IBM's $15B Day, Claude Opus 4.8, & Biggest Earnings Night of Spring 2026 | Ep. 306

The handpicked topics for this week are:

‍

1. Anthropic Releases Claude Opus 4.8: Six Weeks After 4.7

Anthropic dropped Opus 4.8 just six weeks after 4.7, claiming it surpasses GPT-5.5 and Gemini 3.1 Pro on agentic coding, knowledge work, and computer use. Benchmark improvements across the board: agentic coding up from 64.3% to 69.2%, knowledge work from 1753 to 1890, agentic computer use from 82.8% to 83.4%. Three new features ship alongside it: Dynamic Workflows for multi-subagent orchestration inside Claude Code, Effort Control for managing token spend, and mid-task system messages via the API. Fast mode is now 2.5x faster and 3x cheaper. Pat's honest take: what it says on paper is good, particularly on tool triggering and citation precision, but he has lost significant trust in the company and is watching closely.

(The Decode)

‍

2. IBM Commits $10 Billion to Quantum: The Largest Single Quantum Bet in History

IBM announced a $10 billion commitment over five years targeting a large-scale fault-tolerant quantum computer by 2029, landing the same day as the $5 billion Project Lightwell announcement for a single-day IBM strategic commitment of $15 billion. Pat has been calling 2029 to 2031 as the realistic commercial quantum window and calls this the strongest single corporate financial signal yet that the timeline is real. Daniel's framing: IBM wants to be the NVIDIA of quantum, and with a $10 billion commitment, it’s sending a flare to the entire industry that pure-play quantum companies cannot compete at this balance sheet level.

(The Decode)

‍

3. IBM and Red Hat Launch Project Lightwell: $5B to Secure Open-Source Software

IBM and Red Hat committed $5 billion and a global force of 20,000 engineers to secure open-source software for enterprises through frontier agentic AI, anchored by 11 of the largest US and Canadian banks including Bank of America, Goldman Sachs, JPMorgan Chase, Mastercard, and Visa. Pat's read: this is the productization answer to Anthropic Mythos. Mythos found the vulnerabilities. Lightwell is the industrial-scale patching and validation layer enterprises can actually buy on a subscription. Daniel adds that IBM is flexing its engineering talent base as a premium strategic asset, a direct counter to the narrative that AI replaces engineers.

(The Decode)

‍

4. Anthropic Project Glasswing: 23,000 Vulnerabilities Found Across 1,000 OSS Projects

Anthropic's Claude Mythos scanned more than 1,000 widely deployed open-source projects and surfaced approximately 23,000 candidate vulnerabilities, with 1,094 confirmed as critical severity. The Cyber Verification Program now gates the strongest cyber-capable Claude variant behind vetted defenders only. While the tool creates real value, the surface of attack will likely grow as fast as any tool built to defend it.

(The Decode)

‍

5. Anthropic in Talks to Run Claude on Microsoft Maia 200

CNBC and The Information reported Microsoft is in active negotiations to supply Anthropic with its custom Maia 200 inference chip, which would make Anthropic the only frontier lab simultaneously running production workloads on four distinct silicon stacks: NVIDIA, AWS Trainium, Google TPU, and Microsoft Maia. Pat's context: Maia 200 delivers 30% better tokens per dollar than the latest Azure fleet per Satya Nadella, and this deal would be Maia's first major external deployment. Daniel's read: what can be built will be sold right now, and Anthropic chasing every available compute source is simply the structural reality of growing at 80x when you planned for 10x.

(The Decode)

‍

6. The Flip: Is AI CapEx Too Expensive to Earn Its Return?

Pat takes the affirmative. With $725 billion in hyperscaler CapEx tracking for 2026, likely $1 trillion next year, memory has become the choke point making it even more expensive, and open-source models have closed enough of the quality gap for most enterprise tasks that the premium of frontier APIs is increasingly hard to justify. A recent Signal65 white paper shows on-prem payback at 18 months. Daniel's counter: Dell just booked $24 billion in AI orders in a single quarter. Agentforce crossed $1 billion ARR at 169% growth. NVIDIA guided to $91 billion. Only 20% of enterprises are using AI and only 2% of consumers. Both hosts admitted off the flip their notes looked nearly identical.

(The Flip)

‍

7. Micron Crosses $1 Trillion Market Cap

Micron became the 12th US company ever to cross $1 trillion in market cap, surging 19% on May 26th as UBS raised its price target to $1,625, implying a $1.8 trillion market cap. Samsung's Q1 memory ASP jumped 146% year over year. DRAM spot prices spiked 55 to 60% quarter over quarter. Daniel has been pounding this call since sub-$100 and calls it a cycle elongated beyond anything seen in the 27 prior memory cycles, driven by HBM capacity reallocation away from consumer DRAM creating structural shortage.
(Bulls and Bears)

‍

8. Dell Technologies Q1 FY27: The Biggest Enterprise AI Infrastructure Print of 2026

Record $43.8 billion revenue, up 88% year over year, crushing the $35.7 billion consensus by $8 billion. AI-optimized servers at $16.1 billion, up 757% year over year. $24.4 billion in AI orders booked in a single quarter. FY27 AI server revenue guide raised from $50 billion to $60 billion. Non-GAAP EPS of $4.86 beat the $2.96 consensus by 64%. Stock up 18% after hours. Pat's framing: Dell was very clear about what they were going to do. Rack engineering, sales, and service. The basics. And they executed the basics at an extraordinary level while building a special relationship with NVIDIA who views Dell as a market maker for both enterprise and NeoCloud. Daniel's add: play nice and win. Michael Dell navigated the political landscape brilliantly and pulled the entire Dell brand along with him.

(Bulls and Bears)

‍

9. Marvell Technology Q1 FY27: Record Revenue, Data Center at 76% of Mix

Recorded $2.418 billion in revenue, up 28% year over year. Data center at $1.833 billion, up 27% year over year, now 76% of total revenue. Q2 guide of $2.7 billion at midpoint accelerates growth to 35% year over year. Operating cash flow a record $638.8 million. Daniel went on TV and said it’s “written in the stars,” arguing the market had misunderstood this one for too long by conflating its custom AI ASIC story with the full breadth of its connectivity and networking portfolio. Pat's closing: the shorts are eating it now, and the custom AI ASIC versus merchant GPU debate is finally settling into the right answer, which is both in lockstep.

(Bulls and Bears)

‍

10. Salesforce Q1 FY27: Agentforce Crosses $1 Billion ARR

Revenue $11.13 billion, up 13% year over year. Non-GAAP EPS of $3.88 crushed the $3.12 consensus by 24%. Agentforce ARR crossed $1 billion, up 169% year over year, with 28.6 trillion tokens processed, up 152% quarter over quarter. 50% of Agentforce bookings came from existing customers expanding. Daniel flagged the $25 billion accelerated buyback funded by new debt as an interesting signal worth watching. Pat's bottom line: it’s not perfect, but certainly no “SaaSpocalypse” in those numbers.

(Bulls and Bears)

‍

11. Synopsys Q2 FY26: First Full Quarter With Ansys Integrated

Revenue $2.276 billion, up 42% year over year, beating consensus. Non-GAAP EPS of $3.35 beat $3.15. FY26 guide raised to $9.665 billion midpoint. Daniel's framing: every chip runs through Synopsys tools, and the Ansys addition makes it the full-stack co-design platform Jensen Huang keeps talking about. Synopsys is not just the pick and shovel of current AI silicon. It is the pick and shovel of quantum, robotics, and space as well.

(Bulls and Bears)

‍

12. Snowflake Q1 FY27: Strongest Sequential Dollar Growth in Company History

Product revenue $1.33 billion, up 34% year over year, the strongest sequential dollar growth in Snowflake history. Net revenue retention 126%. FY27 product revenue guide raised to $5.84 billion. Natoma acquisition announced for secure agentic enterprise connectivity. New $6 billion multi-year AWS commitment. Daniel's closing: proprietary unique data is the real moat of the agentic era, and that data has to live somewhere. It is going to go to platforms like Snowflake.

(Bulls and Bears)

‍

13. HP Inc. Q2 FY26: Eight Straight Quarters of Growth
    With AI PCs at 44% of Shipments Revenue $14.4 billion, up 9% year over year, the company marks its eighth consecutive quarter of top-line growth. Non-GAAP EPS of $0.86 beat the prior guide. Personal Systems at $10.2 billion, up 13%, with 30% operating profit growth. AI PCs jumped from 35% to 44% of shipments quarter over quarter, with HP guiding to 60 to 70% next fiscal year. FY26 EPS guide raised. Pat's note: They still need a permanent CEO, which would help investors sleep better at night. Daniel's add: the real explosive moment for device companies comes when AI moves to the edge, and enterprises shift from expensive frontier model consumption to on-device inference. (Bulls and Bears)

‍

14. Everpure Q1 FY27: Record Revenue, Rebrand Complete

Record revenue of $1.1 billion, up 35% year over year. Product revenue $577 million, up 55%. Subscription ARR at $2 billion. FY27 guide raised to $4.41 to $4.51 billion. Pure Storage officially completed its rebrand to Everpure. Daniel's emerging thesis: the agentic era has focused enormous attention on memory and compute, but after the inference runs, the data has to sit somewhere. Storage has not seen its full inflection yet, and Everpure is well-positioned when that wave arrives.

(Bulls and Bears)

Watch the full video at sixfivemedia.com, and be sure to subscribe to our YouTube channel so you never miss an episode.

‍Disclaimer: The Six Five Pod is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and reference share prices, but nothing discussed should be taken as investment advice. We are not investment advisors.

PATRICK MOORHEAD:
Do we know what we're talking about, Moorhead? We absolutely know what we're talking about. Welcome back to the Six Five Pod. It is great to be back in the chair. Dan and I both got workouts today. We're doing well, but you know, it's just another week. Dan buys another company and it's his birthday. You know, some people get gifts for their birthday. They receive gifts, but Dan bought himself a birthday gift. Daniel, congratulations. Tell us a little bit about what ETR does.

DANIEL NEWMAN: 

Yeah, thanks, Pat. I love a good self promotional segment to start off episode was is 306 and I did have a birthday and I'm pumped. To announce that we signed a definitive agreement should close this quarter to acquire. ETR, Enterprise Technology Research. It is one of Wall Street's preferred signals, one of the most unique proprietary datasets. They track 10,000 IT buyers. They've been tracking them for 15 years, $2 trillion in spend across a number of categories, cybersecurity, software, SaaS, AI platforms. They're the ones that know before the print what's going to happen. This was a really unique dataset in an era. where all the models are increasingly becoming capable of doing all the things. As analysts, I think what we know that no one else knows is going to be the most important differentiator that we have. So big bet for us. This company serves primarily Wall Street, so its clients are some of the largest institutions. Names, household names you know, and some of the largest investors you know use this data. So it definitely rounded off our portfolio. I've long enjoyed through what we do with Futurum Equities and what we've done in partnership with Defiance ETF and all the work. You and I both do this, talking to the financial markets through CNBC and Fox Business and Bloomberg. But this deeply integrates our day-to-day work with Wall Street and couldn't be more excited. So big bets, some big risk. But I've realized, Pat, I started this company on my birthday. I made a major partner decision on my birthday. And I acquired ETR on my birthday. So, I apparently like to make major life decisions around my birthday. So, thanks for noticing. Thanks for calling it out. 45 and aspiring to be 

jacked. That is my goal.

PATRICK MOORHEAD: 

Yeah. Congratulations, Bestie. Yeah. So, you know, Dan acquired two or three companies a year. I've had my third surgery on Tuesday. That is I'm hoping this time it fixes my sinuses and hopefully I'll be less miserable when I'm on the road, but I'm not making any guarantees. Hey, we have an amazing show for you. I mean, we have so much stuff. I mean, we're talking, you know, Huawei logic folding. We've got a new Opus model out there. We've got IBM making $10 billion in quantum commitments, got IBM and Red Hat coming together on OSS security. We're going to be talking a little bit about the Anthropic Project Glasswing. What have they truly dug up here? And, shocking nobody, is Anthropic talks to run Claude and Microsoft Maya 200. Imagine that. We have a ton of earnings. We've got Dell, we've got Marvell, we've got Salesforce, Synopsys, Snowflake, HP, we've got a great show. I do want to do a hat tip. There may not have been many big events this week, but next week, it's crazy. I'll be at Cisco Live. We've got NVIDIA GTC Taipei. We've got the Signal team out there, analysts from Futurum and more insights. We've got Microsoft Build in San Francisco. We've got Snowflake's Big Tent event as well. I've got analysts going to all of those shows. So Daniel, let's dive into the decode where we're going to parse the signal through the noise, from the noise. Let's hit it. All right, we have got Anthropic releases, Claude Opus 4.8. Daniel, is it good? Does it fix things? Is it a nice try? What do you see in there, bestie?

DANIEL NEWMAN: 

Look, the pace of model updates is so fast. And for most of us, it's, 

you know, remember it was like a yearly cadence and then like maybe six months and now we're seeing new. And by the way, now we get four eight, but, you know, we're getting whispers of a broader release of mythos, you know, the model that was going to take over society and allow every Patrick Moorhead to become a deep, dark black hat. stealing secrets from Russia and China, but in all serious, you know, it's, let's talk a little bit about just some of the benchmarks, right? You know, their claims are that it surpasses GPT 5.5 and 3.1 Pro, and these are on the synthetic AI-centric benchmarks for autonomous coding, which of course is very important, also financial analysis and computer usage. Now let's talk about deltas on the opus versus 4.7. And 4.7 was sort of a I would call it like a mixed release. I think it was very good, but it really brought to the surface some serious issues with Anthropic. It's compute capacity. I know it was 4.7 was sort of where you turned your back on Anthropic for a period of time, got very frustrated with it. But it's basically saying that on the benchmarks, it jumped from 64.3% to 69.2% on knowledge work. It jumped from 54.7% to 57.9% on agentic computer use. It jumped a little bit, 82.8 to 83.4 on agentic financial analysis. And on, what was the last one? Sorry, agentic computer use went, oh gosh, I got this all wrong. I'm reading them out of order. So we might need to go back and do that one again. Agentic coding, 64 to 69. Knowledge work on their scoreboard went up from 1753 to 1890, multidisciplinary reasoning with tools, 54.7 to 57.9, agentic computer use, 82.8 to 83.4, and agentic financial analysis, 51.5 to 53.9. So improvements across the board. It has new dynamic workflows, effort control, which is obviously how hard you have the model working. And whether or not you want a random $500 million unexpected bill that somebody got from not putting the right throttles in place. Personally, Pat, I mean, I'm just getting off the, I'm just getting to use, you know, I'm just starting to put it into play. You know, my early read is the first couple of things I've asked it to do have looked good, but I think you and I would agree that you got to play with these things and kind of beat on these things. Then, of course, it depends on the work you're doing with these things. And I'm also very interested in watching the API costs. You know, we run it on an API, so we're watching those costs very closely. And then, of course, watching how quickly we burned through on our personal pro and max pro credits, how quickly we burned through them. So sorry for all the stats. But I mean, I think as we see models start jumping out as quickly as they are, that's going to start to become it's going to be a benchmarking thing, because it's like, what is the real yesterday to today improvements that we can expect and at what?

PATRICK MOORHEAD: 

Yeah, so what I see on paper is good, right? Big thing for me is tool triggering and citation precision, because most of what I'm doing is high volume systems, MCP dependent and number dependent. So those supposedly will help me with a lot of the workflows. I did move some stuff over to Codex. and which has been an eye-opener. Ironically, I find CodeX a lot easier than Cloud Code. And using CodeX versus, I'll call it, you know, web chat GPT didn't enable a lot of easy MCP connections and tool calls. So, you know, we will see. What it says it's doing is better, but I've lost a lot of trust with the company. So Daniel, let's shift to quantum computing. IBM commits $10 billion over five years to quantum computing. I believe this is the largest single investment. Now we have to also recognize the public companies out there that are making investments that have some pretty high market caps, but I believe this is the largest one to date. And, you know, IBM really took an early position, a strong position in quantum. They did what a lot of people weren't. They actually put a 10-year roadmap out there on the research side, which is absolutely unheard of. Like, what research groups put a 10-year roadmap? And by the way, they actually stuck to it. So that, to me, gives a lot of credibility. And IBM right now is they're engaged in taking it from a research projects and experimentation to delivering real value and they have a benchmarking  formulation, and I don't want to call it a campaign because that sounds marketing, but a structure where their clients can show real value. We saw INQ leaning into that a couple of weeks later as well. So the good news is it's a really good sign for the industry that we're moving quantum from science project to delivering real enterprise value. This also gets on to the government taking equity stake in the quantum foundry. And that's called Andron that was released, I believe, last week. So it's good to see real money on the table. And by the way, these type of signals are also to signal to competitors and even VCs that you need to you need to reach a certain bar. Do you really want to do that? VC and investors to try to keep up with us. So, you know, 2029 is going to be the big, I believe that is going to be the big inflection point that IBM is putting out there to, you know, they already have 90 quantum systems, 325 organizations using them. But 2029 is where we get to big commercial value.

DANIEL NEWMAN: 

Yeah, I mean, look, IBM just had a $15 billion day. We'll talk about all of it, but this is probably the most aggressive thing we've seen come out of IBM corporate headquarter. And, you know, I think they're recognizing, you know, clearly they have a role to play in AI, but I think they have the chance to, I think they believe they have the chance to be the NVIDIA of quantum.

PATRICK MOORHEAD: 

Yes.

DANIEL NEWMAN: 

And so knowing that's the next big planning cycle after AI or alongside, because I don't think AI is going to stop. But, you know, I think one strong signal they sent was look, quantum companies are small, you know, relatively speaking, you know, pure play. You know, we both advise IQ, and I think we both think highly of that company. But in terms of a $10 billion investment, that's not even a possibility. And IQ is bigger than the Rigettis and D-Waves and all the others by a lot. So just being able to put that much in is really kind of sending a flare out to the entire industry and saying, we can spend at a rate you can't even compete with. And I also think this could start to argue that we're sort of spawning an AI capex cycle like for quantum. It's going to be much smaller. But I think what came out of the White House this week, plus this, is starting to signal that the ramp is happening, right? So I think those kind of two things and I think, you know, you made a call, you said 29 to 31 will be the call for when commercial quantum, I've said similar things. I think this is the strongest signal that's come out of the market yet that that timeline is looking increasingly to be the real commercial inflection for quantum.

PATRICK MOORHEAD: 

Yeah, good stuff. And again, it is definitely the next wave. And, you know, the Adder, you know, IBM wants to be the NVIDIA. I think IBM actually wants to be the NVIDIA and the AWS and the Azure and the actual service layer that goes in software.

DANIEL NEWMAN: 

Yeah, you're right. All the above. I guess that's the NVIDIA. But yeah, good call.

PATRICK MOORHEAD: 

So hey, another IBM investment here, that's IBM and Red Hat. Red Hat is part of IBM, $5 billion, 20,000 engineer effort to secure open source software for enterprise. You know, we got a little A little taste of it with Glasswing and Mythos from Anthropic. Maybe it's fake, maybe it's not, but we are finding a lot more holes with AI. Dan, is this even important?

DANIEL NEWMAN: 

Yeah. So as I mentioned, the 15 billion, so 5 billion, 20,000 engineers, frontier agentic AI, secure open source, and it's been anchored by basically the 11 largest US and Canadian banks. So I'd say that's relevant. They all signed into this. It's open source security clearinghouse. And it's a strong, I see it as a pretty strong signal. You know this is you know we talked about them moving into quantum. I mean I think Lightwell is a is a credibility creator for IBM. You know what is their role in AI. We know they were one of the first enterprise governed and scaled platforms. But I think at times people kind of forget. But in the end, enterprises increasingly are going to be depending on a solution that's got governance, compliance, rails. And of course, for AI to work at scale in industries that are super highly regulated, they're going to need this kind of partnership. And so I think it's significant. And I think IBM is leaning into their talent base. You know, we heard walkbacks this week from Sam Altman and, you know, from Dario Amadei about their kind of all jobs are going to die. And I think IBM's a bit making a flex here with their with their engineering team saying, you know, the engineering with engineers with AI is exponentially more effective than, you know, non engineers trying to become engineers using AI. So I think it's a you know, it's a big bet. It's got the right alignment. It's got the right partners. You know, I think, again, remind remind me off the top of my head, but there's a there's a few other partners involved here. You've got the biggest security names, CrowdStrike, Paolo, Microsoft. And so I think. Oh, anyways, I'm reading through my notes here. So this anyways, I got to totally pull that note back there, guys. This is where AI has failed me. Um, so it's not, it's not, it's Okay, sorry. Lightwell is going to genuinely, it's not sitting next to, it's genuinely making IBM security, which has been a little bit quiet, more competitive next to the big names, the CrowdStrikes, the Palos, the Wizards, and the Microsoft Securities. And I think that playing here and playing in open source is an area that IBM can strongly defend. So that was word soup for me, but hopefully with a little bit of editing, producers, we can make that suck less.

PATRICK MOORHEAD: 

Yeah, you know, I see this as the productization answer to anthropic mythos, right? Discovery. I mean, if mythos is finding it, this is actually fixing it, right? And, you know, the 20,000 engineer commitment isn't like a new cost. right? It's being positioned as this as a strategic asset with it is. One thing I really like too, is that it it, you know, it's not just, you know, some some announcement for tech companies, right? You have people like Bank of America, BNY, Citi, Goldman Sachs, JPM, MasterCard, Morgan Stanley, and a bunch of other financial institutions around the world who are in there who have signed up. So I think, you know, this is, you know, we saw the risk with Meethos and what Chad GPT, I forget what their special security sprinkles are called. And this is the actual solution here. So let's go into another, speaking of Methos, Anthropic Project, Last Wing, they found 23,000. This is kind of the result of the initiative that I think started two or three, no, actually about a month ago on bringing this out. And I think, you know, it's funny. There's I think the only debate is not whether AI can find these issues. It's can it do it easier? And can it do it faster? And is the planet about to collapse like Dario talks about all of a sudden. So I think the scale of the discovery is, again, I can't tell you whether 23,000 candidates, 6,000 high critical, 1,000 confirmed as severity is high or low in retrospect, but it's a lot. of issues out there. It is interesting how the limited access to this, I'm wondering, you know, is this truly a risk? I mean, what about the people who didn't get access to this out there? And there is some talk about everybody getting access to it in two weeks. What about the people who didn't get access to it? Are they just shit out of luck? And they need to, they need to, to, you know, Are they just going to be completely hacked out there? I guess we will see. So I think the only question left is, again, who gets access to the model and on what terms do they get access?

DANIEL NEWMAN: 

Yeah, I mean, this is interesting because the cost of doing this kind of penetration testing is massive for most companies. And to your point, such a small subset got access. Does this augment that industry? Does it replace this place? It's really valuable, though. I mean, any company running software to have the ability to very quickly understand their landscape and get a handle on what they need to be protecting. It's a huge risk. And with the proliferation of this technology, the risk on the other side of becoming a victim has only grown. And it's kind of increasingly becoming not the if, but the when. So it is encouraging to think that these tools could help and to create scale, Pat. But I think you make a great point. There's the haves and have-nots here. is going to be the demarcation between those haves and have-nots? And what is the conversation that really needs to be had around AI safety? And how do we make it less theoretical and more operational? Because security has been a longtime problem. It costs fortune to society. And if we can help solve that. But I have genuine beliefs, Pat, that there's probably high likelihood that as AI proliferates, the surface of attack will grow as fast as any tool we're going to build to defend it. So I just don't know that we're it feels very good to tell the story. We'll see in practicality what happens to real breach numbers as this tech continues to grow.

PATRICK MOORHEAD: 

Hey, let's move on to another topic, kind of a reminder. I feel like we've had this discussion with different companies and different chips and different model makers a few different times here. But CNBC and the information reported that Microsoft is in active negotiation to supply Anthropic with custom Maya 200 inference chip. And that would be the first big external deal that I'm aware of. I do not believe that Maya and even OpenAI were strongly referenced with a lot of data. But this would be a pretty big if true.

DANIEL NEWMAN: 

Well, I feel like this is going to be another where you and I just say what can be built will be sold. Clearly, Anthropics chasing compute. They took over Colossus from SpaceX or XAI, because they needed more instantaneous compute to support their rapid growth. Look, all these companies making custom chips have Um, a certain amount of the, the wafer capacity that they've reserved, uh, in most cases from TSMC. Uh, and so right now the companies that are delivering all the inference computer, trying to get their arms around as much as possible. I mean, Maya is designed to provide some, some strong, um, inference numbers. Um, but. I think it's being driven just by the crazy growth of demand, Pat. I don't think it's a I don't necessarily know that this is a sign that it's a great part yet. I think that's going to be found out in time with testing and with utilization. But heck, I mean, look, look at all the news this week. I mean, we saw what happened with Cerebris. Qualcomm is apparently doing a bunch of silicon for Chinese and American companies now. I mean, basically, I just I haven't seen a company yet that's announced a chip. They can't sell it. which I think is more implement, implies more to the demand, but look, this is what's going to get some of these companies off the ground. And as we know, multiple generations of any of these chips, they get better and better with usually get better and better with more turns. So they can get in market and create success. That'll generally be the foundation of getting to my three and my four. And I imagine the, the, the specs will continue to get better.

PATRICK MOORHEAD: 

Yeah, this is kind of the, uh, kind of the law, Patrick's law of always, which is kind of a broader description of what you're talking about, you know, and there's a big, big trough between having a conversation and making commitments. I think the question would be is, does Microsoft have the capacity to do this? I think the Maya 200, the most recent one, had some pretty ostentatious four-bit performance numbers that really put it in a class on its own, kind of reminded me a little bit more of the TPU than, let's say, Tranium. Microsoft is certainly putting up the CapEx. The question is, is it locked down to OpenAI or not? It certainly is nice that if the OpenAI growth, which it looks like it has reduced significantly and is being fulfilled by Anthropic, run it at Microsoft. I wouldn't be surprised if we see this collaboration going on, particularly with OpenAI growth kind of peaking. So if it's true, it means that Anthropic that started off with Tranium now would run on NVIDIA, TPU, and maybe Silicon, right? No loyalty for vendors and just doing the right thing, which is, you know, leveraging vendors across each other. All right, Daniel, let's move into the flip here. And I think the thing that we want to debate in our simulated debate says CapEx is so expensive. It forces a shift to open source and smaller models. And essentially is AI CapEx becoming too expensive to earn its return. Let's flip the newly revised coin with SkinnyDan and NoBeard. Look at those two happy, happy, happy people.

DANIEL NEWMAN: 

So it looks like I got… So you're saying that it has outrun ROI. It's gotten too expensive?

PATRICK MOORHEAD: 

Yeah, I'm saying, yeah, CapEx, yes, has run fast. You're a bubble bear.

DANIEL NEWMAN: 

That's a bubble bear.

PATRICK MOORHEAD: 

Yeah. I mean, this one, again, I'm so tired of winning, but, you know, let's let's go here. So so hyperscaler CapEx right now is tracking to be seven hundred twenty five billion dollars this year. And I think we're looking at at likely a trillion dollars next year. And at that scale, kind of every CFO downstream is looking for ways to not pay frontier model prices. Memory has become the choke point, making it even even more expensive. So what we think is expensive in 2026, hold my beer looking into 2027. And open way models, they have not closed enough for the quality gap for most enterprise tasks. And if we think we're excited about all of the growth now, these are pretty much all frontier models. OSS is nice. And I think it has a place in the future. But, you know, quite frankly, you know, it hasn't closed that quality gap. And I think, you know, sovereign regulated buyers have a non-cost reason to point the same direction, right? Data residency, auditability. You know what? I'm arguing the wrong one. Now I'm confused.

DANIEL NEWMAN: 

No, you're saying AI CapEx is too expensive to earn its return. Which one are you doing?

PATRICK MOORHEAD: 

Okay, you're right. I am.

DANIEL NEWMAN: 

And we are right one or not. You need to start it over.

PATRICK MOORHEAD: 

No, we're a hot, uh, uh, you know, we're, we're a hot mess. Okay. Forget what I said about open way models, right? Open way, open way models have closed enough, the quality gap. Okay. For most enterprise tasks. Uh, we saw it at Dell tech world, uh, a lot of discussions about what really needs to be done. And, you know, when I talked to a lot of the Dell customers there, you know, I guess two or three of the 5,000, they pretty much said that, uh, Hey, we don't need. B300 DLCs to do what we need to do to extract value out of there. And we are assessing and using a lot of OSS models. And if you look at the OSS models came out from recently from ChatGPT, they are pretty darn good. And I think that we already see the non-regulated buyers are using a lot of the Chinese models over there that are pretty good. So the models are good enough and the price deltas versus frontier APIs is large. In fact, you can look at a Signal 65 white paper that we did that clearly showed that the payback was around 18 months and you can save a boatload of money by doing on-prem. Sovereign is just another example of this. Two years ago it was marketing speak and this year it's really driving real revenue. You can look at the UAE, you can look at Saudi Arabia. with Humane, plus a lot of these countries who want their own stand-up AI clouds are building that out, and they're quite frankly not using these frontier models. I think the final comment is that the hyperscalers are a negative cash flow, which I think that I think even we have to question that entire business model of them getting back to profitability. So I rest my case. Daniel, go for it.

DANIEL NEWMAN: 

Sorry, I didn't know. I think there'd be a time limit for how long you can ramble on incorrectly about anything. Look. ROI is showing up. I mean, I could just drop the mic and just say Dell booking $24 billion in enterprise and NeoCloud AI orders. And your points about bringing stuff on-prem, that's not an argument against them not showing up. That's just an argument that we have so much demand, and we have smart CIOs that are basically saying it's going to be an and. Everything about this era is abundance. It's not one or the other. It's going to be one and the other. They're going to use Frontier. models in the cloud, and they're going to use smaller models on-prem, and they're going to do it all because there's going to be so much capacity required and so much demand for these tools that it's all going to begin to show up in ROI. But look, I mean, last quarter, AgentForce went a billion of ARR, 169% growth. We'll talk about a bunch of these in bulls and bears, but look at what Snowflake did. Absolute explosion. a lot of growth on a quarter-over-quarter basis at crushing it. I know it's been a few weeks since the announcement, but NVIDIA's guide to $91 billion, a lot of that revenue, which by the way, was not from the hyperscalers because they realize that we aren't spending enough on CapEx. So what's the next three years look like? Are we going to spend less on CapEx or are we going to spend more? We're going to spend more. And you know why? Because the smartest people on the planet all understand the existential risk of getting this wrong. meta. Sure, Mark gets it. If he doesn't need all this compute, he can create another public cloud and sell it because somebody will need it. We've seen this with the fungibility of every deal that has gone wrong throughout this entire process. Every bubble that people thought blew up, guess what? There was somebody there instantaneously to soak up that capacity. Colossus, the Microsoft deal from Oracle, People need more capacity. We're not actually spending enough. The data centers aren't going up fast enough, and we're seeing growth coming even in the SaaS space where everybody thought it was all over because it's not or, it's and. It's going to be SaaS and AI. It's going to be Cloud and-prem. The future is going to mean big economic growth, trillions of dollars of economic value added because of AI. Guess what? Jobs aren't all going away either. The economy is growing. Life is good. Thanks for attending my sermon.

PATRICK MOORHEAD: 

That's pretty good, Daniel. You're always pretty good at this stuff. Were you on the debate team or the chess team in high school?

DANIEL NEWMAN: 

Oh, I mean, you know, I was a little chub-a-lub. You know, it wasn't until I discovered peptides that I became… I don't know.

PATRICK MOORHEAD: 

I saw pictures of you. You weren't a fatty, but you did have really long red hair.

DANIEL NEWMAN: 

It wasn't red, it was like auburn. Like Trout would not respect me in the redhead community. Like the true redheads would not accept me. But you're right. The blondes and brunettes kind of looked at me as the guy, because you know when I put sun in my hair? You know, like do you ever do that? Like to bleach your hair? Remember sun in? You'd spray it in?

PATRICK MOORHEAD: 

Just look at my hair right now.

DANIEL NEWMAN: 

No, you never did that. Okay. So like when I was coming up, cause I'm a lot younger than you, but when I was coming up, there was this thing like you'd spray in your hair and it supposedly would make the sun. So when you're, when you have red in your hair, you can't get getting blonde is really hard. Just like, so what mine would turn really red. That's the point. Like I'd get super red coloring, trying to go blonde. Um, I think I went bald in college when I actually had them put true bleach on my head so I could have a backstreet boy haircut. Remember when I wanted the tips frosted? They put bleach on your hair and stripped all the color off. I think I burned my scalp and started my balding process early. Anyway, good story, anyone who wants to know me. I had hair in college. That's about the last time I had it, in case you were curious if I ever had hair. It was important, right? It was good hair.

PATRICK MOORHEAD: 

By the way, we had Sonnen in my generation. It was just the blonde girls in high school who did that, not the guys.

DANIEL NEWMAN: 

Well, like I told you, my generation is different. I'm a millennial. Even if I have a lot of Gen X tendencies.

PATRICK MOORHEAD: 

You do.

DANIEL NEWMAN: 

You wish you were in Gen X. I feel like I really don't feel like I belong in my generation. The millennials are too soft.

PATRICK MOORHEAD: 

I mean, I'm just glad you are where you are. All right, Daniel, let's jump into bulls and bears. We have so many to get through. We've got 18 minutes. We're going to plow through these. Let's dive in. All right, Micron, $1 trillion market cap. Daniel, is this just simple supply and demand? We can jack up our prices and crush it in earnings?

DANIEL NEWMAN: 

Well, I mean, the margins are what, over 80%? You saw street high price targets coming up over $1,600. And I just wanna say that I have been, wait, is this a victory lap moment? I think it's a victory lap moment. I have been pounding this one since it was sub $100. It was so obvious to me. should have been obvious to everybody. Unlike the CPU one, which I can admit I didn't get quite right, the memory constraint was super obvious. The difference here is going to be the length of the cycle. So I won't say it's an infinite cycle, because I think some people will tell me I'm wrong about that, but I do think we are seeing a cycle that is elongated beyond anything anyone who's been through the 27 memory cycles has ever seen. They can continue, and the fact is right now the capacity is so constrained, and it's really the reason that it doesn't matter whether it's the input companies, the storage companies, it's memory and it's storage, flash, that's creating all the constraint. So their pricing power is extremely high. and it's going to last for a long time. And so it's not just Micron, by the way. It's SK, it's Samsung. I think the Korean stock market doubled this year. And I think half of, was it Samsung, like trying to go on strike? Like these people realize the asymmetric power they have right now has never looked anything like this right now. Memory still holds the key. Exciting for Micron. You know, how long this cycle lasts, it's probably at least half a decade.

PATRICK MOORHEAD: 

Wow. Five years. You know, whenever I hear like it was an easy one, it was a no brainer. I wonder, did that person like mortgage their house and, you know, pour all their money into that? Or did they keep it into hymns or something like that?

DANIEL NEWMAN: 

So what idiot would do that?

PATRICK MOORHEAD: 

Yeah. Well, what idiot would, by the way, I made a really bad mistake. I just stumbled across in my E-Trade account, how many AMD shares of stock I had at one point.

DANIEL NEWMAN: 

Is there a loser lap here we can do for the victory?

PATRICK MOORHEAD: 

Yeah, I don't know. I don't even know if I want to even even say. You should share this.

DANIEL NEWMAN: 

I think I think I think this is our audience.

PATRICK MOORHEAD: 

Okay, here's that. I did the calc and I was nine figures. um, of, of, of AMD stock valuation had I held all my, all my shares. So you left in what year? I left, uh, 15 years ago.

DANIEL NEWMAN: 

Okay. So you left in 2011.

PATRICK MOORHEAD: 

Yeah.

DANIEL NEWMAN: 

If you'd done nothing, right. If you'd literally just left and let that money sit there, You'd be a centimillionaire from that.

PATRICK MOORHEAD: 

Yes.

DANIEL NEWMAN: 

Alone. Not that you're not already, but from that alone, you'd be.

PATRICK MOORHEAD: 

Well, I'm definitely not. Not that. I mean, I can stretch the value of my company and my joint ventures, but, you know, I'm definitely, definitely, definitely not there. So I mean, I was trying.

DANIEL NEWMAN: 

I didn't want to. I was leaving a little mystery for everyone out there. So, so Pat's kind of race is what he's saying. Yeah.

PATRICK MOORHEAD: 

No, it's the, uh, what did they say on, um, God, what was that show?

DANIEL NEWMAN: 

It was succession where they, you know, you're, you're worth exactly the wrong number where you have too much, but not enough. I remember that.

PATRICK MOORHEAD: 

Back on micron. Yeah, I did not. I did not put all my chips on micron mortgage my house, so my dogs and we're doing like that. Maybe I should have but I mean, The market's following whatever the choke point is of the day. I do like the spectacular 80% gross margin loss to 80% gross margin on the plus side. And it's paybacks. What I'm kind of surprised about so far, is we're not seeing gigantic commitments to increase CapEx as I would expect. The other thing is that the market will respond in one way, shape or another with different architectures that lean less on the lower yield, higher price memory. And you know we're starting to see that, right? HBM, we've got GDDR7 with Tense Torrent, we've got LPDDR systems, you know, even inside part of what NVIDIA is doing. Yeah, it started and I think we're going to see a lot of compression algorithms and techniques being being brought out as well. You know, the one that got a lot of a lot of bank was Google. And by the way, compression is not new, right? And there's always a somewhat of a penalty you pay with compression, but I guarantee you, the market will react. The market reacted with the need for compute. It disaggregated. Right. I mean, remember, I remember people calling me a fool and maybe this isn't it's not necessarily a victory lap. But, you know, I said forever that, you know, there was a chain of compute. And each of them did different things better. CPU, GPU, ASIC, FPGA. And when I said AI ASIC, people thought I was an idiot. And it ends up I'm not an idiot, where we could see on a unit basis as many ASICs as there are GPU units, albeit GPUs get more done than ASICs. Yeah.

DANIEL NEWMAN: 

We got like 17 more. Let's let's go. Okay, we're out my next topic is like the holy freaking smokes one, right? Like, good.

PATRICK MOORHEAD: 

No, it is. Yeah, let's get let's go into Dell. And you know, I tried to come up. Let's go into Dell. Tech here. I tried to come up with a clever tweet right after this. But I just couldn't that it was so good. They literally beat on on everything. I mean, And not only do they beat, they beat in a commanding way, they beat on revenue, EPS, AI servers. all three guides, actually five guides for the year and also for the quarter. They blew away all segment performance and they blew away pretty much all financials, including cash. And I think what I wanna spend my time on is not the why, the why is easy. It's AI optimized servers. Also they had 17% growth on the CSG side. It was only supposed to be up low single digits, right? That's easy to explain a lot of pricing and some pull-ins and Dell just being amazing on supply chain. But it's really about AI-optimized servers where they're absolutely crushing it. Dell is very strong in the Neo clouds. That's where they're driving a lot of this business. They also have 5,000 enterprise customers that went up by 1,000, I believe, in the quarter. We learned that at Dell Tech World. What I'm going to spend my time on is just the awesomeness of going for it. Dell was very clear when they set out what they were going to do. And it was all about rack engineering. sales and service. It was the basics, right? And it turns out that the basics they did very well at and they captured a gigantic part of this market. I know it looks like everybody's partnering with NVIDIA, but I do believe, I know that Dell has a very special relationship with NVIDIA and NVIDIA looks at them as a market maker. for not only the enterprise, but also for for Neo clouds, they've had tremendous, tremendous execution. I know, you know, as we sat in the meetings at DTW, it's like, my gosh, you guys aren't going to acquire a company like, like, come on, like, what? what's going on. I think their plan of, of, you know, knows the grindstone and Jeff Clark's badass, um, you know, leaning in. I mean, Jeff Clark basically came in, took over the client division, uh, is the chief AI guy at the company. And, um, he's operating command and control, uh, along with, uh, with Michael.

DANIEL NEWMAN: 

So listen to our, uh, recent, uh, six, five from Delta with both. We had one with Jeff and Mike.

PATRICK MOORHEAD: 

Yeah, absolutely. So yeah, absolutely crushing it.

DANIEL NEWMAN: 

I mean, yeah, not a lot to add. I think play nice and win is apparently what's happening here. Yeah. You know, it doesn't hurt to have the president of the United States give you a shout out and tell people to buy Adele. Michael did a very good job, by the way, of navigating kind of the political landscape. I mean, he got involved in the invest America. You know, he's he's a great philanthropist, but he's he's been very visible the last few years and he's pulling the entire Dell aura along with him. And of course, you know, if you look, I mean, we saw those results were good. Supermicro had pretty good results. They're being pulled up by this, by the way. I expect HPE will actually have really good results. But I mean, Dell is just it is a bit of a league of its own right now. Its size and its scale in this particular category, they have just absolutely ramped at undeniable pace and it kind of led the whole AI factory model. Um, and, uh, you know, congratulations. I mean, dude, like, I don't know, like you said, you tried to come up with a tweet. I don't know what to say. It's just brilliant. I mean, a brilliant, brilliant.

PATRICK MOORHEAD: 

I mean, at this point, are they just showing off or something? You know, that, that's what went through my head.

DANIEL NEWMAN: 

Yeah, absolutely. So let's hit the next one. Yeah.

PATRICK MOORHEAD: 

It's at Marvell. Very good performance, Daniel.

DANIEL NEWMAN: 

Yeah, I went on TV yesterday talked about I said it was written in the stars. I said the market had just misunderstood this one for way too long. They basically the price action had gone along with the rumors of its ASIC business, specifically related mostly to Tranium, but its other wins on custom AI chips, which is a meaningful part of its business. But when you look across its portfolio, which is a huge connectivity portfolio, and again, as we know, storage related, networking related, it's not all just custom AI chips. Well Marvel has all of it. And so it was able to you know one you know it was able to show that it's its data center business is growing like crazy. I think it's 76 percent of the mix now at 28 percent revenue growth. It accelerates in the quarter to 35 percent growth. So I mean, that's obviously really incredible. It showed some longer term figures into the next couple of years, and that growth continued to stay very, very strong. It's just basically hitting it on all cylinders. And again, this whole debate, ASIC versus GPU, I'm going to continue to beat the drum. It never was either, or it always was and. It's going to be the two in lockstep. And by the way, the more custom silicon they do sell, the more attached content they sell. And I think people are just starting to figure out that as they continue to get the wins and they continue to see these next generation of chips go out, including Tranium, that they're going to sell a whole bunch of other Marvell content. That makes the numbers grow. And it looks really good for the company's prospects.

PATRICK MOORHEAD: 

Yeah, this is one of these. I always saw the real story and knew at some point the shorts would eat shit. And they are at this point. And what I think is finally settling in is all the rumors out of Taiwan. Most of them ended up to be BS. Even the discussion about, oh, they're surties isn't competitive. you know, they're going to crash and burn, then that ended up to be an absolute nothing, nothing burger. And then on the optics side, you look at the investments that they've made, it truly is, you know, has put them in a leadership, leadership position. The one thing on the XPU, I think people finally get is it's not all or nothing. right? There's different ways that Marvell can do ASICs. They can do full, you know, customer does the design, they do the back end all the way to the manufacturing and getting supply at TSMC. If there's potentially another design company doing the back end, they still have IP. that goes into that XPU. So there's multiple ways they can monetize it. And I think people are finally getting their arms around that. you know, data center at 76% of revenue, right? I mean, they're clearly, clearly, you know, what the company has said that it is, and that is the data center, the data center company, you know, it's funny, a good reaction, even though they only beat by a penny, but nobody cared. Right?

DANIEL NEWMAN: 

Oh, this is all about the growth, and their continued trust in their grocery. I'm glad you called that out, though, because the actual EPS and net income might have been the only blemish on the line. It wasn't a blemish. It just wasn't. It wasn't a blowout.

PATRICK MOORHEAD: 

Yeah, that's right. That's absolutely right.

DANIEL NEWMAN: I believe Matt Murphy will be joining us. CEO Matt Murphy will be joining us for Six Five Summit this year.

PATRICK MOORHEAD: 

Yeah, looking forward to that.

DANIEL NEWMAN: 

We've talked to their COO Chris Koopmans a few times on 6.5 recently, but Matt has not been on for a while.

PATRICK MOORHEAD: 

So looking forward to that. Yeah, he was one of the early ones on the show, which was great. Great to have him back. All right. Let's hit Salesforce. Is the SaaSpocalypse true here? A bunch of beats on the results. They beat on revenue, they non-GAAP EPS beat, and they met on the revenue guide, which I think spooked people. But I think first and foremost, investors are looking at the percentage of revenue that's coming from agent force. I think that that is why that is what is is laying up on the stock. I mean, a lot of financial metrics that look good, a lot of color as well. But, you know, as as we've seen in this environment, it's just, you know, it's not it's not good enough if there's anything to be able to pick on.

DANIEL NEWMAN: 

Yeah, I thought it was strong. I mean, look, I mentioned it in the flip. Did you freeze?

PATRICK MOORHEAD: 

I'm still here.

DANIEL NEWMAN: 

OK, you froze. I didn't know. You were totally frozen for a minute there.

PATRICK MOORHEAD: Y

ou start shutting down some things.

DANIEL NEWMAN: 

OK. I'm going to jump in here. Yeah, I mean, look, I mentioned it in the flip, but AgentForce crossed a billion in ARR, 169% growth year over year. And I like the NRR expansion. Half of the revenue is coming from current customers. That's how a company that has that kind of moat gets growth, is they get more products being purchased by their current customers. So that's great. The utilization is up, you know, 28.6 trillion uh tokens 152 growth in the quarter means they're you know they got a nice flywheel of consumption turning um you know i i worry a little bit or i don't worry i guess the i'm curious that they use debt to fund a buyback kind of an interesting strategy um i lost pat they used uh debt to fund a buyback and you know while raising guidance. I'm not sure what the signal was there exactly. It could be confidence, and it could be that their stock is undervalued, and they want to make sure that they take advantage of that. It also could be some type of balance sheet efficiencies, engineering being created. Could be either or, but I think a lot of these companies genuinely believe that their undervalued, and they're taking advantage of that. And in this case, even by raising debts to make sure that it buys back stock, lowers the floats, and waits for that next rip up, which is starting to happen. I think we've seen service now is up like 30% in a week. So that follows. We'll talk about another one of those companies with Snowflake. But not all SaaS is equal. And Salesforce, good print, not perfect, but certainly no SaaSpocalypse indicated in those numbers.

PATRICK MOORHEAD: 

Definitely. Let's go into Synopsys here, Dan.

DANIEL NEWMAN: 

Yeah, I mean, look, it's a true pick and shovel play of the entire AI silicon boom. Basically, every ship runs through Synopsys tools. You know, very little gets taped out without them. I think they had a nice TAM expansion with Ansys. And you're starting to see that cross-sell take place. I think the integration risk that some people had worried about with a deal of that size, it's not really showing up in the numbers. So you can be pretty happy about that. I think the guide up was the biggest and most important thing. They are an early signal for the market. They are seeing things well ahead of what comes to production. And so if they're, you know, if they're guiding up, it means that, you know, the pipeline is not slowing 18 to 36 months into the future. The only other thing I would say is I like this company as an interesting analog to not only the current AI trend, but also to some of the future trends. The whole simulation space, quantum space, robotics, all these things are going to run through Synopsys. So if you're looking for kind of what's the true pick and shovel of those industries, I mean, look, when people design humanoid robots, when people design rockets and satellites to send into space. They're using simulation tools from Synopsys in most cases.

PATRICK MOORHEAD: 

Yeah, listen, you can't do AI without Synopsys and the combination with Ansys makes it a full stack. You know, we always talk about, Jensen talks about co-design, right? And now the whole industry is talking about co-design. Synopsys is the company for co-design. And, you know, I think You know, putting Elliot on the board, I think, could be good. You know, I don't think it was necessary because I think the company was doing what it needed to do, but they got a new board of director member there. You know, kind of like Arm, when Arm wasn't fully monetizing what it could be doing, who knows, maybe that, you know, the Elliott Adder could help it. I'm a little skeptical, I'll be honest, because it's not the same as Arm, who in a sense was irreplaceable over a decade. Synopsys has a lot of competition out there. Say, so let's go to the next one. And that is that a snowflake. My gosh, you know, if Dell if Dell hadn't had had posted this week, Snowflake would be the mic drop. What were they up like 30 percent? I mean, they had a clean triple beat. They raised their guidance, strong sequential dollar growth and product revenue ever. And the stock ripped on it. And on the same day, they put a compute alliance in there with AWS, which to me is not necessarily a revenue driver, but it put them in a very interesting category with the hyperscaler giants. And I think NetNet Wall Street believes that they can more than just survive the AI swarm, but actually thrive. And that makes sense because data is the lifeblood of getting good results out of agents.

DANIEL NEWMAN: 

I mean, have you ever just thought that like, I said this on, I can't remember which show I was on. I think it was I was on CNBC this week and they were asking me about Anthropic being next to the Pope when he did this encyclical and what that means. And I said, I don't think a company has ever moved in this case, more private market cap is public, a private market cap by creating fear than Anthropic. Anthropic has been the company that showed up next to every opportunity to put fear into the market of industry destruction, job destruction, product destruction. And they went and raised money at a trillion dollars now. But none of it's come to fruition yet. I'm not saying never. I am not the Oracle that understands 100% that none of this can be done. But aren't we consistently seeing it show up as and? I mean, look, having a highly available data platform, it's scalable, MCP, API connected with running on the rules of business, isn't that going to be needed to really maximize AI? And if I think you saw the comments from Larry Ellison today where he talked about, look, the proprietary unique data is what your remote is going to be. Well, every company is going to need to invest more in that. They're not going to just give that data and drop it. That data has to go somewhere. And so where is that gonna go? It's gonna go to places like Snowflake. So I rest my case on that one.

PATRICK MOORHEAD: 

Yeah. Hey, let's say a quick follow through on HP, right? They were up 4% after the close. And a lot of that had to do with, they had a revenue beat, non-gap EPS beat. Their guidance was narrowed, right? They had to trim on the top end. And quite frankly, it's memory. The good news though, I think from PC stocks, like revenue units have gone down, but revenue has gone up. Pull-ins plus the memory repricing of PCs. And also when we were on the phone with the company, they discussed selling a lot more premium units. So moving the mix. So everything you might expect a PC and client solutions company to do, they did. They still need to find a new permanent CEO, which I think will give all investors, make them sleep better at night.

DANIEL NEWMAN: 

Yeah, I mean, I think there's a ton of pricing power. You know, we know that there's constraint across capacity. So even if these companies are seeing the AI, sorry, the AIPC rollout been slower than maybe someone expected, that is increasing. But even if that is happening right now, The whole supply-demand matching is creating opportunities. You saw Dell had good numbers. We didn't even talk about their PC business, by the way. Dell had good numbers. HP had good numbers. But it is a mix of being able to apply pricing power and growth. I think a lot of the questions are going to be, when does AI come to the edge and how can these companies really benefit? We go back to our argument about whether ROI is going to come from endless consumption of very expensive frontier models, or will we be running an API to a open source or a more dedicated model on a device at the edge? If that happens, that's going to be an explosive moment for the device companies. That's really what I think we need to all be 

watching.

PATRICK MOORHEAD: 

Yeah, for sure. Hey, let's talk about Everpure, just to wrap this up. Record earnings, right, which was great. They took a hit for their customers, right, on the memory side, right, the gross margins were challenged. And that, I think, was borne out in how the market reacted. When I stand back, though, I mean, just like we've seen storage and memory and compute and optics, I mean, everything has pressure put on it. But I was pleased to see what I'll call enterprise forward progress. We did see some shift. from as a service to actual revenue, right. And that was a little bit of a shift that the market was getting used to as, as ever pure aka pure storage was teaching people that it's about it's about ARR and not necessarily the revenue number, they bang out a huge revenue number.

DANIEL NEWMAN: 

Yeah, absolutely. I mean, look, it was a good quarter, a good result. I think you made a good point. Everpures focused on kind of sharing the load with their customers, which drove down some margin. I don't know that we've fully seen the inflection of storage yet. I still think there's a wave up. If agents take off the way we expect, we've had so much attention to memory, whether that's a new inference model, new inference chips, SRAM, CPUs, orchestration. But after that, where's all the data going to sit?

PATRICK MOORHEAD: 

Yeah.

DANIEL NEWMAN: 

It's going to have in storage. So I mean, we did see some good numbers here. Again, another thing, Dell had some good storage numbers, too. There's a lot of power in that particular business, and there's a lot of opportunity. I could see Everpreet really capitalizing on that. And by the way, watching, are they getting more wins from some of those sort of data native storage companies? That'll be the other thing to watch, you know, the DDNs and the VASTs and the WECAs because they were sort of not counted in that conversation, but I don't think they believe they shouldn't be counted in that conversation.

PATRICK MOORHEAD: 

Yeah, we were on the phone with Charlie. He did talk about some takeouts. And I believe that those were VAST and DDN takeouts. So I love the competition. I love the innovation. Daniel, great show, man. Let's wrap this up. We made it through. We're slightly over here. I appreciate everybody hanging in there. I hope you had a great weekend. We're testing whether we should publish this on Fridays or Saturdays or Sundays or Mondays. So bear with us. Thanks for tuning in. Hit that subscribe button. Take care.

‍