AI Safety Alarms, China's Distillation Reckoning, and Qualcomm's AWS Breakthrough: A Pivotal Week for Enterprise AI

The handpicked topics for this week are:

1. Anthropic Researcher's Doomsday Claim Sparks Congressional Response: Anthropic researcher Jacob Coxon left the company after six weeks and posted a claim that AI carries a greater than 10% chance of ending humanity by 2030, a post that pulled more than 150 million views and landed him on every major network. Patrick Moorhead questioned whether the moment was organic, pointing to Coxon's ties to activist groups. Daniel Newman called the underlying technical claims thin, noting Coxon could not explain a concrete mechanism for how AI would end civilization. Jensen Huang and Dario Amodei both pushed back publicly, and Amodei argued safety claims only carry weight when the underlying work happens in the open. (The Decode)
2. Qualcomm's AWS Deal Validates Its Data Center Silicon Bet: Qualcomm signed a multi-generation collaboration with AWS worth up to $60 billion over a decade, with an initial issuance already covering roughly $9 billion in committed orders. AWS also picked up warrants for 25 million Qualcomm shares tied to the spending commitment. Daniel Newman framed the deal as proof that TSMC capacity plus a credible product lineup is enough to win hyperscaler business, even against entrenched incumbents like Broadcom and Marvell. Patrick Moorhead called AWS the hardest hyperscaler to crack given its in-house CPU and accelerator competency, making the partnership a bigger validation signal than the market's muted stock reaction suggested. (The Decode)
3. Apple's Ternus Era Opens With a Pricier iPhone and a First Real AI Monetization Signal: Apple's first major product event under new CEO John Ternus introduced a $1,999 foldable iPhone Duo alongside a $100 price increase on the iPhone 18 Pro and Pro Max, tied to a 4x jump in memory costs. Patrick Moorhead noted the company already guided gross margin down to 48% and softened the price hike with lease and Apple One bundling options. Apple also detailed its first concrete AI monetization plan, introducing token limits for cloud-based Siri processing beyond what its on-device AI can handle. Moorhead remained skeptical of the actual Siri experience, expecting the rebuilt assistant to stay in beta status for up to a year after launch. (The Decode)
4. Federal Agencies Name Six Chinese AI Labs in a Distillation Advisory: The NSA, CISA, and FBI issued a joint advisory naming DeepSeek, Moonshot AI, Alibaba, Minimax, StepFun, and Z.AI for running industrial-scale knowledge distillation campaigns against US frontier models. Daniel Newman linked the advisory to a separate Anthropic security report describing API exchanges in which Chinese labs extracted answers directly from Claude. Patrick Moorhead pointed to his own August analysis of Kimi K2, which he said was distilled from frontier models months before the government confirmed the pattern. Newman estimated the distillation activity generated roughly $40 billion in revenue attributed to Chinese labs, separate from the $5 billion Anthropic removed from its own reported ARR after tracing distillation activity back to Meta's Muse model. (The Decode)
5. IonQ's Investor Day Shows the SkyWater Bet Paying Off Early: IonQ used its September 9, 2026 investor day to show tape-out cycles falling from nine months to two months since acquiring SkyWater, alongside a raised combined revenue guide of $450 to $460 million, up from a standalone range of $280 to $290 million. The company confirmed Qolab as its first merchant foundry customer, with roughly a third of SkyWater's wafer capacity currently booked. Patrick Moorhead flagged that IonQ still has not disclosed gate fidelity and error rate figures for its Superion 256 roadmap, data he believes is more important than the qubit count headline. Daniel Newman placed IonQ alongside IBM as the only two companies pursuing a fully vertically integrated quantum stack, running at what he estimated is a $100 million quarterly pace. (The Decode)
6. The Flip: Does the Distillation Advisory End Enterprise Use of Chinese Open Weight Models?: Daniel Newman argued the joint federal advisory makes Chinese open weight models like Qwen unsponsorable inside major enterprises, regardless of legality, because no general counsel wants a breach traced back to a model the FBI specifically named. Newman pointed to open weight alternatives like MBZUAI's K2 Horizon as a viable substitute already gaining ground. Patrick Moorhead argued the advisory specifically targets malicious extraction techniques and leaves legitimate distillation untouched as a valid technical method. Moorhead cited Meta's reported $5 billion in distillation spending against Anthropic's Claude as evidence the same technique operates as standard practice among US labs too. (The Flip)
7. Citi's Bullish Call Ignites a Broad Chip Rally: A bullish Citi note on the semiconductor sector combined with CPU pricing chatter and renewed custom silicon momentum to send Intel, Qualcomm, AMD, and Broadcom all higher on September 8, even with oil prices falling. Patrick Moorhead credited the move to growing confidence across the entire AI buildout chain, from NVIDIA's line of sight toward $1 trillion in bookings to the financing and power capacity required to support it. Dell shares climbed almost 12% and HP gained roughly 11% on the same data center capex read-through. (Bulls and Bears)
8. Oracle's Backlog Keeps Growing as Free Cash Flow Turns Negative: Oracle posted $19.3 billion in quarterly revenue, up 30%, with cloud infrastructure revenue up 121% and remaining performance obligations climbing $209 billion year over year to $664 billion. The company delivered more than 300,000 GPUs and signed over $30 billion in new AI cloud contracts during the quarter. Patrick Moorhead flagged that Oracle generated $23 billion in operating cash flow, funded largely through customer prepayments and equity, in the same quarter its free cash flow turned negative. Daniel Newman added that Oracle's cloud gross margins trail the larger hyperscalers, a tradeoff he said the company is betting will resolve through pricing power once customers are locked in. (Bulls and Bears)
9. Adobe Beats Estimates Again, AI Revenue Still a Sliver of the Business: Adobe posted record revenue and non-GAAP EPS that beat estimates, with AI-first ARR up 150% to more than $650 million. Patrick Moorhead pointed out that figure still represents only 2.5 percent of Adobe's total revenue base, following a June pricing reset that redirected roughly $500 million toward a freemium push for Firefly. Adobe shares lagged the broader software rally that lifted Salesforce, ServiceNow, and Snowflake sharply higher on separate AI announcements during the week. (Bulls and Bears)

Watch the full video at sixfivemedia.com, and 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 video, we may discuss companies that are publicly traded, and we may reference their equity share prices. Nothing discussed during this webcast should be considered investment advice or a recommendation to buy or sell any security. We are not investment advisors, and you should not rely on this content as financial advice. Six Five Media collaborates with technology companies and industry leaders to produce research-driven interviews and multimedia programming for enterprise technology audiences.

Daniel Newman:

Hey, hey, we are back, everybody, Friday. It is a day that requires a moment in remembrance. It is 9-11, 2020. 26 at 25 years. Remember, hope never forget whatever term you need. So where were you? It's funny. I was at my dad's office. So this was kind of in a tricky spot in my own life. Cause this was like, right. Uh, three months before the birth of my first child. Um, and I was all of 20 years old. So I was, I was going through it at that point. And I remember I was sitting in the office working for my dad, you know, still going to school in the city. And I, um, They had, you know, it was like a 13 inch little color TV that was in the, it was a trucking company. So it was like a very kind of, it was not a pretty office like we think about in tech, right? And it was like sitting on top of an old file cabinet. And I just, like the news came on and I remember, like my dad told me to go home, but then I remember sitting in traffic and it was just, it was like the most surreal day ever.

‍

Patrick Moorhead: 

So yeah, it's crazy. Yeah. I was a VP at AMD at the time and I was in San Jose, uh, going to visit the headquarters. It was in the morning. And then, uh, uh, my wife kept calling me and she may have been sleeping and kind of woke me up. And, uh, And then just ended up watching it on TV for hours. I actually, I don't know if you know that, but they stopped all airplanes. I was in California for probably a week with about two days with the clothes. But yeah, what a crazy, crazy day.

‍

Daniel Newman: 

I was in New York about 10 days before that. Flew American Airlines to New York two days before that.

‍

Patrick Moorhead: 

Yeah, I was there two weeks before too at World Trade Center.

‍

Daniel Newman: 

Yeah, just wild. Like I was talking about like sliding doors, like just the smallest shifts in moments in our lives. But anyway, just everyone out there, you know, at least on this show, we remember this. I want to go, you know, shout out to everyone who. was affected by it. And I hope we put, you know, however you, however you cope, pray, whatever you do, it's a big day. But on a more positive note, the world's going to end and everyone's going to die. So we had that happen this week. Only a 10% chance though, Daniel. Well, that was that person, but didn't someone else come out? Like an open AI guy says like 70% chance. And then And then there was a whole lot of great mockery and memes. I saw some really, really funny ones that came out where, like the Apple, the guy, I'm an engineer at Apple. Nobody's building foldables responsibly. Anyways, but yeah, crazy week. A lot to talk about. Welcome back to Six Five. In case I didn't say, I got right into the dark, deep stuff. But hey, you know, It's the recording day. And of course, you're probably listening to this a few days after. But we are back. Pat's on vacation, you know. But there is no vacation. Pat was on TV this week. Pat was blowing up my phone trying to do deals this week. Pat's coming up to the podcast. He's slowly morphing into me. where he's basically working on vacation. I thought he was better than that. He's not. But how are you doing? I don't want to say where you are, but you're somewhere beautiful, mountainous, fresh air. Feeling good?

‍

Patrick Moorhead: 

Yeah. Feeling great. I'm in the Aspen area. So we'll, uh, we'll just leave it at that. Not in a snow bunker. It's no snow period, but no, listen, Daniel was an entrepreneur and you know this, and you know, I've been running my own company for over 15 years now. And you know, you just, I mean, you run the company, you're, you're always on. Uh, what makes a vacation for me though, is not to do end to end zoom calls, right? That's, that's, I just, You know, and this is the only the second time I'll be on video. I'm here first with CNBC and I couldn't miss the pod. But that is a vacation to me doing big thinking, introspection, working out a lot. Did you did you come up with anything? I have I have and I can't wait to share it. Can't wait to share with you when I get when I get back or we can even talk this weekend.

‍

Daniel Newman: 

Well, maybe I'll make time. When are you coming back? Actually, when are you actually back? Monday?

‍

Patrick Moorhead: 

Monday. Yeah, I head up to Seattle for some customer meetings up there.

‍

Daniel Newman: 

Okay. I got to go to San Francisco. Lucky me. Dreamforce. There we go. Yep. So I think we'll be like ships passing in the night or something like that from when I… I'll look out my window and wave. Yeah, you can do that for me, buddy. All right, so what do we got on the docket? Let's do that. It's always an important thing to say, hey, what are we going to talk about on the show today? We've got end of the world. That's pretty interesting. Kind of, yeah. We're going to talk about some big data center deals because people are going to keep building, even though the world's apparently going to end. And then we got the really interesting stuff, foldable phones. But hey, since you went on TV and we're all baller, we need to talk about that. China is doing things again. At least Americans do not believe are ethical. Again, different rules, different regions. We'll talk a little bit about that. And then, you know, we'll talk a little bit about Investor Day for our IonQ and our personal friend, Niccolo Domasi, another big one. But you made some interesting, you and your team made some interesting calls on quantum and we published some new research on their Skywater acquisition actually just yesterday. I wrote it myself. I know I'm a quantum guy. What can I say? And then we'll go to the flip. That's where you and I simulate an argument, but there's no simulation of who wins. And then we will hit the markets. It was kind of the last week of earnings for the wave. And now we're going to get a five or six week reprieve. where we can then go through all of the various media doom narratives about why AI is going to end the world and why inflation is going to run rampant and why 90% of us won't have jobs. And am I sounding cynical today? Is this coming out? Tiny bit.

‍

Patrick Moorhead: 

OK, well, we all we all we all get our days, do that.

‍

Daniel Newman: 

No, we'll have our day. It was a leg day. So it starts with self punishment. Uh, anyone that knows, you know, a real leg day, not the one where you just get on the extension machine and do a few leg extensions. And then you do some leg curls and you're like, you know, I'm talking about the one where you actually lose oxygen to your brain. And when you stand up after you're done with your set, you fall over. That's the kind of like that I did just so I can let everybody know. Oh man. All right, Pat, no more yapping, no more BSing, no more opining except for about tech things. Let's go ahead and hit the decode. All right, I feel like this is gonna be like the majority of our show here. Like we'll hit all the other topics, but like probably should spend some time on this one. An anthropic researcher named Jacob Coxon basically came to X, drives 150 million plus views, becomes an overnight sensation, goes on every TV network. This guy has never tweeted anything or had like five tweets in his entire history. And basically comes out and says there's a greater than 10% chance that AI will be the end of humanity by the end of the decade. And basically said that as an insider at Anthropic and OpenAI, the labs are irresponsible. I'm not saying more, I'll give you the mic, but what happened this week?

‍

Patrick Moorhead: 

Yeah, so, you know, about once every month or so, something new that shifts the narrative and, you know, essentially, I would say, primarily Anthropic has been trying to get everything to slow down. Every once in a while, Sam will chirp in and talk about how he supports that. And the whole theory is we're building so quickly that we're not putting the safeguards in that could potentially destroy humanity. You know, this guy was at the company Anthropic for, I think, six weeks. And he's also affiliated with different groups that I would say are, I'll call it the far left, and which I think is an adder. I think connecting those two doesn't necessarily mean, oh, you know, this is necessarily a setup, but The guy's been in there for six weeks. I mean, can you imagine, you know, I show up to AMD and six weeks after, you know, making weird and odd claims, I think everybody would just ignore me. I think the press is hyper tuned into this this this conversation. I think. you know, if I look at, you know, if we can make an AI so powerful, and this was, I wouldn't say it went viral, but, you know, if we can make an AI that's so powerful, why can't we make an AI that is so powerful that it can not destroy humanity? It was pretty wild. And as you'd expect, kind of the far left, the socialists kind of jumped on here. You had Bernie Sanders in our, loser Austin representative Greg Cesar. He was a nightmare when he was on the city council. That's kind of when the city started going down. They're trying to use this not only as a we need to slow down, but make this a make this a political statement. And Bernie went so far as to say, hey, you know, there's going to be a two year prison sentence. if you violate some sort of, you know, safeguard or restriction. I'm all in on some sort of a kill switch, which, by the way, could be, you know, turning off a data center, or what would be easier is turning off the internet through the very few true carriers that, you know, do like L3 connectivity. So I think this is just another wave of craziness. I like the conversation. I don't think it has to be as polarizing though, right? Do you think this is organic? At this point, I think overall in totality, I don't think it's organic. I think it's a planned campaign. I don't know if this specific example is and I know you were challenged on on X on it on it being a campaign or not. But the fact that he was there for six weeks and he's affiliated with these these far left groups kind of leads me to leads me to believe that he could be part of the campaign. Did you listen to any of his interviews? I did, and it's funny, I'm not a researcher, but it sounded like nonsense to me. He couldn't answer some basic questions on how exactly the Earth, humanity would come to an end. And I was expecting something like, you know, taking over this facility and backdooring this and getting into, you know, the launch codes on the magic football that our presidents carry around with them. I didn't I didn't get in any any detail around that.

‍

Daniel Newman: 

Yeah, it's weird. It's just the whole thing to me. Super weird. It's. You know, going through just the football that we've been playing for three years, it's like we're just continuing to push the envelope or escalating. It's funny, all the online chatter about, you know, bubble and the circular, it all died this week. Like, I don't see anyone talking about that now. We're all just talking about the fact. So we basically, you know, my sort of hyperbole to the hyperbole was, so effectively now we've ended up where we all agree it's too powerful. It's so powerful and so capable and so important to the world that it could actually become, you know, stationed or what's the right word I'm looking for sentient is the right word I'm looking for and decide to go robo terminator on us. And by the way your point about like what it can do because I feel like this conversation becomes more important when the physical robots hit their next level of intelligence. Like what we've built with drone technology scares me more. What we've built with autonomy scares me more. Like, you know, we have now, you know, recursive drones that can learn to target very specific, and by the way, if you read that anthropic report that we'll talk probably a little bit about, like the people using AI to try to get answers to do things, like that to me, like humans doing things that are more accessible to them because of AI. But all this is, is it's like the internet on steroids, right? We had the YouTube era and the Google and the dark web era where you could buy, like people who wanted to do nefarious things could find access to how to build a, you know, a weapon of mass destruction, right? Online and build it. But it still was a human. So the real question is, at some point, do the robots start using the models to build the weapons to basically, I mean, it's like the Terminator shit. It's like a Terminator movie plot. And maybe it's possible. What's interesting to me is just how gullible we are and like has a society to just absolutely wrap our arms around this thing and go. Yeah. Like this guy, what did he work six, three months or something at Anthropic? Six weeks. Six weeks at Anthropic and a few years, I guess, at OpenAI. He has all these really interesting affiliations. Call them what you will, but they're all tied to, um, you know, uh, these very virtuous groups, justice groups. I mean, the trail is very interesting and where it goes. It falls closer to the socialist viewpoints. It falls closer to the, you know, you saw Bernie Sanders really got on this one, jumped on this one. And to me, like I said, there's just no true example here, but this does bring up another point. Like, are we, do we need to have more alignment between the regulatory shape and the technology innovation that's going on? And I think that conversation needs to be had. I guess the outstanding view or the overarching view right now is that we won't slow down or even regulate or give enough time to regulate because the innovation is happening so fast. And unless the world believes this is existential to our existence, no one's going to stop it.

‍

Patrick Moorhead: 

I did like Jensen, you know, Jensen loses a lot, you know, being hyper aggressive on certain things. He said, Coxon's claims were outlandish and deeply untrue, which, you know, back in June of 2025, Jensen said he disagrees with almost everything. Dario says, he added, if you want things to be done safely and responsibly, you do it in the open. Don't do it in a dark room and tell me it's safe.

‍

Daniel Newman: 

Yeah. And but can we just say like everyone, you know, go back to what Benny said on our pod when he said everyone's talking their book. You can agree that to Jensen, it's beneficial for things to be open. And I think for most of the world, it's beneficial, by the way. For Dario, it's it's beneficial to have closed regulatory captured and maybe a two company race, right, or a two or three company race. They're weeks away from an IPO. I mean, There's so many things here that, again, I'm not going full conspiracy here, but there's so many things that are sus. And the history of this particular company is anytime they feel they're losing market momentum is to come and say something absurd. Right. I mean, he said within one to five years that we could have 10, 10 plus percent unemployment. And 90% of white collar entry level jobs will be gone. Now we're about going on about two years since that comment was made. And that timeline gives him some room here, but we've actually, the economist came out this week and said, we're at a million net positive jobs because of AI. And, you know, I was one of the probably more cautious, like I saw what AI could do, and you and I were like, you know, optimization maxing for a while. Like, look at all the things that we can, and like in the end, like, it's like, there's still this kind of last mile need for humans in the loop, because AI does a lot of stuff well, but it just can't finish the job. So maybe at some point when we have these like great humanoids that can do the rest of the stuff that needs to be done in the physical world, but like, there's just parts that don't, anyways. So long story long, absurdly overplayed this thing. We need a balanced debate to go on here. I'm glad that we saw some logical voices come out. And I think the strongest voice was probably Sachs that came out and said, we have to stop the IPO. Like, if this is true, we need to stop the IPO. And let's be realistic, like if this was not just some media and PR ploy, this should be taken very seriously. should be taken very, very seriously because, you know, you've got 8 billion lives in the balance. All right, so something a little less morbid, you know, the week did start out with another deal. You and I both, you know, had a lot of comments out there, a lot of media pickup on our comments out there, but, you know, Qualcomm, entered in a multi-generation collaboration with Amazon to build custom silicon for a large scale AI data center, focused around their AI inference and connectivity. Not a ton of details on exactly the parts, unlike some of the other deals that they announced at their investor day, but what attached to this deal was a partnership. And the partnership was in a, you know, Marvell like AMD like stock type deal where Amazon has the opportunity to buy a large block. of shares, I think the, remind me the number, was it 25 million shares? It was, yeah, 25 million shares. 25 million shares, and that's tied to them spending about $60 billion over a decade. And it looks like based on the initial issuance, about 9 billion of orders committed. Now, again, 5 billion was their expectation in this fiscal year, which I think starts October 1st for Qualcomm, then 15 billion by 29. So this was part of that forecast. This was sort of what I would call the proof behind where their numbers were coming from. you know, the, the company's stock jumped pretty sharply immediately, but then tapered off pretty quickly. I think people might've not known whether this was incremental to their forecast. And, and, you know, but what I read out of this pattern, I'm sure you've got takes is this is the kind of proof that Qualcomm needs. Like I, have been continuously saying that in this abundance era, it does not need to be at the cost of Broadcom, the cost of Amazon's own in-house silicon projects, Marvell, or NVIDIA or AMD for that matter. The demand is so outlandishly large that if you have TSMC capacity, right now and you can build, you have the capability to build a quality product, which is important. You actually have to have the know-how, the capability. You can sell it right now and I think that's where Qualcomm's initial market share comes from. They've got product, they can build CPUs, they can build accelerators, they can build networking connectivity that can expand data center. You saw the size of the commit for Amazon. I think Microsoft this week made a massive leap in its commit to build. I mean, the amount of compute these guys are standing up is an all hands on deck. Qualcomm has capacity, Qualcomm has products, Qualcomm has capabilities, and that you're starting to see this work. Literally, you're talking about a single low digit market share at this point. And so I think this is tons of upside. And my real positive read on it is at least now the Apple woes of the past should be forgotten because they've added a much higher margin, much better and much more secure long-term set of businesses. And this will not be their only massive partnership with a hyperscaler.

‍

Patrick Moorhead: 

Yeah. You know, it's interesting, Daniel. The whoever has compute people will buy. I used to really agree with that, but I really don't anymore. Because if you're doing it out of out of TSMC, wafers are fungible at the same at the same node level. And I think Qualcomm has something really unique here. And, you know, I think that this is proof positive. The interesting thing is that this has not, you know, I don't know for a fact that this has anything to do or not to do with HBC. But what people forget sometimes is that AlphaWave, the IP company that Qualcomm bought, was very much enabling AWS Silicon. They were already in there. Yeah. And this is this, you know, over a 10 year deal. You don't know exactly what that means. Right. 10 years, 60 billion dollars. You know, it could could include CPUs as well in into the future. But I think this is monumental for Qualcomm and I would say proves the naysayers. And I think the hardest company to get in, in my estimation of all the hyperscalers, would have been AWS. They have competency on the CPU side, they have competency on the accelerator side, and this was a huge surprise for me. I don't think that the stock didn't move for the exact reasons I said it wouldn't move, which were on a call that you and I were on, which was that um you know sell side looks 18 months maybe two years out and without granularity and without you know somebody saying hey the 60 billion dollars is is a locked in purchase order and irrevocable um the market didn't digest it, even though the likely terminal value of this thing is probably going to be $60 billion plus. But that's the primary reason. And I think Akash CFO got on, I think it was either Citi or Another conference gave a little bit of light on the margins on, I think you had mentioned, did you mention nine, $9 billion? Yeah. So a lot, a lot of good stuff here, but I do think that the market is underplaying this, this point for reasons that, you know, Qualcomm has traditionally been looked at as, as a smartphone play. has really diversified a lot given things like automotive and IOT and a bit of a late comer to the data center AI market.

‍

Daniel Newman: 

Yeah, absolutely. All right. We'll hear more about that. I mean, look, this is one of those where I just think the market is going to continue to demand proof. I think there's a healthy skepticism. It took AMD a while to get through this too, as you recall, for the companies that have sort of played in and entered and exited. But again, the demand curve here is just completely different than those eras in terms of the opportunity, the demand. You know, as you can tell, all these companies that have made massive investments in their own capabilities are still buying outside. It doesn't matter if it's the hyperscalers buying extended power and compute or silicon or it's because, again, we just, you know, standing up 200 gigs or whatever of power and compute over the next handful of years. You know, what do I have? Fourteen trillion dollars of spend, you know, so Qualcomm getting its little share can become a really nice business. So you went on TV. I said, I said, Apple, take my money. I legitimately am excited about the foldable, not because it's innovative or cool, just because that size is just perfect. Like I'm always in between. Do I bring my iPad or don't bring my iPad? Like it's all something for me. So Apple had its big, big, big moment and new CEO. Like what was your what was your read and working while you were not supposed to be?

‍

Patrick Moorhead: 

Yeah, so three different things going on, not going to go into each product and screen size and colors and stuff like that. Yeah, three things I was looking at were going in was how's Ternus going to do? Is the delivery going to be, you know, completely different than Tim Cook? The second one was increase in pricing on the iPhone, because iPhone is really the epicenter of all the profits that the company has. Sure, they sell Macs and they sell iPads and watches, but attach onto iPhone plus services is the big deal. And the third one is, and we've been asking this for three years, maybe longer, how does Apple monetize AI? And the first thing is I think there were a lot of let's start with the iPhone $100 increase on. on iPhone 18, they've already guided to reduction in gross margin from 51% to 48%. They added some lease programs. And I think the $100 increase was less than I'd expected, given there was a 400% increase in the cost of memory. So this makes a lot of sense. CNBC asked me about the super cycle, and I basically said, you don't create a super cycle by raising prices. But I do think Apple hedged itself. If people were offended by the extra $100, hey, do a lease program or do the Apple one, all you can eat program. The second thing I thought was interesting was they finally gave a little bit of detail on how they're going to incrementally monetize AI. Apple has a hybrid AI architecture of what gets done on the smartphone and all the private data stays on there. And then it bursts to the cloud, which essentially are a bunch of Macs sitting in a rack to be able to give it services. And there will be, quote unquote, token limits. I doubt that Apple will call it that, that will be in iCloud for Siri, And they say they will charge incrementally if you go over those limits. So there was no pricing introduced. So people can't model it. Nobody's used it yet. So, you know, nobody's done the research to say, well, you know, typical user might get, you know, 27 minutes of cloud AI. So, but I do think it's a positive for the company. and how they're going to monetize that. I will be super skeptical on the experience given how poorly Apple has done so far on the smartphone and it will still be beta. It might be beta for a year once the, literally the software goes out to everybody.

‍

Daniel Newman: 

I mean, they had five years or more from the time these things started hitting the market to play with it.

‍

Patrick Moorhead: 

I mean, yeah, this is the the new the new Siri, though.

‍

Daniel Newman: 

And I get it. I just I just mean, like, at least on the foldable side, I mean, these things when when did when did the first one launch like 19?

‍

Patrick Moorhead: 

Oh, I think it was twenty eighteen. Or maybe I'm confusing that with the eighth generation of the Samsung.

‍

Daniel Newman: 

It was around that. My point is, I think it was pre pre covid. So I'm just saying, I was at a lot of time to try to kind of get this right. I mean, we always kind of call them as a bit of a fast follower, you know, it's not even fast in this case. Like they just kind of got around to it now. And it was like, after all these years, that was the best idea. Doesn't mean I don't like it, you know, as someone that uses an iPhone. And yeah, I mean, I thought your hit was good. I mean I still don't have any like conviction about Siri. I still just think the Apple is the place you're going to use Claude or GPT. Like that's still kind of my read on it but I do still think that's where people go and I do think that you know having that particular profile of you guys like us that kind of bounce between when you're always on the move it's kind of nice when you're going meeting to meeting to meeting to be able to just kind of break your phone out and open it up and use it for 80% of the value of an iPad? Yeah.

‍

Patrick Moorhead: 

I don't know, man. I, every time, you know, I use a fold, I'm super excited at first. And then it just, it just kind of wears off. And the simplicity of just having a single, single screen phone, I gravitate toward, I mean, a tri-fold might be different. Right. Cause I carry a 10 inch iPad or I guess it's 11 inches now. And I would love to be able to have a trifold. So I truly would only need one device.

‍

Daniel Newman: 

Yeah. I guess we might get there at some point, but I still carry a bag full of tools. I carry three laptops around. I got a Mac, I got a XPS. I've got an iPad. I've got two phones. I mean, I'm, I'm, you know, I'm never. Never traveling light. Plus I gotta bring all my supplements with me. True. A whole bag for that alone. This doesn't happen naturally. Anyway.

‍

Patrick Moorhead: 

I do have a question for you. Any thoughts on John Ternus or how he did?

‍

Daniel Newman: 

I mean, I think it's too soon to have a full verdict. The one interesting thing about John is just, from what I'm hearing is, you know, this is still not totally John's show. It's interesting. And I mean, obviously he's new in the seat. How much influence Tim continues to have from his chair versus how much John, you know, I told you, I liked that he came out and he's being social and he's kind of talking. Cause I think that was a missing, a bit of a, everything Cook sat as CEO, even when he went to X always felt very antiseptic to me. It felt very like, you know, like it feels like, like John might be a little more personable and definitely back to product. But like I just this isn't him yet. I mean, this is not his like he's still kind of reflecting the Cook era. Let's give him the next launch and I'll have more to say about that. So let's keep moving. Another top story. China. Maybe surprisingly to you, maybe unsurprisingly to you is apparently like has gotten the attention of basically the CISA, that's just the Cybersecurity Infrastructure Security Agency. Yes, we have that. Coordinated with the National Security Agency, the NSA and the FBI. And they released a warning that China-based AI companies are running aggressive, malicious and targeted industrial scale knowledge distillation campaigns against US frontier models. They named DeepSeek, which came out with this great new model this week, by the way, Moonshot AI, Alibaba, Minimax, StepFun, and Z.AI. And they've been doing it for at least a couple of years. They're extracting billions of tokens across exchanges. And by the way, this sort of aligns to a post this week that came up from Anthropic, where Anthropic put out a security report that basically went down a litany of different things it's working on, where it called out that not only are these companies using and distilling, they're actually in some cases through API exchanges, having people using their platforms and having the answers come directly from cloud. Wild. I mean, absolutely wild stuff. Again, you and I have talked a lot about like technologically speaking, China is certainly doing really interesting things architecturally to make their models run better. Pico, your son actually tweeted something to me about this, like being very impressed by some of the architectural things that are being done by DeepSea. So this is not to say that they don't innovate or do anything. I'm just saying that what apparently is the case is these, basically these models are very near cousins or inbred siblings to our leading edge frontier models. And the reason that they're so damn good is that they basically use their industrial distillation techniques to build the entire model using our intelligence and then they put their capabilities on top of them, make them really, really good and deliver them really, really cheaply. Although some of that's a little still mythical because I've heard things about API costs rising for some of these Chinese models because they can't actually afford to deliver these at this scale and at the size of these prices for very long. But we all know China is willing to buy a market if it has to. This brings some really interesting questions. There's a lot of people that are debating about distillation as a whole, because distillation is a technique that is very commonly used in ML. There's a lot of debates about the ethical nature of all this, because obviously companies like Anthropic have paid large fines for taking intellectual property that was not theirs in the process of building their own models. I believe they recently settled a lawsuit for that. But I think this also brings to attention some questions about China's own regional ethics and how they handle things. Like I do not believe that China believes that doing any of this is breaking any rules. I don't know what you think, but clearly this debate continues on. And by the way, these models were not trained using Huawei Ascend chips, maybe some, but that's not how this happened. I still believe there's a lot of NVIDIA in the background here. And by the way, I don't believe it's all H200s either. Singapore has a lot more NVIDIA than I think it needs to serve its market. So anyway, what is it serving over there? Good question to ask, but I don't know, Pat, I'll let you weigh in. I mean, obviously we're talking about four, five, six different things, but to have these major agencies bring it to the attention and calling out what some of us, including you and me have been saying for some time, it doesn't give me any real validation, but does it change the framing or does the market just shrug it off and not care? Or is this just what we expect from China?

‍

Patrick Moorhead: 

Yeah, two months ago, when Kimmy3 came out, I put out a post, got 235,000 views, and I had some very special comments on where I made the claim that Kimmy3 was distilled off of Frontier Models. Henson Wong said, LOL yes, you have no clue what you're talking about and rightfully got dragged Paul Marin Patrick you should stop this is embarrassing Even though your boomer audience is less savvy to understand the nuance. Please have some self-respect before before peddling Bullshit, so this was you can look at you can look at all the comments Maybe our producers will will put that put up there, but I I call this retribution and a gigantic a victory lap. And I'm sure some people say, oh, it's the it's the NSA. I mean, they were super specific. Go and look at the report on what happened. This is not hard to figure out. Like we can monitor traffic coming in and out of the country like like crazy. It was just proof that this this did happen. And Daniel, I am very much thinking that the frontier models are okay with this because it drove so much revenue. Discussion about the Anthropic S1 was that they had to redefine ARR based on that they removed the $5 billion a year from Meta. By the way, Meta apparently with Muse is distilling Claude. And the other I think was $40 billion that came from the Chinese labs. 40 billion in revenue. Yes, 40 billion in revenue. Wow. Yeah, yeah. I didn't see that. Yeah, it was pretty interesting.

‍

Daniel Newman: 

I think I saw your tweet about that though. As they're going into filing for their public IPO offerings, How much AR do you get to count? But I guess as long as they keep putting new models out, these will all keep distilling them, right? I mean, this doesn't stop because the one thing I hope you, I think you'd agree with me on this is like, there is no amazing open source, open weight models that are almost as good as Claude and OpenAI. If OpenAI and Claude don't keep doing OpenAI and Claude, like it requires them to put the money up and make these, because that's what people I think miss is like, oh, open weight will be great. But if we don't have a frontier, There is no one investing big. The open way to open source stuff will lack. It won't have anything to still and build.

‍

Patrick Moorhead: 

Yeah, I totally, totally agree. Totally agree.

‍

Daniel Newman: 

All right. Is that it on that one or you got anything more you want to add? I'm good. I'm good. Got that out of my system. I've got a serious bout of haterade on my timeline right now. About? About my comments about that. I keep saying, okay, can we talk about this now? Are we allowed to talk? Yesterday when Deepsea came on, can we talk about how it's so good? It was before that actually dropped. Can we at least acknowledge that it's only this good because it's got the… Anyway. All right. Well, someone had to build the first Ferrari so that Lamborghini could come out. I mean, it's okay. You build the next thing, but we have to acknowledge there is someone at the front. One of the things I have heard, by the way, that is interesting to talk about is, dude, the closed and frontier models become really, really highly paid for solving incredibly difficult. We saw the mathematician stuff. the cancer stuff, like the very big problems that warrant the big dollars. But obviously at some point, like these models do become good enough to be like the respondent to your like, hey, my Starbucks order was wrong. Like, I don't know how much further to the frontier we need to go to get like the, oh, you're right. Your latte was wrong. We're going to make you another one. Like the point is, is like, you know, what we need frontier intelligence for versus what we can use a N minus one, N minus two model for. I think the market starts to bifurcate. And so, um, and obviously having unlimited amounts of compute to throw at problems is pretty awesome. You know, I was talking to my tech team cause we launched our API this week and I was like, look at what they're doing for just 10,000 a day in tokens. You know, but like, you know, if every one of your engineers had 10 grand a day to spend, like, we could build some cool ass shit.

‍

Patrick Moorhead: 

I'd be running swarms all day long and ultra fast.

‍

Daniel Newman: 

Oh, everything like your biggest, hairiest, most audacious problems. You're just throwing endless resources at. And then you're like, well, how smart we are. Anyways, all right, last topic for the decode, Pat. I and Q, you know, you graded the September 9 outcomes versus the August framework. What'd you guys think of the day?

‍

Patrick Moorhead: 

With a company in a stage that the quantum industry is, I think it's important to focus on the products and the technology, right? I mean, the numbers, obviously it's an investor day and they matter, but I think the thing that we were most impressed with was the iteration cycle, chip design cycle, right? Before and after owning SkyWater, right? Tape outs, iteration speed went from nine months to two months. And that to me is absolutely insane. A new roadmap came out, Superion 256 ships in 2027, 10k qubits, green check mark. Neutrality on Skywater, that was the big question. Would they be able to do that? A company called CoLab, I hope I'm saying that correctly, Q-O-L-A-B, is their first Merchant Foundry customer. And only about a third of Skywater's wafers are currently on queue. Combined Revenue Guide, super duper, man. $450 to $460 up from $280 to $290 on a standalone basis open. And stuff that's still open that the company should communicate at the right time is kind of combined EBITDA guide. Plus, the company is losing, even at a loss of 120 million last quarter. Segment disclosure would be nice. And on Superion, it's one thing to talk about qubits, but as we all know, fidelity among those qubits is, and the error rates are the most important thing you can bring out there. But all in all, a pretty good day for INQ.

‍

Daniel Newman: 

Yeah, I mean, look, this is a clearly an important industry, government continues to invest, you know, not necessarily in all cases in IQ, IQ sort of challenging to become the biggest of the fully vertically integrated quantum offerings, whereas other smaller players that are maybe focused on one part compute or focused on networking or focused on sensing, they're doing all of the above. They're clearly challenging kind of the status quo in terms of buying SkyWater and building their own foundry, where I think they now serve nine different quantum companies. So they've become a bit of the you know, they have the opportunity to not only deliver, but also be the TSMC of the industry. Maybe, you know, an IDM with a fab worked well for Intel. It's working well now, working better now. But like, you know, one thing is, is through a combination of acquisitions and organic growth and successful scaling up of systems. And of course, deeply being capable in areas like security and networking, which is really where I think the short-term opportunity lies, getting into, you know, as a service, compute as a service, you know, since not every company wants to buy a quantum computer, but they want to be able to use a quantum computer, you know, they are now basically running a hundred million dollar a quarter business. And That's like 10 to 15x all the other independent quantum companies. You know, I've kind of said there's certainly some technology across the wave and across Continuum and others that is notable. But I do think it's become a bit of a bit more of a two company race at this point for like, you know, full quantum stack. And that's really IBM and INQ. That's kind of the end of two I see right now with other companies sort of in the ecosystem and around it. And the market will be big enough for more than two. But at some point this could be interestingly the NVIDIA AMD race of quantum computing. And of course you always have the Googles and Microsofts and Amazons playing around quantum. But it increasingly feels more like they're going to be a distribution layer of compute assets than an actual, actually trying to be a full stack quantum offering. But we'll see. So successful day. Stock kind of didn't move a ton on this, but there's a lot of external, you know, the market factors right now are palpable between interest rates, Iran, bonds, very hard to know what's moving markets right now. It's, You know, all I would say is ask Pat what he's buying and short it. That's the, that's your safest play.

‍

Patrick Moorhead: 

And I'm just. That's fine. Yeah. Yeah. I've got, I'm looking at a Sennheiser ripping today. Yeah. Okay.

‍

Daniel Newman: 

So that means software's now.

‍

Patrick Moorhead: 

Course. Yeah, my number one, my number one decline, but but only snowflakes down by three quarters of a percent. Okay, so it's not too bad. Yeah, top top of my list right now. My my tracker list is Dell up almost 12%. HP following up 11%. HP Inc up almost 10%. In actually it's in for a ripping, then chips.

‍

Daniel Newman: 

Yeah. I wonder what the, you know, here in the dark, like what the, the, the big catalyst is because NVIDIA is only up 0.5%. So, but, uh, I mean, Oracle had a decent day, but it's given back almost all of its gains straight away.

‍

Patrick Moorhead: 

Yeah.

‍

Daniel Newman: 

Um, weird. Yeah. I mean, Dell is just, I mean, incredible. That's all I'm gonna say is incredible. Like the valuation is out of this world. Congratulations, Michael. I mean, you have just absolutely crushed it. You too, Jeff. Okay. You think they're listening? I think they listen.

‍

Patrick Moorhead: 

I think they might listen. They do.

‍

Daniel Newman: 

Yeah.

‍

Patrick Moorhead: 

I'm pretty sure Michael sits with his Apple TV on a Saturday morning. No, he can't do that actually. Because we publish on Monday. Yeah.

‍

Daniel Newman: 

So it's the, if he has, he's listening to last week on Saturday. Anyways. Okay. Pat, this is the fun part. This is where I have to embarrass you. I'm sorry. We're going to, we're going to, we're going to jump into our simulated debate and we're going to basically decide whether or not this joint advisory that we talked about coming out of the NSA, CISA and FBI is basically It basically means that U.S. enterprises are going to have to go no, no, no to using Chinese open source, open weight models. So I'll ask it this way. Does this advisory mean no more enterprises using distilled U.S. Frontier Lab models being sold at a fraction of the cost? Let's see who's going to say yes.

‍

Patrick Moorhead: 

Ugh. Toxic. You got it, buddy. Explain to everybody the toxicity of open models.

‍

Daniel Newman: 

All right. Listen, nobody's going to ban Quentin. Nobody has to. The moment that three federal agencies put six Chinese labs by name in a joint advisory The question at a Fortune 500 stopped being, is this model good? And became, who's going to sign off on this? And the answer is nobody wants to. That is what off the shopping list means. Not illegal, just unsponsorable. The against side will say the advisory doesn't prohibit anything correct, but also irrelevant. Export controls are commerce's job. This is the NSA, CISA, FBI, a trio that shows up when the government wants boards to hear about national security. The general counsel's calculus is simple. If a breach or an IP incident ever traces back to a model the FBI named as running targeted distillation campaigns and it was legally deployable is not a defense, a CISO survives. The standard for enterprise deployment has never been, is it allowed? It's can I defend the choice in a deposition? Also, the cost of being wrong is asymmetric. I mean, I talk all the time about the hard ROI era. So what does a Fortune 500 bank gain from Quen over a non-Chinese openweight alternative? Marginal capability, maybe some costs, but what does it risk? A headline that pairs the company's name with a three agency advisory? JPMorgan Chase goes against the FBI, the NSA, and the CISA and deploys Quen. and gets hacked and gives all of its important customer data away. Can you believe that? Can you imagine that? No chance. You know, and we now have the argument for substitution and you'll like this, right? You know, a month ago, you maybe had a case before this happened. You know, these models were pretty good. Pricing was good on the shelf. Maybe you'll have some people talk about error gapping and other things. Really not, you know, it's just no longer true. And you talked about the, what, the M-B-Z-U-A-I, K2 Horizon. You know, they're models like them that are fully open. So that's an option. And by the way, they're not hacking and stealing all our stuff. You got US open weight efforts now, much better. The open AI ones are shit, but META has come a long way. And so when something credible within the jurisdiction hits and they're already hitting and you see more and more of that with the support of NVIDIA, that is basically going to make the case for itself. Anyways, I'm sure you're going to say, oh, but distillation is legitimate. Yep, that is absolutely true. But the advisory talks about this isn't just standard distillation. This is aggressive, malicious, and targeted. So you're kind of using the disclaimer, but you are skipping the headline. And by the way, There's no caveat here. So I'll get to the point. I'll say this all. There's a bunch more I could basically say, but enterprises play defense first. They've spent more than two years now building AI governance frameworks precisely so they'd have a process to say no. This advisory is the first thing that the process was designed to catch. Chinese Open Wake, it's not bad. It's just something that no one at a major enterprise is willing to put their name on anymore after these advisories have hit. So anyways, good luck, good shopping. It's over. There's no argument. I don't even know if you need to rebuke this.

‍

Patrick Moorhead: 

Daniel, it's only been a week since Signal 65 brought out Pinnacle, and Pinnacle showed that for a known good accurate output, you can get a 90% cost reduction using open models, which I think is amazing. Essentially, you're espousing inflation, which sucks. But the good news is, if you actually read the advisory, it doesn't order some enterprise-wide exclusion. It's only about malicious extraction. That's the only thing that this is about. And I think the proposition conflates a company, a model, and the deployment. So the specifics, the exact model version, hosting arrangement, data access, and the workload matter. you know, why would any enterprise want to have a defensible procurement decision to save 90% on your bill and run stuff on premise? Legitimate distillation still remains a valid technical method. And this advisory said nothing about that. And You know, I think you made my point when you talked about Horizon K2 with the fact that it's, by the way, it scores the highest on openness for anything. And it's a top 20 result model as well, right? And even with all the new models, all the Chinese models that were brought out even this week, it's still a top 20 output and intelligence model. So, you know, this one's an easy one, Daniel. You just need to pay attention to what it actually said and what it didn't. What the report didn't even say is what is illegitimate and what's malicious, right? Is it a certain number? Is it that you're coming in and spoofing, kind of like your Kimmy example, is it… I don't think we're going to stop distillation. You can bring a hard drive full of weights, full of your model into a country that enables this setup shop relatively easily to be able to do all of this distillation. And by the way, if distillation is so bad, why did Metamuse do it on an Anthropic and spend $5 billion apparently on on that right? Arrest my case, Daniel. What do you really think? I think once they define what malicious is and that models will be put into that category and you you likely won't be able to use them. I think critical infrastructure is the biggest tipping point of, of this whole thing. The funny part is that report didn't say, Oh, Chinese models dangerous. So I think it was more about just setting the record straight.

‍

Daniel Newman: 

It's so interesting that when you bring up all the the financial benefits to these labs, though, of all this distillation. You know, everybody thinks that the spend is all enterprises building shit, but the biggest spend is actually competitors building copies that they can sell cheaply. Yeah, it's slightly enhanced.

‍

Patrick Moorhead: 

Also, we have startups, right?

‍

Daniel Newman: 

It's wild. I don't know. It's interesting. All right. Quieter market week, but we got a couple of things to hit before we close the show. I think we had a couple of earnings, a couple of significant earnings and some news. There's a Citi event this week where lots of companies are speaking, including Intel. Let's get after it. All right. Pat, we're in the home stretch here. I know you viewers are going to be sad when we get to the end of the show. I know Pat and I are always sad when we get to the end of the show. I got to get back to my vacation. That always depresses me. All right. So I mentioned Citi made a bullish call this week on Intel.

‍

Patrick Moorhead: 

What's going on there? A lot of different things going on. And I can go kind of marker by marker and talk about Intel, Qualcomm, AMD, and Broadcom and why they went up so much. But even with oil down, chips just skyrocketed. What happens is the alignment, I'll call it super alignment, of all the different factors that come into play to keep the growth going. You have NVIDIA that's hinting to 700, 750 billion line of sight to 1 trillion and the market kind of yawns. But it says, hey, what's the power that it takes to do that? What's the type of financing that's required to build this out in the entire chain? Oh, are we going to have enough photonics, silicon and capabilities to be able to keep this rally going? I mean, you know, when the type of cement that's used in different data centers, you know, somebody has tracked down some deficit in the amount that we need, or some capacitor or series of capacitors we get, that just seeds doubt out there. And I think the combination of what came out. There were two conferences, I think Citi was one and I forgot the name of the other, where each of these chip companies got up and talked through exactly what their future looks like. So, Citi and Goldman. net, net, the, the growth, the $14 trillion, you know, up and down the chain became a lot more believable this week.

‍

Daniel Newman: 

Yeah. I, I, I mean, I don't even, it's interesting. Cause like you said, these are a bunch of different things happening at once. And, the guides and outlooks just keep going up. I mean, I'm literally watching, I was trying to read while you're talking about what the heck is going on with Dell. Like, how is it up this much? And I mean, look, all we're seeing now is it's like the market is just vacillating between believing and not believing what it's seeing. And I think it was a, what was the CapEx number from Oracle? It was like 115 billion. And I guess Dell is HPE are seen as a couple of the biggest beneficiaries of all this continued CapEx growth. But I mean, look, it's showing up everywhere. And unless Bernie Sanders gets his way and we stop, You know, game on. That's all I'm going to say. So Oracle reported, to your point, and we got some new data there. I think they made some people feel better. And of course, there's still some real questions. Like the growth was good. They beat their cloud IaaS business plan up 121%. So they're still on the infrastructure side growing at a clip that's significantly faster than the bigger clouds. So from a much smaller base, you got to be very clear about that. They're not growing 121% from AWS's size, but they're, you do this long enough and eventually you become very big. And that's kind of how this works. Their RPO jumped 209 billion year over year to 664 billion. I think their backlog is the size of like Microsoft's at this point. And they got about $30 billion in new AI cloud contracts in the quarter alone. And they delivered about 300,000 GPUs. So some big headline numbers. And again, it was a beat all around, raise all around, just nothing to really complain about except for the fact that the debt and the cashflow, you know, I always talk about the free cashflow mafia. I mean, this is the, probably the most sensitive balance sheet out of all the balance sheets in the space. Like they are massively committed and, you know, meaningful slowdown in the build and meaningful failure in the, in the chain, you know, Again, I always say there's the fungibility, but if for any reason, the demand and the ability to charge flows down meaningfully, this is the company that I think has the highest sensitivity to that. But again, overall, like, you know, their, their infrastats growing 62%, the info alone, 121%. They got, you know, 850 megawatts of additional data center capacity that they're delivering on. They booked all that revenue, Pat, like, You know, again, not much to complain about the numbers alone, just the same bare story exists that this is a company that's taking risk on another level than its contemporaries.

‍

Patrick Moorhead: 

Can you imagine the whip when this thing, the risks of that funding just start trickling away? I mean, the whip on Oracle is going to be massive and whip up. But net-net, so let's talk about cash, right? Oracle posted $23 billion of operating cash flow, but a negative free cash flow in the same quarter. And again, this gets to the funding model. But the good news is that the customers are paying up front for it. And when you and I or actually it was just me, I was at their financial analyst day in Vegas, the last one. They made this clear then, but this just keeps coming up. And I just think it's the amount of capital that we're talking about, right? So customer prepayments plus some equity, which is kind of unique, but kind of not. Some pluses, like backlog is becoming revenue, 84%, then 93%, now 121%. on that on that conversion. So that's a really, really good good take. So and I say the bear case, right, if you're getting the discussion is it's moving from demand to dilution, debt, and interest, right? No question about the demand they're getting. It's just now really about, again, dilute, debt, and interest. Say that six times fast.

‍

Daniel Newman: 

Yeah. The only thing I'd add is margin profile for Oracle has been part of how they've grown so much. underpricing in some ways, if you look at the gross margin on their cloud business versus the other three. And again, I don't know that it's a bad strategy, because I think that the bet is that over time they can raise prices, they can add services, they can find different ways to incrementally monetize once they have the customer in. But that's just something to watch is, you know, the growth on, you know, lower margin means less, again, also means less cash flow. But yeah, that backlog is is I'd like to have a 600 billion dollar backlog. Would be bad. Mine's slightly smaller than that. No, but I hear it's growing fast.

‍

Patrick Moorhead: 

It is, man. We've got a juggernaut going. Superstar.

‍

Daniel Newman: 

Pat's a superstar, everybody. All right. He's doing deals while on vacation. Got to do deals from the vacation. Our last one, Pat, Adobe, it's going to zero because people are just going to vibe code it, right? That's what happened.

‍

Patrick Moorhead: 

Yeah, you know, Adobe got got initial lift up there when Salesforce went up 22% in one day. We saw ServiceNow ripped. It's not in sympathy. I think it's more in support. of of those software plays, then you saw a frickin snowflake up another 20%. Okay. And what was the, you know, what's the reason? The reason is not just Oh, getting Dario up on a video with with Mark Benioff. This is all about what degree it's very simple. What degree of your services are led by AI, and what do we believe in the future? Adobe is not lifting nearly as much as some of these other software, and it makes sense. And getting back to the basics, the BDON revenue, BDON non-GAF EPS, missed on fourth quarter guidance, but they raised FY 26 guidance. So, you know, some mixed messages down there. So, And I'd like to say, look, the freemium conversion is what people need to be looking at inside of Adobe. They did a June reset that deferred Credit Cloud pricing increases to fund a freemium push across Firefly at the cost of roughly $500 million of second and a half. ARR. AI First ARR is above 650 million, up 150%. That's real, but it's still, and this gets back to the, I think the crux of the conversation, it's only 2.5% of their entire revenue base. So, I mean, overall, plus new CEO transition, I'm not going to even pretend to say his last name, but it's good that they actually have a CEO in place, and I look forward to meeting him.

‍

Daniel Newman: 

I've had a few conversations with him over the years. He's an insider. They brought an insider. I know the market doesn't always love that, but I think I think that blends very well with the culture. I mean, look, the market's been a little erratic about how it feels about software, but I do think we've worked through the darkest days of the SaaSpocalypse. I think what we're realizing is there's deterministic and, you know, there's probabilistic and for most systems of record. Now, Adobe's a little different because it's a lot of what it does is create. I mean, it's got all those parts. It's got parts that's very data-driven in its business. It's also got parts that are much more design-oriented software and tools. But it still to me just feels like even though there's been some cool models that do some design type stuff, like people are not abandoning what's considered to be the corporate standards yet. Will that happen at some point? I don't know.

‍

Patrick Moorhead: 

I think Adobe is most at risk of that, given how much the models are putting into this. But, you know, Adobe can be that enterprise Switzerland where, hey, you know, regardless of the next new model that comes up, We can integrate that and keep you clean on the data side.

‍

Daniel Newman: 

The model you have to do a lot more for governance. Excuse me. I'm not bored. I'm just tired. The model you have to do a lot more for governance. if they're going to take this over. But yes, I think the average schmo like you and I will just turn to a model to design something for us very, very quickly, for sure. So, but we're also not necessarily putting out things that are going to be copyright, copywritten, or that are going to need to be trademarked or, you know.

‍

Patrick Moorhead: 

I mean, you know, Daniel, we accidentally used a copyright image on on one of our, you know, two of our posts probably 10 years ago. And we had a horde of lawyers from Getty come after us. It was crazy. Like, you know, threaten lawsuits, or you can just pay us $16,000 and we'll stop.

‍

Daniel Newman: 

Yeah, I've had a few of those over the years. I'm not even sure if they're real, but they're real. Anyways, the net of it is, it's a story to be continued, but I mean, record revenue, massive growth, good quarter. The end has not come yet. Let's just put it that way. And if it is coming, it's not showing up in the actual numbers. Anyhow, Pat, we've done it. We've made it through the show. I've got meetings to go to. You have hikes to take, babies to snuggle, outlandishly expensive dinners to eat. You do you. I've been on this trip with you before. I couldn't afford to go again. But the food was tasty. And I did get a nice message from someone not named Pat, wait, maybe named Pat, but it's not you, that said, I am deeply missed on this trip. So next time, bring a friend, man, bring a friend.

‍

Patrick Moorhead: 

You know, you're always invited.

‍

Daniel Newman: 

But thanks, everybody, for being with us. This is a crispy show. We had some fun here. It was a little different, covered some ground, strong opinions. But that's what we do here. You know, we're not we're not here to feed you the news. We're here to tell you what it means. So thanks for tuning in. Thanks for being part of the Six Five. See you all next week. Bye bye.

‍