

Yeah, weekend was great. I spent it at the lake house with the fam, including my grandson. It was really good. And yeah, I really didn't pay too much attention. I mean, I'm just, listen, I'm always on X. So there wasn't a lot going on X. And yeah, so it was a it was a good relaxing fourth to hang out. I was the I was basically the the bus driver of the boat. So yeah. Yeah.

Daniel Newman: 

Now I don't have any fireworks or not.

Patrick Moorhead: 

Well, I did bring out the guns. I mean, it was the you know, big reveal, right? Took the shirt off flexed a few times did get a few comments, which were they good? Good. Yeah. Which is always good to kind of keep it keep it going. Oh, hey, what's up?

Daniel Newman: 

My son keeps me humble. Comes up to me, walks up, he grabs my skin around my belly, and he just calls me a fat ass all the time. He doesn't say ass, though. He just says you're a chunk or a chub. You need to get yourself one of those. That keeps you humble. That's not nice. It keeps you humble. So anyway, we got a great show. Tons going on. We got, you know, for those of you, maybe first time listeners, I'd ask where you've been, but, uh, you know, we got the decode where we can break down the big topics, themes and noise of the week. By the way, did we, do we, you know, some of this is not last week. We might have to. We're going to talk at all about this Kyber thing, because they're not on our list. We'll talk about it later. Then we're going to go into the flip, where I win a simulated argument about something related to AI. And then we'll do some market madness. What do we call it, bulls and bears? We call it different things each time. But bottom line here is it's where we do our best to manipulate the market so we can buy into our ETF at a better price.

Patrick Moorhead: 

I've heard rumors of people doing that, Daniel.

Daniel Newman: 

Yeah, I don't know. All I'm going to say is my roadmap is on schedule, Pat. It's intact.

Patrick Moorhead: 

Maybe it's not aggressive enough.

Daniel Newman: 

For everybody that doesn't get the inside joke, you're going to have to look it up. I'm not going to give it away. But what we do need to do is we do need to get to the D code. So let's get after it. Oh, all right, Pat. We've got the American AI champions doctrine. Open AI is going to give maybe 5% to the government. And what do you think, man? I'm gonna just start with that one. There's some other stuff here, but what do you think? Is that the case? Are they gonna hand over 5%, 50 billion?

Patrick Moorhead: 

I mean, listen, if you wanna lock in kind of DOD, government projects, at least for the next two years, it wouldn't be the dumbest idea that you could have done. And we've seen, it's funny, I have not seen, I have not heard of any outward pressure tactics with, let's say, Intel, as an example, that vow must use them not even rumors, right? Like, you would think by this point, somebody there would be a New York Times article or the information. article that would say that. So I, listen, as long as it has no true ability, no board seat, it's not the worst idea out there that I've seen. And, you know, I would hope that it wouldn't dictate roadmaps or things like that, but I I just don't necessarily love the idea, but I don't necessarily hate the idea. It is coming at a time where GPT 5.6 is being gated out there. ChatGPT or OpenAI, you can say, hey, they're just savvier than Anthropic, or they're just later than Anthropic. out there, but it is out there at 20 pre-approved companies under a White House security rule. I don't know if these two are tied together, but, and who knows? I mean, the value of open AI, I think, I don't think Sam wants to go public until he can find a trillion dollar valuation. So, you know, we might be sitting there for a while. I think you'll be closer if 5.6 is all that.

Daniel Newman: 

Yeah, I think it's the move to make to ensure your distribution and your IPO. We all know that President Trump and his support meaningfully moves the market. Pritchard Dell was up 5% today on another nice comment from President Trump about Dell's participation in Invest America. I'm pretty sure that you go back to Micron, you go back to so many of these companies, Intel, that he's had a hand in. So we know there's a tension between Anthropic and the US government. It's, you know, Dario asked to be regulated. He challenged the use for what he thought was unethical reasons. And it seems like at each opportunity, Sam and the OpenAI team has been more than willing to raise their hand and say, we'll play ball. We'll do whatever you want. And so this is another case of, is $50 billion of fake money in the stock to get that guaranteed underwriting, guaranteed placement, assuming your product is competitive? I think it's probably a worthy path if you want to ensure all your current investors are going to get out. I mean, he basically underwrites the continued capital raises and supports the entire AI build out. So, I mean, of course, you know, according to Michael Burry, AI is over. It ended this week. Although we're having another face ripping rally as we speak, but never mind. You know, he just says it every day forever. He'll be right eventually, right?

Patrick Moorhead: 

No, that is true. I mean, statistically speaking, I mean, yeah, he's going to be right. It's going to go down. And I think he's really only looking for like a 10 point 10 point swing, right? Oh, sorry. 30 point swing. That's kind of his. Of course it's going to go down 10% at some point. 30% would be vicious, but anyways.

Daniel Newman: 

Yeah, I think these people keep forgetting that there's so much stimulus and there's so much willingness to move policy quickly to avoid any sort of real market meltdowns. A lot of people have said like, well, if these five things didn't happen, we would have had a big sell off. Yeah, but they did happen. And the government did all of the things it needed to do to create the incentive for the market to not capitulate. So I'm not saying it'll never happen. I'm just saying like, People like to look at history and the information history. Even the information symmetry that existed even in 2008 is different than today. The speed of which a news story impacts the price of a stock or can impact the overall sentiment of the market, it's in hours now, minutes, minutes to hours. And it's just as fast. that a change of course, a simple comment or a tweet from the president or from the right politician, or like I said, a refutal from the right CEO of a piece of news can completely turn the perception of the market.

Patrick Moorhead: 

Yeah, a lot of states getting involved here. I mean, you've got, you know, Japan, sorry, Korea is looking at a $625 billion investment, you got MGX, 49 billion, sovereign capital investment. And then supposedly, this voluntary AI standards framework is is supposed to come out. as well, which is basically, which models can you use and by whom and what proof do you need and becomes a documented federal process. So the hard part about this for like CIOs and investors is like, like if you've got a cap table that includes labs and governments and things are gated by security access, you really need to think about who kind of who you're aligned with. I think all of this kind of throws this into open source. But then again, some people are pontificating that open source will be will be shortly regulated as well. I mean, you know, Dario is very much under, you know, wants open source regulated.

Daniel Newman: 

Yeah, well, at this point, open source, we'll talk about that a little bit more later, maybe one of the biggest, you know, arguably one of the biggest threats. Oh, anyway, um, let's see, Anthropic, they had a pretty good week, you know, all that they haven't yet announced. Um, They haven't yet announced whether or not they are, sorry, just going to give 5% or 10% to the government, Pat. They haven't made a commitment yet. But they did get the commerce approval to lift Fable 5. And is it partial on Menethos? Menethos is partially back. That's only been rolled out to still a number of vetted institutions. So, you know, I got a lot I don't know about you, but I got a lot of. Inquiry from press about this particular topic and my read on this 1 was. We have to put a process in place for these new models. Like, there's a lot of people thinking government intervention meant like. this is going to be a forever blocked up, messed up process. And every new model was going to be awful. Like the fable one, where it's going to get, I think we're just figuring out the right approval process and what's going to end up happening. I don't know what you think, but going forward is that nothing new will ship until it kind of gets the green light. Meaning that, you know, what happened with fable will not likely happen again because it won't ever be released without getting the, the kind of seal of approval. Is that, is that how you see it? Or do you think it's going to be different?

Patrick Moorhead: 

I, um, I don't know what to think anymore, to be honest. I think that, um, I think we'll get back on. I think, I think we will, it'll be turned off and reel back in and we're probably going to get some sort of approval in one way, shape, or form based on a smaller partner track, which I think will all come through the major hyperscalers. Because at least up to this point, the major hyperscalers have been the only one, I think, to really demonstrate who uses their stuff, the observability around it the ability to to shut it off as opposed to even a Even a neocloud that again, there's there's nothing wrong with their technology Their technology is is is good from the standpoint that has low low overhead But I think that we're gonna just you know have a Yeah, I think it's just going to be a bunch of fits and starts until we get to the right place. This is going to be like, I think, trying to keep people from using NVIDIA GPUs, right? And you can go back and fact check us, audience, but we were always kind of skeptical about how difficult it would be to do that. So I think best efforts is what we're going to see.

Daniel Newman: 

Yeah, I agree with that, but I think there's going to be more of a process. I think it's going to get more defined. I don't think it's going to be as loosey-goosey dropping these kinds of things and challenging. I think there's going to be a pretty well-defined process. And then, by the way, on the other side of this, though, I thought was pretty interesting was Alibaba in China. ban Claude Koch. So in the US, everybody screams and cries about the US is stopping the Chinese progress. We should be sending our best ASML machines. We should be shipping all of our best GPUs over there. And why not let them steal and distill all of our models and then sell them back to us for a fraction of the price through their labs, right? You know, I mean, Alibaba blocking Claude, I guess I don't really know if I care that much, but good. I mean, you know, they block Facebook and Meta and they block Google and they block, you know, a whole bunch of other things over there, don't they? What's any different about this?

Patrick Moorhead: 

First of all, I think it's, I think it's a bunch of BS. Um, I think, uh, Alibaba saying that it's blocking is, is, is just a bunch of show showmanship. Okay. Uh, doesn't mean that somebody can't get a VPN, uh, use a credit card and, and, and do it. I mean, none of it, uh, stops anything. So I'm not buying into, uh, I'm not buying, I'm not, I'm not. I'm not going down to that narrative rabbit hole. For what it's worth, I mean, shit, I couldn't even get Fable to give me a reading back on my own personal blood test, right? And it was funny, it bounced me down to, you know, a lower level after that, right, 4.8. And then it said, nope, nope, we can't even do it on 4.8 anymore, let's bump you down to fricking Sonnet 6. where it kind of gave me an iffy response. So I don't know if that's just, you know, what everybody should expect, but I thought it was absolutely absurd. And I looked at the terms and conditions. Basically, you're just not allowed to use Fable to build some biological weapon. I mean, I guess my blood scores are either so good or so bad, it could be a biological weapon. But still, if this is the quality of of there is somebody doing something nefarious. It sucks.

Daniel Newman: 

Yeah, well, TBD, I've not had the problems with Fable so far. It's actually never denied me anything so far. So maybe I'm just not challenging it enough. You know, maybe I need to I need to ask it about my, you know, my what is it, my inflammation. What's the ApoB? I need to ask about Apo. I don't know. I need to ask it something harder. I just keep asking it to tell me what my body fat is.

Patrick Moorhead: 

Can't you take pictures of it?

Daniel Newman: 

Oh, yeah. And it just comes back and it's like, you look great. What are you doing? I'm just kidding. Anyway. By the way, models are very complimentary, naturally. They can fill a lot of gaps emotionally in your life if you need to. Like Pat, this article you wrote was so good. It's like, yes, Fable, you wrote it.

Patrick Moorhead: 

I just, listen, the more I abuse my agents, the better they do. And I think this has something to do with, there's something there.

Daniel Newman: 

Maybe. So let's hit open source. We've hit OpenAI. We hit Anthropic. So NVIDIA Palantir made an announcement. Nemotron, I mean, that's been kind of off the radar. You know, I kept hearing about GLM 5.2, and I keep hearing about Quen. All I hear is that US has no good open source and now NVIDIA and Palantir and Alex Karp is going on TV and telling people that they're giving all their alpha away and you just need a Palantir with an NVIDIA Nemotron and you don't even need a frontier model. I mean, what's up with that?

Patrick Moorhead: 

Yeah, I posted Alex Karp's video on CNBC. Some people called it unhinged. By the way, I got almost 200,000 views of it, which I was pretty happy about. I really hit a nerve basically saying all CEOs should look at this. And yeah, I mean, NVIDIA is making a credible run. at good or open source models. I've always said that the open source models don't need to be the best. They don't need to be if they're really good or they're the best six months later. I mean, look at how slowly the enterprise moves. They do not run very, very quickly. I do acknowledge though, when you have a hot model that comes out, it might be able to do something that you've never had to be able to do. It's hard to get off of that, even though if you've got an independent harness or a structure to rip and replace different models. I think though, what, it's interesting, The whole trope about don't give your information to frontier models because they will basically steal your IP. I always kind of like push that to the side because of you can turn off data sharing, right? But once Fable 5 hit and you had to share data either based on you know, security concerns or something like that, all bets were off. And we saw everybody in the industry who was relevant, basically tell their people, do not use Fable 5, including Microsoft, who's a partner of, a distribution partner, and an IaaS partner of Anthropic. So essentially, it is bringing up this whole concept of sovereign AI. which Satya had been making this point for a while. All of the on-prem vendors, whether it be Dell, HPE, Lenovo, and Cisco, they have brought this out. I can finally say with my confidence that yes, if you use a frontier model, they will have the chance to steal your IP and build products that are potentially better than yours. So therefore, you have to have some sort of an air-gapped type of solution. 100% for defense and regulated industries. I think, my gosh, you know, we saw even Anthropic talk about how it's going to start designing its own drugs. And then I look at who the drug companies that Anthropic has done some deals with and reportedly lower priced deals. than one might. And so I look at also this whole idea of token maxing, token efficiency. I heard a really smart analyst cited that at one point, I think it was even before Dell Tech World. His name may have been Daniel Newman, but Anyways, so it also plays into it this debate. It's funny. It's amazing. It's a debate that running smaller, more focused models on-prem can actually save you money. So I do think there's something there. Even the Signal 65 testing that we announced at Dell Tech World said you could save 84% 84% cost savings by running the same tokens versus Anthropic CLI. There's a lot of wind in these sales, baby, for enterprises and CIOs, and even the risk committees of public companies need to fully comprehend this at this point. It could be an actual just a boon for on-prem infrastructure providers and may force the hyperscalers to go even deeper into on-prem infrastructure.

Daniel Newman: 

Yeah, I think it's interesting. I mean, first of all, great interview if you didn't watch it. Some people thought he was unhinged. I just thought he was great. I really appreciated the framing. I think he is the forward deployed model belongs to Palantir. Five years ago, they were picked on for being a consulting company. And today, the sexiest thing you can do is be a consulting company with a thin software wrapper wrapped around you. And the fact is, ontology is the application of a unique skill set to data to create really high value from AI. And that's what every company is trying to replicate that now. But moreover, you know, this is one of those interesting inflections of when you need the leading intelligence, which you would argue the frontier labs has the most advanced highest capability intelligence. But these open source models, they're not orders of magnitude behind their single digit percentage points behind in many categories. And We need, first of all, an American champion. Second of all, what Karp said is basically, if you apply the alpha of your business entity organization, the special data that you have, this is something you and I have said endlessly, that you have that is unique to one of these models. And he basically said, if you do it with Palantir, I can give you the same output quality that you would get from the most advanced model in the world. And to your point, Pat, we do not want to necessarily share all of the alpha that exists inside of every business. I think J. Cal said this from LN, but he said that basically these labs are just going to steal all the best ideas. because people are building and just freely sharing. And, you know, I'm guilty sometimes of doing that. Like sometimes I do incognito, sometimes I do it in COVID, sometimes I'm just searching stuff and I'll drop a document in, right? I'm sure you've done it too, where you don't need, it's got a revenue spreadsheet or something like that. What I'm saying is that you probably wouldn't share, but you know what? It's become very easy and comfortable to just share proprietary information with these things and have no idea how that information then goes on to be used. And so I think when you're talking about a sovereign nation, a state and local government, you're talking about a health care company where your trade secrets are your patent portfolios and your APIs and all the things that you're working to build, just freely sharing that with these models with no rules and rails of business, governance,

Patrick Moorhead: 

You know, I mean, the only thing worse than that, you would have thought we would have been the governance guys. But think of it this way.

Daniel Newman: 

I mean, we, I sure as hell criticize China a lot for stealing IP. I mean, the entire, the entire frontier model lab business is all stolen. I mean, I don't, I don't mean to like say this cynically, but like it literally, they stole the internet and built a trillion dollar product on it. Yeah. At least Google, in its first wave, organized the data and then sent people to your stuff so you could get credit. Now it's just distilled inside of a block of text where, at best, you get a citation that no one will ever click. So it's an interesting dichotomy of what's happened here. So we've got to move on. We're dragging here. We've got nothing else to do. All right, Pat, speaking of the FDEs, so AWS spends a billion. Microsoft stands up a $2.5 billion business. Now there's $3.5 billion in fresh spend to potentially, is this going after the Accenture's and Capgemini's and Deloitte's, IBM's? Is this something new? I mean, the way I read an FDE and this whole model is, You know, instead of needing 20 consultants, it's going to be one really highly competent and capable engineer that deeply understands these businesses and understand the application of a technology or tool to get maximum value. It might not be 20 to 1. It might be. You know, 5 to 1, but the idea is the FTE model broke the traditional consulting model that was extremely high labor utilization driven and it was much more about, you know, a low labor utilization coupled with technology to drive exponential outcomes. And so. My big question about these efforts, because I think Palantir has proven the value. My question is, how much marketing is this versus reality? Meaning, can a Microsoft or a Google or an AWS do this better than an AWS, or sorry, better than Accenture, better than a Deloitte? And is this ultimately the model those companies are going to have to go to in order to we turn to desirability within the market. But I think, Pat, this is a trend line that every software company, most infrastructure companies, they're all going to go down a path like this. It's a good story. I liked it so much, I told one of my own.

Patrick Moorhead: 

So it's interesting. I thought this was interesting, not based on the threat to Accenture, Deloitte, Capgemini, and IBM Consulting, but more around how this tied to what Alex Karp and Satya Nadella had aired out, which was, you need an intermediary for this. I wrote a Forbes article about, can CIOs trust Anthropic? And I'll let you guys read it and look at it in the show notes, but essentially I said, hey, as long as you use a buffer for these reasons, yes, you're gonna pay for the buffer, But you cannot just hook your wagon to an anthropic and I wasn't picking on anthropic I think the same for open AI, you need to be able to have something with. the tooling to be able to move models in and out, use open source models. And I think in the long run, it will actually save you money, but more importantly, protect your IP, because that was really what Satya was leaning in on this as well. And I got a lot of questions for the press, which is, well, why couldn't they do this before? The thing about it is they were. When you have a company as large as Microsoft and you want to speed up, you have to have a focus. And typically, an organizational construct like this is being created. Look at Satya. He's like the chief product manager of AI. And then he put Altaf, sorry, what's his name? His first name? Chief- Justin. Justin, yeah. Judson. Judson, sorry. Sorry, Microsoft. Put Judson as CEO of the commercial business. And this is exactly what it is to increase decision time. sorry, decrease decision time, decrease execution time, and likely increase the quality of the engagement. So, yeah, that's what I find interesting. And, you know, it's funny, I've done a couple of CIO roundtables in the last year, Daniel, and I did get the feedback that said, hey, we're going to try to dial down the management consultants and try to have more direct relationships with the vendors. And I thought this was probably over a year ago that I had a bunch of these conversations in London. And it just seems like that's exactly what is happening here. A lot of customers just want to go directly to the source. I will note in the blog that the SIs were cited here.

Daniel Newman: 

Yeah, I think your point is that there's a little bit of a marketing story that sits underneath a transformational story. Because to your point, these kinds of roles have existed for a long time. I do think a net new focus on it, but I also think the way it's articulated and then how we are going to get to the point at some point where these companies are going to have to prove ROI on all this spend. It's not yet, but it's going to happen. The receipts are going to be required for people at some point. These kinds of roles can go in and help more immediately generate value from these investments that are being made, also creating a lot of good case studies in the process. All right, last topic on the decode, Pat. Let's talk about meta, meta the cloud co, meta cloud. Is Meta going to be the next AWS?

Patrick Moorhead: 

Listen, when you fail at your consumer journey, which I would say that they had an early start, they did well, they had Lama for open models, the future was bright, and then the wheels just fell off. Lama ended up being a who cares, they money whipped a ton of engineering talent that they brought in and literally made them a billion dollars worth of commitments to folks to come in and we've seen nothing yet. And when you're sitting on a ton of GPUs, you have to take them somewhere else. And I think people looked at XAI that was in a similar position and said, we can make a boatload of money off of these things and increase our capitalization. So There is no doubt in my mind that they can provide what a Neo cloud provides, which is high performance, low latencies, GPU as a service. Can they build the services that you would expect if you were an enterprise? Uh, not a chance. They failed, uh, at least 25 times and everything B to B that they've tried to do outside of, of advertising. But I do think that if you're an anthropic or an open AI or, or somebody else and, and need GPUs. Yeah. So this would be a tactical thing. It'll get stopped. Um, uh, when it becomes a not as, um, interesting for them, it's not a long-term commitment.

Daniel Newman: 

All right, so I have a very simple thesis. He followed into this watching SpaceX. Yeah, basically said buy all the compute you need for my pet project and whatever I can't actually deploy, I will sell at a higher than neocloud market price and I will make There's like dollars per share of value on this cloud compute being sold. So I think he goes and buys cheaper cloud from Neoclouds. We're going to be forced into this competitive frenzy of selling down their cost per megawatt, gigawatt. And then he'll take his highest caliber, high capacity, high trust name, and he'll sell his compute on at a premium, and he'll arbitrage the whole thing. My rate. We'll see what happens. But I think it's smart. I think it's smart because it's a hedge. People are saying he's overspending. Now you can say, well, he's got a plan B. He's got different ways to monetize this investment. If he overbuilds, he'll sell it. If he strikes gold on a product or service using the technology, he has the compute at his disposal. Whereas that's going to continue to be a really scarce resource. So that's my read, path, good decode. Um, you know, pretty good pace, not as fast as I can normally go, but, uh, you know, it's, it's been a really tired after a long weekend of not doing anything. What can I say?

Patrick Moorhead: 

Um, I don't think you didn't do anything. Just like I didn't do anything.

Daniel Newman:  

Let's, uh, let's, let's go.

Patrick Moorhead: 

Yeah, no, I crushed it. I crushed it today. Six 35 for eight on the, on the belt squat.

Daniel Newman: 

I was like, you know, I'm a third of a ton. I'm doing a third of a ton now. All right. So let's let's go. Let's have a little debate, Pat. We're going to, you know, choose a take prediction inside on the latest news. Maybe what topic we've already talked about. We'll flip a coin and see who's for or against this take. Today, we're going to hit enterprise AI value has shifted from model providers to implementation partners. The buyer's next big check is who built it, implemented it, not who trained it. Well, is it me? I'm four? I got to do this?

Patrick Moorhead: 

Yeah, jump into it.

Daniel Newman: 

All right. All right, listen, the model war is over and everybody lost. Prices per token are collapsing. Capability gaps close in months. The four biggest names in AI just told you where the money went in one week. Amazon committed a billion dollars to AI, to forward deployed engineering. And Microsoft, two days later, announced $2.5 billion. Not to mention, Alex Karp came on TV and said, every enterprise in the world hates the Frontier Labs, and they only want to use open-source models with Palantir. That's not exactly what he said, but if I had to take his entire 15-minute interview and put it into a sentence, that was it. And listen, when Microsoft and AWS both write 9- and 10-figure checks, do the same thing in the same week, not a trend. That's a confession. And the confession is that model hours don't retain customers. The team standing inside the building when AI finally works does. So let's talk about where this comes from. You know, it starts in Nadella's essay back in June itself. It's 60 million eyes on this thing and foundational models are becoming interchangeable and it's happening fast enough that no provider can count on the model alone as durable revenue. Basically, the largest distributor of models on Earth, Microsoft, just said, model isn't the mo. You know who said that before him? I did. And so, you know, MIT's Project Nanda found 95% of enterprise Gen AI pilots deliver zero measurable P&L. That is a hard ROI number. Now that number got played out, but it wasn't about AI. It was about the fact that the people required to successfully deploy AI, the people between the demo and the dollars are forward deployed engineers. So it's gonna be the people standing on the other side. That's where trillions of dollars in enterprise value gets captured. models, thing of the past. And by the way, the labs get this. I mean, heck, Anthropic and OpenAI themselves stood up FDE groups. So I mean, when you're the model company yourself, and you're putting these in place, and you're partnering with private equity firms, and you're working with banks, and you're driving out and engaging consultants, it's because they know that their models without the right technical help, standing next to the people deploying them means not enough value. So it's very simple. You know, you follow the services math, Microsoft Enterprise and Partners did what? $2.1 billion in the March quarter. You know, they bet big on Frontier rather than on licenses because they need to re-accelerate that line. It's very, very simple, Pat. It doesn't matter if it's Palantir's early model. It doesn't matter if it's AWS and Microsoft following suit. It doesn't matter, you know, if it's Satya's confession that there is no moat in the models. We know for sure that the future success, all the margin, all the upside, all of the alpha that a business has is going to be caught when they can take AI, they can take their proprietary data, and they can partner with the technical, forward-looking, capable people, just like a forward deployed analyst from Futurum, to get maximum value from their business. Models are a commodity.

Patrick Moorhead: 

Nice plug, Daniel. So, listen, I don't want to live in this fantasy world that says, like we're just gonna reach AGI and all innovation in models is gonna stop. So I think we saw very clearly that in the latest Fable 5 models, and even some people with the latest chat GPT models, we are very, very far away from AGI. Now, once we do get to AGI, These model makers are going to start making very narrow solution models that are going to be incredibly cheap because they've got the scale, they're doing direct deals with Neo clouds, even chip makers, and will be driving their costs down. I mean, look at the revenue crawl chart of OpenAI. In 2022, there were $28 million. Right now, they're 10x at $28 or $30 billion. In 2022, Anthropic was $5 million in revenue. And now they're also looking at $30 billion. And if you look at 2029, a mere three years away, you're looking at $150 billion in very high margin revenue for these companies. So a lot of people aren't believing it. The valuation of these companies are nearing a trillion dollars because they understand what's going to happen situationally with these companies. And, you know, what do you think is going to happen when somebody kind of belly aches and says, hey, I you know, I want to buy into Anthropic because they've got the best small, you know, let's say we're in 29, uh, we've had AGI. We they've got the best small models for manufacturing for sales. Uh, and if they're at a, at appropriate price, making a ton of money, uh, it sure looks like that, that they would throw in an open model. If you, if you really wanted it, Heck, even Dell is doing an on-prem deal with OpenAI. And there were rumors that it would even be anthropic. So I can see, like if you read teamness and look at what could happen with a threat if this whole sovereign thing takes off, I think there's a lot of different chess moves that Anthropic can have before we get to the point that they don't, you know, there's nothing to keep them from doing what the hyperscalers are doing. There's no massive moat out there. Or rest my case.

Daniel Newman: 

Yeah. I just wanted you to awkwardly sit there for a while. And, uh, sulk in your defeat.

Patrick Moorhead: 

I mean, I'm fine with awkward.

Daniel Newman: 

I'm giving you a red card. Ballora and Bogalt ball again. You're getting a red card. You're out for the rest of the game. You're just locked. You're Mexico, I'm England. I'm playing down a man mentally, but I still won the game. I'm just kidding. Look, I think that this is truly a case of lying somewhere in between. I do think that the idea, I actually think like the SaaSpocalypse, a lot of that stuff's faded because I think people have come to the realization that it doesn't freaking work. You can't home grow a massive ERP system to run your, you know, billion dollar company without a ton of governance, a ton of protection of data, a ton of security, some real engineering talent. Not to say never, but I'm just saying we're not there. It's not there yet. Things like that I think people have come to realize, which has changed the discussion a little bit. And I do think companies that can apply talent that really understand the practical application of AI are going to move a hell of a lot faster. And I don't know that that's traditional consulting firms. You know, I don't know. I think this is what's a little different in this way is not to say they can't amp up and get those right people, but just bodies thrown at this doesn't fit this. The scale is different when they are not just more bodies. you have a very powerful technology that needs the right capable person. You know, it's just like with you and your son, right? Like how much can he do by himself versus you to try to solve the same problem? Because the understanding and all the pieces that go together, you know, one person can be exponential. Not to say you're not really great at building apps. I've heard you've built a hundred and they each have one user. Come on, that was funny, laugh. That was funny. built a hundred. They all have one year's payment. All right. It was really funny in my head, though. I thought you'd at least chuckle about it. Come on, man. That breaks me. I don't even know if I can host the show again now. Hurt me over here. Actually, what's happening? Pat's answering inquiries right now. He's he's he's double double tasking on me. OK. All right. Well, let's wrap this up, man. It wasn't a crazy week in the markets, but there was some news for bulls and bears. So let's let's let's give the market what the market wants. Some patent and buy high, sell low. All right, man. So. First, what less than a couple of weeks since its IPO and SpaceX is now part of the NASDAQ 100. I don't know if you have a lot of comments on this, but mine is I guess we're going to change all the rules.

Patrick Moorhead: 

Yeah, I mean, this was the fastest IPO to… NASDAQ 100 ever 15 days, right? And you know, all the index stuff hits there. And which is just crazy, which which will give it a, a kick. I still can't believe that this valuation though, the underlying valuation, I mean, it's better than crypto. But yeah, when you got Elon, but this is this, these are the things that give me pause on we are in a stock market bubble here. stuff like this. I'll just I'll leave it at that.

Daniel Newman: 

That's a bubble there.

Patrick Moorhead: 

Definitely not about there.

Daniel Newman: 

I'm just kidding. I'm just kidding.

Patrick Moorhead: 

I like that.

Daniel Newman: 

I got Ben to say bubble there.

Patrick Moorhead: 

No, I saw that he didn't even put the trademark. Mark, I told him it was okay.

Daniel Newman: 

Yeah. Now he's, now he's got buried between them. He's, he's officially in the, he's in the, he's in the club now. Oh yeah. I don't want to say about this one. You know, their first change, first break all the rules. I mean, they need this to succeed. They need this, this, uh, to become a huge part of the index. It creates a lot of passive buying, um, you know, Again, I'm torn on this one. Five years from now, I want to own it. I just want to buy it at like 100 bucks a share. So make it happen. Make it happen, market. So on Friday, Pat, the Six Five Mayor may not be in New York this week to do a little chitter chatter about a deal that might be happening this week. I can't say if there's anything to do with this. But let's just say SK Heinex is doing a $28 billion ADR. So they're going to basically list here in the US. I believe this puts them at like a trillion dollar valuation. This is a company that's, you know, alongside Micron, going to be very interesting, being made available in US markets. It's going to show just how much bullishness there is. But I mean, SK is the preferred partner of NVIDIA. They've been deeply partnering with them. They're spending big money, right? Was it the number you threw out earlier, like 600 billion? That's it, yeah. Yeah, like 600 billion. And, you know, we're gonna get some foreshadowing this week, because not only is SK coming in and doing this big ADR announcement and listing here in the US, we're also gonna get Samsung's results out tomorrow. It looks like they're gonna crush it on Pat. I mean, look, memory is back. Samsung is back. How can you be calling a bubble, dude?

Patrick Moorhead: 

Yeah, I bubbles are good. Bubbles drive investment, right? And the only reason we call it a bubble is it gets really, really big. And some bubbles, they don't pop, right? I think bubbles are a good thing. And I think we're seeing that. We would not have this crazy investment like we're having right now, which 100% will pay off. But no, it's good to see. It's so funny. I doubt most people even knew who SK hynix was, but now they absolutely do a trillion dollar, you know, a trillion dollar out or trillion dollar ADR. I think the listing was was twenty eight, twenty eight B's. Right. And then combine that with Samsung's literally 18 X profit jump. Um, and I think this thing hits tomorrow, but it's exciting. It'll give us at a minimum, something else to, uh, ruminate on and stew on and debate. I'm waiting for the, uh, you know, the people to say, Hey, this money needs to be shifting into NVIDIA. Not, um, not memory.

Daniel Newman: 

Important question is, are you buying?

Patrick Moorhead: 

I only buy when it's at the highest.

Daniel Newman: 

You're gonna let this come out, you're gonna let it bubble up and you're gonna wait, you know that you're gonna use the Kathy Wood trigger when she buys you buy?

Patrick Moorhead: 

No, I mean, I pretty much got my own personal index tied to hers.

Daniel Newman: 

So yeah, it's called the top.

Patrick Moorhead: 

Yes.

Daniel Newman: 

But don't worry, you'll panic sell the bottom.

Patrick Moorhead: 

No, exactly. I mean, it's a tax strategy, Daniel. I make so much cash.

Daniel Newman: 

By the way, you know she said that, right? She talked about all the tax savings she has from the giant losses she's taken.

Patrick Moorhead: 

No, I didn't know that. That's pretty sad.

Daniel Newman: 

It was actually, that happened. I'm not kidding.

Patrick Moorhead: 

It was more of a joke.

Daniel Newman: 

Yeah. Well, that felt like a joke. It was an interesting one. It was pretty funny. It also just shows good ideas and timing matter both, right? It's, it's both. So, you know, quickly on this one, um, the half one mag seven, Pat NVIDIA eeked out a 7% game. Microsoft shed 23%. What is going on? Like what's your read on Microsoft? Why is the market hate that company right now? I mean,

Patrick Moorhead: 

So they got driven to extraordinary lengths when they were the open AI, the only one that open AI. And then, um, I think, uh, who was it? Brad Gershner did a, uh, podcast with, uh, Sam Altman and Satya, uh, did one and I forget who, who he did his with. And then it pretty much all came crashing down, which, which was, and once you had the, the aura of open AI going down. And then it's funny, somehow, Microsoft couldn't win on either, which was higher valuation because it's being disconnected from OpenAI, who's losing some of its luster to Anthropic, you would think it would go up. I firmly believe that it gets more bundled with SaaSpocalypse than it being a leading IaaS provider or a PaaS provider, regardless of

Daniel Newman: 

how much uh cash it prints so yeah it uh they just got spooked yeah no i mean i think it's roi i think it's uh some of it's the exposure to open ai i think some of it is the backlog and that exposure to open ai some of it is it's Software business is different than the other infrastructure companies, but I think it's mostly nonsense. I would say probably a generational buying opportunity here. If Microsoft, I think it's been way oversold. I think they have the capability to execute well and it's going to present a really good opportunity for people. Um, you know, NVIDIA has done nothing for like a year. So on that side of the house, like, um, I know by the way, like God, so much, so much going on with the Kyber thing and semi analysis. I don't know, but I got a, I got a confirmation. I think you got a confirmation from that. The roadmap is intact. Now this is interesting because you know, that's. Obviously, that's not specifically addressing the note. It's generically addressing the note. But legally speaking, putting out an official statement refuting that is a pretty strong rebuttal. to a note like that, right? Because these companies are regulated. And so if they come out and say, you know, in refuting that, that their roadmap is in fact on time, and then that turned out to be true, there's culpability. I'm not sure there's any culpability on the analyst reports though, that come out. It's been really interesting to see how this has played out. Now, I had had channel checks and you'd had channel checks that had been somewhat consistent with the report. that there were delays. Now the timing of delays, I would I never heard a year, but I definitely heard it was doing back.

Patrick Moorhead: 

I mean, for me, it was the difference between, like volume and actually shipping, right? Like ramping, and for ship. Right. FCS first customership versus versus volume. And that was a lot of the a lot of the consternation and hand wringing. But but there was also a challenge. I mean, it was a very aggressive, really aggressive. Design that they were going for, and I think we've been here before. I mean, even, you know, I remember the Blackwell metal spin freak out. They ended up doing some material pull-ins to make up for it. I see NVIDIA spending a bunch of money, or they need to, to pull this in. I do wish that I knew what, I mean, when I was doing products and either building or managing products for 20 years, our roadmap is intact to me means either same committed timeline, same features, or a new roadmap that means the same features and same timing as the original. Is that the way that you read that, Daniel?

Daniel Newman: 

Yeah. I mean, look, I don't have the hands-on product management that you had, but what I would say is following this industry for a long time, when a public company that's regulated comes out and refutes a report like that, it's a pretty strong statement that what they promised shareholders is still true. But did they refute it though? Well, again, that's the, that's probably the real question to ask is they didn't specifically say Kyber, uh, and the PCB backplane and the, and the whole, um, you know, one 44 versus 72 by two, you know, is, um, is, is exactly like, they didn't come out with the details of that architecture and specifically call it out. But at the same time, they presented that architecture as their roadmap, right?

Patrick Moorhead: 

Yeah.

Daniel Newman: 

So, I mean, again, words matter, but they haven't announced a new roadmap, have they? No. So what I'm saying is the roadmap they presented at Investor Day, the roadmap they presented to their shareholders, the roadmap that sits inside the deck on their website right now that you can download, is what they're saying is intact. So again, I'm sure they've had their lawyers deeply review that statement, and I'm sure the reason they picked the words they did, but if nothing else, Um, I think there's a massive liability on one side of this when, when, you know, so if they are lying, then they deserve the liability that will come with it. They'll get sued by shareholders. I don't know if there's any liability for, you know, this is kind of the, what was the Hindenburg thing? Remember the Hindenburg research guys, they published the report. And I mean, it's a little bit what semi analysis is doing here, isn't it?

Patrick Moorhead: 

it is definitely a similarity. Right? For sure. For sure. So not exactly the same. But I mean, some analysis has a lot of other businesses that it does, you know, but it'll come out of the wash.

Daniel Newman: 

I'm not saying they're short sellers. I'm just saying that that was the game that was played to move a market. And it works. But what I what I am getting at is like, we'll see in a year, right? I mean, the bottom line is, is if in a year it ends up having this delay, they can hold up their belt again and be the absolute champions, right? But if they're not right, or even if they misqualified, like to your point, I think the most interesting point you made, Pat, though, was about the scale, meaning that could both be right. Time to first customer is intact with the roadmap, but the time to ramp is like really long and it's going to take a long, long, you know what I mean? And I think that's probably the most likely landing point sits in the middle is they won't be able to ramp this thing in a timely fashion. All right. All right. Enough of that one. Beat it to death. We did. Are you writing a novel? Stay with me, kid. We're almost done. I'm just trying to be like you, dude. What are you, tweeting rockets? What are you doing? Emojis and, yeah. Emojis, emojis. All right, last topic. Just real quickly hit this one. But quietly, the end of the sas-pocalypse, I don't know, the end of the cyber-pocalypse, Palo Alto and CrowdStrike had a comeback. Not only did they have a comeback last quarter, they had their best quarters in history in that Q2. So after Methos was going to end cybersecurity, and after Claude was going to destroy all SaaS-based tools, you have a CrowdStrike, which could be considered both, went up 95% from April to June, and Palo Alto went up 113%. Good earnings, but the bottom line is it looks like the market no longer believes that we're going to deploy a frontier model and replace all of our cyber tools. I don't have much else to say about that one. You?

Patrick Moorhead: 

I mean, I never thought that they should ever be lumped in to SaaSpocalypse because security is one of AI's biggest risks. I mean, nobody woke up and said, the biggest risk to AI and the biggest opportunity is CRM, right? Nobody woke up and said, supply chain or HCM, right? Pretty much everybody said security. So yeah, it was done to lump them in, I should have bought, but now that they're high, I'm gonna dive in.

Daniel Newman: 

Now you're going to go all in. All right.

Patrick Moorhead: 

Perfect time. Yep. And in my head it's going up to 200.

Daniel Newman: 

We'll call it the PATF. Yeah. All right. That's it, man. That's all she wrote. So, you know, good show, Pat. A little different start in the week. You're going on vacation, dude. I am. Going on vacation.

Patrick Moorhead: 

I'm going to be in Wimbledon at the men's finals on Sunday. I'm going to go a couple of days early, hang out with my wife of 35 years. Nice. Yeah.

Daniel Newman: 

I'm, uh, I'm, I'm going to be relocating to a bunker somewhere where the water's warm and the skies are blue. I'm going to do on Sunday. I might, I might, I might podcast with you while I'm there. I don't know. We'll think about it. We'll think about it.

Patrick Moorhead: 

Please go to a fricking Island. You can get out of.

Daniel Newman: 

Yeah. I can't promise I'll be back in time. I was going to rush back for an event, but I'm hearing rumors that no need to come back so fast.

Patrick Moorhead: 

Yeah.

Daniel Newman: 

I heard they made a terrible choice. Terrible, terrible, terrible choice.

Patrick Moorhead: 

Horrible choice.

Daniel Newman: 

Ms. Charles Barkley would say, that's terrible.

Patrick Moorhead: 

Terrible, terrible, terrible. Love Charles Barkley.

Daniel Newman: So I won't say what event it is. Nobody needs to be called out. Maybe eventually, but not right now. Anyway, great show, Pat, great being with you. Thank you all for being part of this. We appreciate you. Hit that subscribe button. Join us for all of our six, five, this is episode 311, and it's going down, down, down, and out. See you all later.

‍