AMD reports record earnings but faces valuation skepticism amid supply and regulatory concerns
Summary
Speakers analyze AMD's record Q4 revenue of $13.4 billion and aggressive guidance, noting a structural shift where data center AI revenue surpassed server revenue. They highlight AMD's strategic moves, including the Helios rack-scale systems and acquisitions of Talas and World Labs, aimed at capturing agentic AI and physical AI workloads. Despite strong fundamentals, the stock fell 17% due to market fears regarding regulatory risks, peak capex narratives, and widening credit spreads. The speakers contrast AMD's high valuation multiples with NVIDIA's lower multiple, citing concerns over free cash flow and customer concentration. Ultimately, the hosts view the sell-off as a political overreaction to a strong earnings report, though they maintain a cautious stance due to execution risks.
Episode sections
- 0:00 – 4:21
Market Volatility and AI Sentiment
Discussion on the recent market sell-off driven by fear of regulation and capex peaks, emphasizing a contrarian approach.
- 4:21 – 10:46
The Role of CPUs in Agentic AI
Explanation of CPUs' critical role in orchestrating AI agents and managing logic, challenging the assumption that only GPUs are needed.
- 10:46 – 13:48
AMD's Helios Systems and GPU Ramp
Details on AMD's launch of full rack-scale Helios systems, allowing for better optimization and significant shipments in 2027.
- 13:48 – 16:27
Earnings Performance and Growth Guidance
Analysis of AMD's record Q4 revenue of $13.4 billion and guidance for $17 billion, highlighting a structural shift toward AI workloads.
- 16:27 – 23:00
Supply Chain Health and Regulatory Risks
Discussion on potential supply chain cracks due to construction delays and regulatory red tape, despite AMD's secured long-term commitments.
- 23:00 – 27:47
Software Ecosystem and ROCm Improvements
Examples of improved ROCm software, such as Anthropic's rapid deployment, reducing switching costs from NVIDIA's CUDA.
- 27:47 – 29:43
Strategic Acquisitions: Talas and World Labs
Overview of AMD's acquisitions of Talas and World Labs to hardwire inference into silicon and hire Fei-Fei Li's team for world models.
- 29:43 – 31:07
Capital Markets and Execution Risks
Concerns about the ability of capital-intensive companies to execute on delayed ramps and access cheap capital in a tightening credit environment.
- 31:07 – 35:24
Credit Markets and Supply Chain Constraints
Analysis of widening credit spreads for companies like Oracle and the critical bottleneck of HBM memory supply.
- 35:24 – 41:15
Valuation and Market Skepticism
Discussion on AMD's high valuation multiples compared to NVIDIA, citing concerns over political regulation, peak capex, and weak free cash flow.
- 41:15 – 47:19
Investment Stance and Business Outlook
Speakers express a 'lean bullish' stance, viewing the sell-off as a temporary political overreaction to a strong earnings report.
Claims worth verifying
NASDAQ had its worst day in six years during the recent AI sell-off.
Unidentified speaker (UNKNOWN) ·0:18
How to verify: Check NASDAQ composite index daily performance data for the week of November 5, 2026, against historical data. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
AMD Q4 revenue was a record $13.4 billion.
Unidentified speaker (speaker_0) ·3:31
How to verify: Review AMD's Q4 2026 earnings report filed with the SEC. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceAMD guided to approximately $17 billion in revenue for Q4.
Unidentified speaker (speaker_0) ·3:31
How to verify: Review AMD's Q4 2026 earnings report filed with the SEC. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceAMD expects revenue to grow more than 80% in 2027.
Unidentified speaker (speaker_0) ·3:31
How to verify: Review AMD's Q4 2026 earnings report filed with the SEC. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceAMD stock fell 17% (or 18%) on Thursday, November 5, 2026.
Unidentified speaker (speaker_0) ·3:51
How to verify: Check AMD stock price history for November 5, 2026, and compare to the previous close. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Anthropic engineer brought a frontier model onto AMD MI355 racks over one weekend.
Unidentified speaker (speaker_0) ·23:28
How to verify: Review AMD's Q4 2026 earnings report or press releases from AMD/Anthropic regarding ROCm performance benchmarks. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
AMD acquired Talas for $3 billion in cash.
Unidentified speaker (speaker_0) ·28:07
How to verify: Review AMD's Form 8-K or proxy statement filed with the SEC regarding the Talas acquisition. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceFei-Fei Li joined AMD as Chief Scientist via the World Labs acquisition.
Unidentified speaker (speaker_1) ·28:43
How to verify: Review AMD's press releases or SEC filings related to the World Labs acquisition and executive appointments. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceData center AI revenue became larger than server revenue for AMD in Q4 2026.
Unidentified speaker (speaker_0) ·14:58
How to verify: Review AMD's Q4 2026 earnings report filed with the SEC. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceMichael Burry called out a new trade idea capitalizing on the AI sell-off.
Unidentified speaker (speaker_0) ·0:48
How to verify: Check Michael Burry's public statements or filings (e.g., 13F) for Q4 2026. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceCredit spreads have been blowing out for Oracle and some other companies, showing cracks in debt markets.
Unidentified speaker (UNKNOWN) ·31:42
How to verify: Check credit default swap (CDS) spreads for Oracle (ORCL) and other megacap tech stocks on financial data platforms like Bloomberg or Reuters for the period leading up to November 2026. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Samsung signed an HBM4 supply agreement for AMD's MI455.
Unidentified speaker (UNKNOWN) ·33:11
How to verify: Review press releases from Samsung Electronics or AMD (AMD) regarding the MI455 supply chain and HBM4 specifications. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Foxconn, Dell, and Supermicro are starting to ship Helios systems in Q4.
Unidentified speaker (speaker_0) ·33:31
How to verify: Check quarterly earnings reports or press releases from Foxconn, Dell Technologies (DELL), and Super Micro Computer (SMCI) for Q4 shipments of AMD-based Helios systems. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
AMD announced 14 gigawatts of AI deployments and expects data center AI to more than double next year.
Unidentified speaker (speaker_0) ·33:51
How to verify: Review AMD's Q3 2026 earnings presentation or press release for the specific 14 GW figure and forward-looking guidance. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
AMD trades at 40 times forward EPS with a PEG ratio below 0.5x.
Unidentified speaker (speaker_1) ·35:44
How to verify: Calculate AMD's forward P/E using consensus analyst estimates for the next 12 months and compare to the trailing 12-month EPS growth rate to derive the PEG ratio. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceNVIDIA trades at 14 times earnings.
Unidentified speaker (speaker_0) ·38:36
How to verify: Check NVIDIA's (NVDA) current P/E ratio on financial data platforms as of the transcript date (November 2026). For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceAMD's free cash flow was only $500 million.
Unidentified speaker (speaker_0) ·40:16
How to verify: Review AMD's Q3 2026 10-Q filing with the SEC EDGAR database for the reported free cash flow figure. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Primary sourceAMD's data center AI revenue is expected to more than double next year.
Unidentified speaker (UNKNOWN) ·44:26
How to verify: Review AMD's segment revenue guidance for Data Center AI in the Q3 2026 earnings materials. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
The stock fell 17% in one day on Thursday, November 5.
Unidentified speaker (speaker_0) ·45:51
How to verify: Check historical stock price data for AMD (AMD) to confirm the closing price on November 4, 2026, and November 5, 2026, to calculate the percentage decline. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.
Do your own research
- Verify AMD's Q4 revenue and guidance figures in the official 10-Q or earnings press release.
- Check the current PEG ratio and forward EPS estimates using financial data providers like Bloomberg or Yahoo Finance.
- Review AMD's 10-K for details on warrant dilution and free cash flow generation.
- Assess the status of HBM supply agreements with Samsung and TSMC for the MI455 chip.
- Monitor credit spreads for megacap tech companies like Oracle to gauge liquidity conditions.
- Investigate potential regulatory impacts on data center permitting under current political leadership.
- Confirm Anthropic's deployment timeline and technical details regarding ROCm compatibility.
- Analyze the competitive landscape for agentic AI compute, specifically CPU vs. GPU demand forecasts.
Full transcriptexpand
- 0:00
- Unidentified speaker (speaker_0): Welcome back to the deep end. Today we are going deep into AMD. But before we get into the company, I think uh Brad and I kind of have to talk about the elephant in the room. The market absolutely crapping this week, especially the AI drawdown, where all of a sudden
- 0:18
- Unidentified speaker (UNKNOWN): being
- 0:18
- Unidentified speaker (speaker_0): a contrarian in the market is being bullish on AI when nothing truly fundamentally changed except for a potential.
- 0:29
- Unidentified speaker (speaker_0): red tape bureaucracy on his AI build out that I think is going to geographically change, but nothing about the underlining compute demand did. But clearly the market doesn't care because NASDAQ just had its worst day in six years. Um, all the AI bears, including Michael Burry, are flexing the chest a bit. Even Michael Burry uh is capizing capitalizing on it uh
- 0:48
- Unidentified speaker (speaker_0): by calling out a new trade idea. because clearly he looks like the genius in the moment, but we'll see if that's actually going to be the case down the line. But before we get into uh AMD Bradley, you want to talk about um how how your weekly for me I I'm I had to put makeup on because I have two black eyes.
- 1:06
- Unidentified speaker (speaker_1): Yeah, I uh
- 1:08
- Unidentified speaker (speaker_1): it's I mean we it seems really just feels like markets are racing between everything is way more amazing than. it truly is, and everything is way worse than it truly is. And we get one headline or one reason for momentum to swing the other way, and then we just go so far in that direction so quickly and so aggressively that it's it's uh I I think
- 1:28
- Unidentified speaker (speaker_1): keeping your head right now, I think taking deep breaths and laughing at the ridiculousness and focusing on affordable companies with solid rates of profitable compounding. with good prospects that are steady or brightening. Um, it's it's it's always an edge in markets. It's always an edge to free ourselves from having to outperform for our clients every 90
- 1:48
- Unidentified speaker (speaker_1): days. But it is especially an edge right now when these dislocations happen. And it feels so shitty in real time. It's like I I'll I'll be on X, greedy when others are fearful, like time to lean in, let's get excited. You have to trick yourself into feeling that way. You have to hijack your mind. and fight the natural instinct you have to want to put a hole in the wall when stuff like this is happening. Um,
- 2:08
- Unidentified speaker (speaker_1): because you have to zoom out and understand that in a couple of years it'll feel like a blip on the radar. And right now, being rational and being calm and process oriented is a k a large edge.
- 2:19
- Unidentified speaker (speaker_0): Oh, it's massive. I think that's where outperformance gets cr uh it's the beginning of where you outperform the market. It's times like this where you're just
- 2:29
- Unidentified speaker (speaker_0): in the discomfort, you're stewing in it, but instead of shying away and listening to all the FUD headlines that are very easy to kind of digest and follow, you stick with your routine. You the conviction you build on the research, this is where it shines. You add when there's blood in the water. And I also, before we move on to AMD, like one last thing. An introduction of a policy
- 2:48
- Unidentified speaker (speaker_0): is not the same thing as a new law.
- 2:51
- Unidentified speaker (speaker_0): It's data center build out's going to happen. The idea of floating of it's going to take a little longer than expected is not the same thing as all data centers are going to be halted in the majority of states in the US. Not the same thing. But um this is why we also wanted to kind of talk about AMD because uh during times like this, everything
- 3:11
- Unidentified speaker (speaker_0): that's associated with AI gets sold off. Does not matter what you reported, uh, what the earnings are because the market doesn't care uh right now. So that's why AMD is really timely video because they've had an they had an o home run earnings just this week, this past week. Uh QV, Q4 revenue kind of exceeded
- 3:31
- Unidentified speaker (speaker_0): expectations. And now they're talking about 80% revenue growth in 2027. So you basically have the stock uh move 30 percentage price. points in two sessions because they went up 10% after the support. Then on Thursday, during um it's kind of a block Thursday. It was like the worst day in six years. You had AMD down 18% because
- 3:51
- Unidentified speaker (speaker_0): of all the fear around AI regulation financing, whether growth rate of hyperscaled capex peaks in 2027. But this is why it's a very timely episode because we're gonna dig into why we think AMD is potentially I actually I I'll green green light it. I think it's a great opportunity to trade and add to AMD right now into this FUD. Uh so Brett, I kind of
- 4:11
- Unidentified speaker (speaker_0): pass it to you in the first section on why CPUs are cool again, because that's kind of one of the biggest tailwinds on AMD story in 2026.
- 4:21
- Unidentified speaker (speaker_1): Yeah, and it very I mean the the theme of the week is sector sentiment. So I I think distancing ourselves from that and focusing on how they're actually doing and and how they actually may be rewarded when sentiment is a little less aggressively negative makes a lot of sense. And for for AMD, CPUs are cool again. I think at the beginning of the cycle,
- 4:40
- Unidentified speaker (UNKNOWN): we
- 4:40
- Unidentified speaker (speaker_1): all assumed accelerate every workload, every single use case, every single application needs to have GPUs running every part of it. And and that was not correct. So so GPUs are the the workplace. work, and and we'll get into why um AMD is having a lot of progress there too, right now, but focusing on CPUs for a second. While GPUs do all of the very hard, complex work and and
- 5:00
- Unidentified speaker (speaker_1): do basically are the engine behind the magic that we see when we plug in an answer and see an output that that that still blows my mind sometimes. Um but CPUs GPUs are the uh supervisor or or the ground floor manager, I hear is an analogy. I hear other people um use. use a lot very frequently. And what that means is you need something to
- 5:19
- Unidentified speaker (speaker_1): orchestrate all of this. And the best thing to orchestrate it is a CPU that is well versed in static um static workflow orchestration and is very well versed in understanding how to efficiently conduct um and and again orchestrate that work for GPUs and and on their behalf. So while GPUs are doing all the work. while
- 5:39
- Unidentified speaker (speaker_1): they're um while they're doing this this impressive stuff, CPUs are behind the scenes kind of making sure they're pulling pulling the string strings like a puppet master to make sure they're behaving um in optimal ways that aren't uh leading to rogue agent activity, that aren't leading to exploding costs because you're not using the right model when you could use a cheaper one. Um CPUs are really the engine
- 5:59
- Unidentified speaker (speaker_1): that allows for uh order amid all of the the promising but also intimidating chaos that GP rack provide.
- 6:06
- Unidentified speaker (speaker_0): Yeah, I mean I think a genetic AI is a big part of why that is. It's creating a completely new reason for general purpose compute to accelerate again. I mean Lisa Sue's been very vocal about this as well where I think she recently talked about how personal agents are the next killer application for CPUs in the cloud. And that makes a lot of sense to me because again like the first couple years of
- 6:26
- Unidentified speaker (speaker_0): of AI was all about, I call them chatbots because it was just conversational tools that people use on ChatGPT and Cloud. Uh, those are incredibly GPU heavy, and that's why NVIDIA is just printing cache. They still are, but the center of intention or center of gravity in that era was just GPUs. And GPUs might do the heavy reasoning, but agents
- 6:46
- Unidentified speaker (speaker_0): still need a ton of ability to call. a database, run code, talk to an API, pull information from another application, check the results, decide what happens next. And then they repeat that process over and over again. And a lot of that orchestration still happens on CPUs. So that's why I keep coming back to something like we joke about uh just I think it
- 7:06
- Unidentified speaker (speaker_0): was like in May when AMD was always CPU first and almost got lucky that Agentic AI became one of the defining themes of 2026. because the architecture of agents just suddenly made one of AMD's historical strongest businesses become super important at the time where their I won't say the GPU business was
- 7:26
- Unidentified speaker (speaker_0): lagging, but it needed to catch up. And it was always having this narrative of oh, second source of this AI build-out, build out, why buy them when I can buy NVIDIA? CPU bucked that narrative, and I think MetaMuse.
- 7:39
- Unidentified speaker (speaker_0): is a great example because that was a ChatGPT moment, I believe, in this agentic AI theme was Muse. They passed, I think it was 10 million activations in six weeks. And eventually it's going to become hundreds of millions of personal agents constantly running tasks. Then 10 million people occasionally asking I don't know the chatbot question. So a question for
- 7:59
- Unidentified speaker (speaker_0): you, Brad: like, does why does agents need more CPUs rather than everyone simply moving on to the accelerator? It's probably a good question of why CPUs are cool.
- 8:09
- Unidentified speaker (speaker_1): Yeah. Because they I mean
- 8:13
- Unidentified speaker (speaker_1): there there needs to be a layer of step-by-step logic-oriented planning and and and conducting for all of this work. And GPUs are it it would sort of be like using you. using a rocket going to the moon to fly from Michigan to Ohio almost. Like there is something that is just so much
- 8:33
- Unidentified speaker (speaker_1): better in terms of the amount of the amount of resource intensity required to run this workload um and and the amount of cost required to run it with a CPU versus GPU that it just makes rational sense to to use to use these cheaper um CPUs in order to make sure that all the GPUs in in those racks are
- 8:53
- Unidentified speaker (speaker_1): conducting all this work that we want to. They are doing the research that we need to. They are um helping me edit my article so that um I make fewer stupid typos in in a post. Um it's just it's about how to get from point A to point B in the most impressive yet most cost-efficient manner. And CPUs are a big part of that, again,
- 9:12
- Unidentified speaker (speaker_1): logic-oriented planning where GPUs. would be running around like a chicken with their head cut off, not really knowing what to do. It's like that Spongebob meme where he's inside his mind and everything's on fire, kind of thing. And CPUs are sort of the teacher making sure all the students are um following along and and and not cheating and um not talking out of turn and raising
- 9:32
- Unidentified speaker (speaker_1): their hand when they have something to say, like all those things that that's what CPUs are making sure that they're their hall monitor, teacher, um supervisor, whatever you want. whatever analogy you want to use, that that that's their role, and they do it far more cheaply than a GPU counterpart.
- 9:43
- Unidentified speaker (UNKNOWN): I
- 9:43
- Unidentified speaker (speaker_0): mean, as you're saying this, like I think a good comparison that comes to mind because I I'm just watching the Bear and Hulu trying to finish the last seasons, the worst season of the show. That's I digress a bit. But uh like you guys know how chaotic CARM is. Like CARM can't run a kitchen.
- 9:58
- Unidentified speaker (UNKNOWN): I
- 9:58
- Unidentified speaker (speaker_0): view GPUs as like the chef, I view CPUs as a kitchen man. manager who runs the kitchen to make sure the restaurant's operating because the chef's super important. That's the quality of the food for the whole restaurant, but it cannot cook as well as plan all these things to make sure the kitchen runs efficiently. That kitchen manager is the CPU. So
- 10:18
- Unidentified speaker (speaker_0): GPUs are like Carm in the bear. Uh So I think that's a good way of like thinking about my end of like what the difference is. But also,
- 10:25
- Unidentified speaker (UNKNOWN): I think
- 10:26
- Unidentified speaker (speaker_0): it's a good. transition and talk about like why GPUs can become dramatically more important uh still while CPU demand grows right beside them. And I think that's what makes AMD unusually interesting because it has credible positions on both sides. So I kinda wanna pass it to you on talking about the Venice revenue shipments that Our are beginning in
- 10:46
- Unidentified speaker (speaker_0): Q3, the move to TSMCs, TNM, 2NM, and then like the cloud rollout. in 2027. Like there's a GPU story here that's not just CPU for AMD.
- 10:56
- Unidentified speaker (speaker_1): Yeah, I think so. Their new Helios, React Scale systems are big, big moment, big unlock for them. Um Venice CPUs, and then their their latest MI GPUs. Um, so basically, this I mean, there was a there was a large competitive gap with with NVIDIA offering something like this, and an AMD just offering more individual components. in more of a piecemeal manner.
- 11:16
- Unidentified speaker (speaker_1): And the ability to offer companies full racks, um, it not not at the chip level, but at the at the full rack level, and be able to perform optimizations because their their software ecosystem is getting a lot better too. Have having an army of developers geared towards optimizing every part of inference and training work that happens for every single
- 11:36
- Unidentified speaker (speaker_1): component instead of having several things. third-party components and trying to optimize how they're working together while trying to get that kid to communicate with how AMD's components are working. And just it is it is far easier to drive performance gains and cost gains and keep that annual cadence of platform launches rolling where you're
- 11:56
- Unidentified speaker (speaker_1): justifying those launches because you have have taken a big enough step in the right direction. Um having um a a rack. scale design is extremely important for that. It's extremely important for more customization opportunities where they're working hand in hand with their large anchor clients to build them compute architecture that is best suited for their cloud environments, sort of kind of emulating why custom
- 12:16
- Unidentified speaker (speaker_1): processors have been so popular in certain use cases. MD takes a very active role in terms of crafting um their their Helios products um in several different malleable ways that cater. to different cloud environments. Um and I think when you pair Helios with the fact that their software ecosystem is now creating or now enjoying critical massive developers that
- 12:36
- Unidentified speaker (speaker_1): are working on AMD level equipment, um, there are advantages that NVIDIA has enjoyed for several years now in this AI build out that maybe still are there but do seem like they're dwindling a a little bit, especially because um this this open everything approach that AMD AND takes makes it a lot easier to build uh or make makes it a lot easier to create that kind of
- 12:56
- Unidentified speaker (speaker_1): critical massive developers in your platform because you're not threatening them with vendor lock, you're not making them work on certain tools. Um so so yeah, they they've the this this second half of 2026 ramp um is when things are really supposed to get exciting as Helios um has has now shipped um a few units as of uh this quarter to uh who's who was it, Anthropic, Meta,
- 13:16
- Unidentified speaker (speaker_1): Microsoft, OpenAI or Cloud.
- 13:17
- Unidentified speaker (UNKNOWN): Um,
- 13:18
- Unidentified speaker (speaker_1): but 2027 is really the year where you're going to see this take off and then some, which is why triple digit growth for the sec for the um segment is the expectation. And also why now they've moved from 80% CPU growth in 2027 to 100% plus growth there too, just tying it back to why AI is so positive um for that part of of the tech t technology stack.
- 13:38
- Unidentified speaker (speaker_1): Um, but this is AMD's moment to be able to pitch and argue that they have something that rivals uh in in pretty much every way um what NVIDIA's had for a few years.
- 13:48
- Unidentified speaker (UNKNOWN): I
- 13:48
- Unidentified speaker (speaker_0): mean, I also think that's why this recent earnings report was just so massive for AMD, because the story obviously on the GPU side for AMD for years has been the narrative that they're going to be that second source in AI accelerators beyond NVIDIA, sub ten 10% percent share will do no like wonders for the company. We finally got the earnings report that showed that they're not just
- 14:08
- Unidentified speaker (speaker_0): a second source in the narrative, they're the second source in production. And this was like the first quarter where you can see the real MI455
- 14:18
- Unidentified speaker (speaker_0): X revenue as well as the Helios revenue. And you already have like Had Anthropic Meta, Microsoft, OpenAI, and Oracle putting those systems online. And I also think that we had HBC. HBE, Supermicro, and Dell begin shipping these Helio systems in Q4. So like I think that's why we finally got proof in the pudding that this is a real story that's happening right now.
- 14:38
- Unidentified speaker (UNKNOWN): And
- 14:38
- Unidentified speaker (speaker_0): the guide was extraordinary. Like AMD is like calling for $17 billion of revenue in Q4. I think estimates were $16 billion. dollars. So that's like nearly 30% sequential growth. And more importantly, for me, it was like data center AI revenue. is actually becoming larger than the server revenue for the first time in Q4 this year. And Lisa said that I
- 14:58
- Unidentified speaker (speaker_0): think it was data center AI is going to double again in 2027, while data center itself is going to become more than 80% of company revenue. That's a completely different story than what AMD's Ambie's investors had in this theme in this company a couple years ago. And I think that's why. the sell-off they've experienced like last week on
- 15:18
- Unidentified speaker (speaker_0): Thursday is just so compelling because I don't know, multiple analysts started talking about hyperscalar cap acts like peaking in 2027 and the semiconductor complex, the cycle basically got repri. I mean, though all semiconductors got repriced on that's on the second derivative of that sales side tsunami of
- 15:37
- Unidentified speaker (speaker_0): updates. But AMD just told us that its own AI revenue is about to double again next year. Those two things can coexist because semis can absolutely derate before CapEx itself itself declines. But man, if you believe that growth rate isn't peaking and this is just a multiple compression story, AMD's earnings just provided a lot of
- 15:57
- Unidentified speaker (speaker_0): evidence that their business is accelerating at a rapid pace. Um
- 16:01
- Unidentified speaker (UNKNOWN): so
- 16:02
- Unidentified speaker (speaker_0): I kind of want now to pass it to the next theme of this uh
- 16:07
- Unidentified speaker (speaker_0): topic of discussion is any signs of psycho health or cracks. Because I think that is the name of the game for semiconductors after all the FUD we've heard about 2027 peak earnings. That's obviously gonna dictate the Take whole semiconductor complex. Doesn't matter what you do in the complex. If you think growth is gonna peak going to this year.
- 16:27
- Unidentified speaker (speaker_0): you're gonna get punished. Like I think this is a really important topic to talk about. So I'm gonna pass it to
- 16:32
- Unidentified speaker (UNKNOWN): you.
- 16:33
- Unidentified speaker (speaker_1): It doesn't, I mean, when you hear about I mean not peak 20 peak caps in 2027 or in 2026 doesn't feel likely in 2020 or in 2027, but peak growth rates definitely feel very, very likely. Um, and when you think about all these component players and where how are they gonna going to find the rapid, rapid time. top line compounded with margin expansion and amazing pricing power
- 16:54
- Unidentified speaker (speaker_1): that justifies them continuing to zoom higher? Um, I think some of not all of them, but some of them probably do need growth rates and capex to stay very, very lofty. Um, but for something, someone like AMD that's so supply constrained, um, that has such high quality and in and well capitalized customers um and and has such tight relationships with those customers to be able to plan over the next
- 17:13
- Unidentified speaker (speaker_1): few years, it doesn't seem like there are huge cracks forming on the demand side, especially when Muse is coming online and and AVs are starting to proliferate and and the implications that has for free physical AI and all the compute need that that is needed there. But what is starting to look more dicey um is uh when we AAOI and Intel and and these companies um and and
- 17:33
- Unidentified speaker (speaker_1): IRN Iron say saying basically saying there's amazing demand, but we're having trouble figuring out how to get it online as quickly as we said we were going to, and there are delays now. Um, and there, I mean, these are massive undertakings, big, big projects, massive buildings, complex supply chains. Um, one delay, one very short delay, can easily turn into more um and can kick can kick the can down the
- 17:53
- Unidentified speaker (speaker_1): road while uh these amazing demand tail ones we're currently enjoying uh continue working their way through the system. So I think the the risk here or or the the potential. potential sign of cracking is not from a natural demand point of view. I think it's from a supply point of view and making sure that these companies, that a lot of them are not proven in terms of these
- 18:13
- Unidentified speaker (speaker_1): build-outs that they're planning and telling companies to expect and are now saying it's going to be a little slower, just makes everyone a little bit nervous and appreciating their ability to actually do what they're they say they're going to and and to make the amazing sell side estimates that that are in place. over the next few years um realistic to expect. So from a an end market demand point of view um no
- 18:33
- Unidentified speaker (speaker_1): I I I don't really see many signs of of demand cracks. I mean there's people who cite who tweet out things like uh the black well GPU rates are are crashing and and then you look at charts that that aren't anecdotal and in the opposite looks like it's happening. They just keep going up and to the right um but everything that needs to look good from a demand perspective looks good. It's
- 18:53
- Unidentified speaker (speaker_1): just okay, now ball is very much so in your court to conduct a very tough few years of strong, clean execution um to make sure these ramps that everyone is now expecting and everyone is now baked into their base case um take place um like they're supposed to.
- 19:07
- Unidentified speaker (UNKNOWN): I
- 19:07
- Unidentified speaker (speaker_0): mean, I think I think that's why it's so important to isolate fundamental information from stock price because Andy. Amdi, I mean, we got so much clarity from Lisa this week, just alone. Like Amdi's telling us already that they've secured the wafers, substrates, like advanced packaging and memory required for its 2027 growth plan. Lisa Sue also made commentary that they moved from
- 19:27
- Unidentified speaker (speaker_0): playing from one or two years ahead towards three to five years ahead because the customer commitments are becoming just a lot more there's a lot more duration to these commitments. So uh I think that I'm trusting what she's saying over what the FinTwick gurus are saying on uh as the AI bears. And I also think that the fact that
- 19:47
- Unidentified speaker (speaker_0): AMD has secured the supply chain around such a huge growth guide guide is huge because you don't reserve these wafers, HBMs, like packaging years ahead based on vague indication that customers might eventually show up. Like they're getting clear confirmation in supply and and arrangements that are telling them we're this build out is happening
- 20:07
- Unidentified speaker (speaker_0): at the same pace it's been. Get all the all the tools you need because we you're gonna have a lot of demand. And I think that market can still decide doesn't want to pay the same multiple because 2027 might represent peak growth. But that's very different from saying that the actual build out is disappearing next year. But I think it's a good time to talk about, I mean.
- 20:26
- Unidentified speaker (speaker_0): a bit on the cracks, because I think there is uh a bit of an inventory build like I talked about uh I just mentioned right now that uh who know like that's a risk and also some margin pressure and a memory cost. Like I want to pass Steve on some cracks in the story.
- 20:42
- Unidentified speaker (UNKNOWN): Yeah,
- 20:42
- Unidentified speaker (speaker_1): and I would say just one more note on regular regulatory and at risk of of introducing politics into the conversation, but hopefully doing so in as innocent a way as I possibly can. When we have these very loud and and kind of annoying elections where you see these political ads saying the other person's crazy for eight months, I mean, I both both wings have a vested interest in making
- 21:02
- Unidentified speaker (speaker_1): it sound like the other wing is crazier than they are, so that they get more seats and they get elected. I mean, it's a natural motivation. So we're hearing from people like that who won the Texas governorship and who everyone was really afraid of just clamping down an irrational. in ways um on data centers speak a bit more rationally and a bit more reasonably than I think many were fearing, um, which
- 21:22
- Unidentified speaker (speaker_1): does bode well for this build out taking place uh in a bit more of a status quo way than than many are are fearing uh right now. And I will there there are there are people there are crazies on both sides um that have that have wild ideas that um get get a lot of attention. But but the vast majority of these people are are ra I mean it's kind of kind of funny to call
- 21:41
- Unidentified speaker (speaker_1): politicians rational and insane, but they're not trying to destroy the world. Like they're they're not trying to make everyone's lives miserable. They're they're they're making decisions with smart people in their ears um based on what they think is best after they've had all their campaign rhetoric, after they've said all the stuff for the sound bites. So I think it's always a little like the the the the takes that I hear after midterms,
- 22:02
- Unidentified speaker (speaker_1): and then we'll get through politics are everything has changed so drastically, and you need to change a lot of stuff because now these companies are screwed for the next two years, and these companies are in a way better spot for the next two years. And I would just say that those opinions are are seldom correct. Um they're they're usually they're usually um good to take a contrarian
- 22:21
- Unidentified speaker (speaker_1): view towards, and things don't change as much as we think they're going to between election from election to election. in this country, um, where
- 22:30
- Unidentified speaker (speaker_1): where there's there's checks and balances, where there's um just a lot of reason to make this amazing consumer economy that that powers the entire world um stay healthy.
- 22:40
- Unidentified speaker (UNKNOWN): I mean,
- 22:40
- Unidentified speaker (speaker_0): I think I was really well set on what the cracks are. Uh I think we can move on to the next section, even uh talking about kind of the software and inference push, because AI inference is a major challenge. tailwind on AMD and NVIDIA. But AMD specifically is got the luxury now all of a sudden of closing that gap. Because I think one of the biggest reasons I've become more bullish on
- 23:00
- Unidentified speaker (speaker_0): AMD, especially as the inference boom has begun, is it beginning.
- 23:09
- Unidentified speaker (speaker_0): Some version of the silicon Looks pretty good, but I don't know, CUDA, which is NVIDIA's software offering. makes switching way too painful. And I think AI itself is starting to attract, attack that part of the switching cost a bit more because Rock M10 is obviously a better software in its own right because AMD is improving cluster startup, agentic performance, and running through
- 23:28
- Unidentified speaker (speaker_0): Rock M AI. But management said that uh work that historically required weeks of engineering optimization can actually happen in hours. And that example uh we we're gonna come back to is Enthropian. anthropic because AMD said I think it was like one uh anthropic engineer used claw to bring a frontier model onto MI355 racks over one weekend. And the anthropic
- 23:48
- Unidentified speaker (speaker_0): room actually includes using claw to help accelerate Rock M development. That's a massive feedback loop because the same AI coding technology driving demand for compute is also making it easier to move workloads between all these different types of compute. I'm not saying that. CUDA suddenly loses its moats. NVIDIA is still an enormous ecosystem. And
- 24:08
- Unidentified speaker (speaker_0): years of developers tooling around it is a real mo because these developers are very hard headed, they're stubborn, they don't like to go out of their comfort zone. But I don't think it moves the competition. Uh but I do think that it moves the competition more towards total cost, performance, and availability if the software migration becomes a lot easier and at least. Lisa Sue's comments around custom
- 24:28
- Unidentified speaker (speaker_0): ASICs were pro were important for the same reason because she basically said the customer ultimately cares about total cost of ownership and tokens per dollar. And I agree with that. That's the new like KPI that everyone needs to focus on in the inference era. And it doesn't have to make every workload run that on the same exact architecture, it just has to kind of give a
- 24:48
- Unidentified speaker (speaker_0): customer a portfolio that optimizes different workloads. more efficiently. So, like, Brad, question for you. Like, do you think AMD is actually catching CUDA at this point, or is it almost trying to go around CUDA by making profitability and like, I don't know, the economics more than just the ecosystem itself?
- 25:06
- Unidentified speaker (UNKNOWN): No,
- 25:06
- Unidentified speaker (speaker_1): I think they I think they have matched and and and and gotten a lot of the value that having that closed ecosystem that NVIDIA is built. is so impressively it provides. And and I'll say, and I mean this is why they bought companies like Talus and and they're trying to specialize in more low latency inference and just get better at that and more efficient at that. Um, but
- 25:26
- Unidentified speaker (speaker_1): but at the end of the day, you you convince these people again, like you're saying, you convince customers to buy your chips and you convince them to to deploy them in their cloud environments or in their enterprise environments because you deliver great ROI. and because it it they they they make a lot more money than they spend. I mean it's simple enough. I don't I everyone understands that idea. But
- 25:46
- Unidentified speaker (speaker_1): one of the key input costs or not input costs, operating costs of of of running these um these these next gen um assets is depreciation. And and when when you're seeing things like um them catching up in software level optimizations and having more developers that are are geared towards making sure older generations of chips are seeing
- 26:06
- Unidentified speaker (speaker_1): more performance and and and more efficiency extracted from them, that is profoundly positive for depreciation schedules and useful lives for the chips that AMB is providing and something that that NVIDIA has enjoyed over them a lot more for the last few years, that that I think I I I don't know if they fully closed the gap. I I I it's hard for me to say to to to rank score them and give an Apples to Apples
- 26:26
- Unidentified speaker (speaker_1): comparison. but I will say that AMD has made a ton of progress in closing that gap and and and going from um something that doesn't really have much here to um someone that has something that is is is highly formidable and capable of um capable of of creating uh of creating an environment where their Helios GPU systems, they argue, are are better performant than other products on the market.
- 26:46
- Unidentified speaker (speaker_1): And when they say that, they're talking about NVIDIA. And I'm sure um I I'm sure Jensen and NVIDIA would say, You're wrong, and we're higher performing. And then they'd argue about different ways of of of classifying that and both both walk away feeling right. Um, but basically they they have they have closed a a large gap in their product ecosystem that
- 27:06
- Unidentified speaker (speaker_1): allows them to more perpetually and more meaningfully and more frequently make optimizations to the other very important parts of their ecosystem that all that and and in turn that makes them a much better competitor for NVIDIA. It allows them to
- 27:19
- Unidentified speaker (speaker_1): create products and deployments where companies are are comfortable with five, six, maybe even longer useful lives for these chips. So they can they can pallet the kind of depreciation intensity and overcome that with the revenue they're generating to to m create good returns and to look up and say, man, we should buy more of this compute because that was a great idea. Um so
- 27:39
- Unidentified speaker (speaker_1): software is a big, big, big part of that and and their focus here has been has been warranted, it's been well placed. It's gone well. Um yeah. No,
- 27:47
- Unidentified speaker (speaker_0): I mean I think their MA strategy has made it really clear that AMD is trying to build a broader AI ecosystem rather than tr trying to catch the king NVIDIA generation by generation. Tal you mentioned talus, Talos. like I think talus Talos is one side of that strategy. Like AMD is basically buying technology that can hardwire a model directly into the silicon for extremely fast and cheap
- 28:07
- Unidentified speaker (speaker_0): inference, which gives them another architecture for workloads where general purpose GPU may not be the most efficient answer. And I also think that deal being $3 billion in cash is just it's so tiny when you're talking about one trillion dollar company. I also think the other side of their strategy has been world labs. Uh That's basically a huge.
- 28:27
- Unidentified speaker (speaker_0): huge talent and research bed because there's no revenue, but you're bringing uh oh my god, I'm gonna butcher her name. It's Feife Lei, I think it Feife Lee, um, into AMD as a chief scientist and getting a team working on world models that can you're
- 28:43
- Unidentified speaker (speaker_1): not you're not tight
- 28:43
- Unidentified speaker (UNKNOWN): with
- 28:43
- Unidentified speaker (speaker_1): Feife.
- 28:44
- Unidentified speaker (speaker_0): I'm not tight with Fei Faye Fei. Fey. But I think she was a huge aqua hire because it does bring it gets a team working on world models. that can understand the physical environments, aka physical AI, robotics, autonomous systems. Like these endpoints of AI are going to be massive in the next couple of decades. So they're going to become the next major compute platform after generative AI. It's not
- 29:04
- Unidentified speaker (speaker_0): a matter of if, it's just when. So world models can become super important in that era. And AMD wants to have the architecture ready before the market fully exists. Love that story. strategy. So that's why I call World Labs a genius acquisition when it was announced, because Lisa Sue is buying the intellectual layer that can influence
- 29:24
- Unidentified speaker (speaker_0): what the hardware needs to become three or four years from now, rather than trying to bolt AI talent onto the company after physical AI has already taken off. Uh So I kind of think we should pass it to Warren, Warren a second. look. Uh the supply chain. What do you think about this?
- 29:43
- Unidentified speaker (speaker_1): Yeah, I I I just I I get nervous when when I hear companies just delaying ramps by three to six months after they've spoken so confidently about having everything in place to pull it off. Um and it's like it almost like I f like when I when it reminds me of like analyst profit estimates. anal Analyst schedule estimates are coming. or company schedule
- 30:03
- Unidentified speaker (speaker_1): estimates, if they're moving in one direction, they're probably going to keep moving in in the same direction for a considerable amount of time longer, barring some large catalyst that that creates a large shift. But if there are delays now forming and if we have moved from everything is is in check and and and and we're gonna deliver on this and we're gonna execute to well, we're gonna execute, but it might take us
- 30:23
- Unidentified speaker (speaker_1): six months longer. Um I I don't I I don't have a huge amount of confidence in the fact in the fact that these issues aren't gonna keep popping up um as we continue stressing out the grid more, as as we continue potentially seeing more red tape enter the um regulatory backdrop. Um, maybe not in a severe way as people think, but still probably some more red tape nonetheless in
- 30:43
- Unidentified speaker (speaker_1): certain states where there's been important data center projects. Um
- 30:47
- Unidentified speaker (speaker_1): to me, it's that that's a concern. Um and something that needs to be worked through and needs to be proven out. And then the other massively important thing for for a supply chain is access to cheap capital, especially because a lot of these companies are far from profitability. If you look at any of the neo clouds, they're a few years away from making free cash, from generating positive
- 31:07
- Unidentified speaker (speaker_1): free cash. You need capital markets to be wide open so that you can um you can raise debt in in a way that's not going to blow up. your balance sheet and and your income and your interest expense line. You need capital markets to be open and and investors willing to accept the d the inevitable dilution that's going to come your way or you're not going to be able to fill all fulfill all
- 31:27
- Unidentified speaker (speaker_1): these promises that that probably can come with good returns and probably can make you some money but need an exogenous backdrop to cooperate if you're going to be able to realize that outcome and and on honestly, I I
- 31:42
- Unidentified speaker (UNKNOWN): there
- 31:42
- Unidentified speaker (speaker_1): there have been there have been signs of of cracks forming in debt markets and in credit spreads blowing out for Oracle and some other companies.
- 31:51
- Unidentified speaker (speaker_1): Even for megacap balance sheets, where if if they're in trouble, everyone's in trouble. We have seen some some spreads widening a bit. Um, I just think people are a little more antsy, but maybe they might have come back in in the last week. I haven't I or in the last few days I haven't checked. since midterms. But they there are reasons to look at this very, very closely. And to
- 32:10
- Unidentified speaker (UNKNOWN): well,
- 32:10
- Unidentified speaker (speaker_1): there's always reasons to look at this very, very closely, but there are reasons to be a not not not scared, but a little bit paranoid and constantly obsessively seeking out signs that um people providing the liquidity to fund this these gigantic build-outs build outs and make sure there aren't bottlenecks that get in the way is is still affordable enough for these. companies to access. And NVIDIA will step in and provide guarantees
- 32:30
- Unidentified speaker (speaker_1): way more than AMD has. These companies will step in and try to alleviate some of the concerns and guarantee some of some of the financial risks here. But there are big, big financial risks here. And that means credit markets need to be willing to accept those financial risks. Um and that relies on good liquidity, uh, good, good, good enterprise sentiment, relies on a
- 32:50
- Unidentified speaker (speaker_1): lot of things. that are still true enough right now, um, but need to remain true for these companies to get from point A to point B and to start generating the profit that they need to sustain their own build-outs without relying on capital markets.
- 33:01
- Unidentified speaker (speaker_0): No, really well said. I think the only thing I would add to that is uh obviously, mem, we should talk about the memory side of the supply chain. I think that's probably the most important constraint because
- 33:11
- Unidentified speaker (UNKNOWN): I don't know
- 33:11
- Unidentified speaker (speaker_0): what Samsung just signed the HBM4 supply agreement for M. MI455 in HBM4. That's probably the tightest component in the entire industry right now. And that's what why it's a big reason for me that I continue thinking that memory has pricing power, even though I worry about the long term supply cycle, because AMD can design an incredible GPU and still miss the revenue
- 33:31
- Unidentified speaker (speaker_0): number if there is not enough HBM available to ship the system. And then you move out to the rack level, where Foxconn is one of the major AI server. builders. Um, and HB, Dell, Supermicro all are starting to ship Helio systems in Q4. So, like when AMD tells us that 14 gigawatts of announcement announced AI
- 33:51
- Unidentified speaker (speaker_0): deployments are coming, uh, and they it expects data centers AI to more than double next year, there is an entire semiconductor Semi conductor supplier ecosystem underneath that that's growing. And rather than AMD just capturing every dollar itself, it's lifting a whole system below it. So I think that that's why I keep coming back to memory. I think that HBM is the
- 34:11
- Unidentified speaker (speaker_0): easiest place where demand visibility and physical scarcity collide. I mean, every additional MI450 deployment pulls so much more HBM alongside it. And AMD isn't the only customer trying to secure that supply. You have NVIDIA, Google, AMD. Amazon, every other ASIC program are competing for the same ecosystem. And
- 34:31
- Unidentified speaker (speaker_0): guess who's pr uh who's the factory doing all the work for everyone? TSMC. They and they benefit enormously. So um I think that right now, supply chain, the is the biggest worst for supply chain is memory, obviously, has a as scarce input, but man, capital markets. Luckily, we have principal Jensen. Uh, they were
- 34:51
- Unidentified speaker (speaker_0): they print $200 billion of cash every year uh starting next year. I think that it's very difficult for me to kind of get a little worried about the cracks on the supply chain when you have one company just making that much money. I get the circular financing nature of it, but if you believe AI is real, it's continuing, this build-out's What else gonna happen. There is interesting
- 35:11
- Unidentified speaker (speaker_0): lanes that you can go besides the capital markets. Uh so go to the next section. I mean, why so cheap? Where's the love? Kind of think um it's probably a good time to lay out the valuation, Brad. I'm gonna pass it to
- 35:23
- Unidentified speaker (UNKNOWN): you. Yeah,
- 35:24
- Unidentified speaker (speaker_1): good, good segue here because I think why so cheap has a lot to do with everything we just talked about and all the risks that people are anxious about, cycle longevity, how long this lasts, where the cliff is is felt, how long all of the signs are going to stay positive, like they still are, by the way. Um but for from a valuation point of view, it I mean it trades at 40 times forward EPS. So it's not like traditional
- 35:44
- Unidentified speaker (speaker_1): valuation framework is cheap, but then you look at the fact that they're going to compound EPS above a 90% clip for the next two years, and the peg ratio is is below 0.5x. So that really is investors saying we're baking in a ton of doubt and a ton of uncertainty and a ton of skepticism to you guys realizing the ramps without again falling off a financial cliff at the end of
- 36:02
- Unidentified speaker (UNKNOWN): the day.
- 36:03
- Unidentified speaker (speaker_1): that. And and and as stocks are forward looking, it doesn't matter if if they they're looking out 18 months and expecting um $100 in earnings, if they expect that hundred dollars in earnings has a good chance of going to ten dollars in earnings, and I'm using extremes here, then the Ford multiple is going to be based on that ten dollars. It's not gonna be based on the hundred dollars, it's gonna be based on where things are going and where ex
- 36:23
- Unidentified speaker (speaker_1): visible expectations are heading. Um so why so cheap? Um I I think the the the two charts that that we need to I need to get better at being able to just display charts in here, but we'll we'll get there. Don't don't worry. Don't worry uh listenership. But the the PE chart for AMD over the last 10 years looks very, very normal. It is right around 10-year means. Then you look at the
- 36:42
- Unidentified speaker (speaker_1): 10-year price chart, mean at 5X currently at 11 11x. X. Margins are insanely good for AMD right now because pricing power is so insanely good because demand or supply demand mismatches are so you guessed it insanely good. Um so as those things remain intact, you can see uh you can see an earnings multiple look very, very
- 37:02
- Unidentified speaker (speaker_1): reasonable while they're trading at a much higher multiple of their revenue base than they have in a long time. Um, but that that really is the source of the cheapness. It's people wondering what happens to the margin profile when there's not such an aggressive gap between what is needed and what is available. And it's it's people figuring out and trying to make their
- 37:22
- Unidentified speaker (speaker_1): best educated guesses on when that happens. But that that's all I again, this is all future anxiety. And that's where all of this is coming from. People trying to be the hero to find that signal that that that confidently says that marks a peak. But if if you look, if you listen to all these companies, if you take them for their word, if you see all of the moment momentum happening on
- 37:42
- Unidentified speaker (speaker_1): consumer agent side of things and A V proliferation, there is reason to believe that and and you listen to all these CEOs talk, which I know they're talking their
- 37:50
- Unidentified speaker (UNKNOWN): own
- 37:50
- Unidentified speaker (speaker_1): books, and they're not going to be the first people to
- 37:51
- Unidentified speaker (UNKNOWN): tell you
- 37:51
- Unidentified speaker (speaker_1): things are bad. Um, but it's it really does feel like this hasn't ended. And it does feel like the the aggressive volatility we're feeling right now, it's Friday, November 6th, sixth. so you uh y it was a few days ago. I'm sure you guys remember how fun of a week that was. Um it it's really tied to elections and tied to um sentiment,
- 38:11
- Unidentified speaker (speaker_1): but not real tangible evidence of financial deterioration being imminent.
- 38:16
- Unidentified speaker (UNKNOWN): Yeah.
- 38:17
- Unidentified speaker (speaker_0): And I also I'm I'm gonna restrain myself from saying the obvious answer on why so cheap. Um and that obvious answer is you have Nvidia trading at 14 times earnings. They're growing just as much as AMD. So why bother paying? I don't even know what their earnings multiple is after this drawdown. Maybe it's like 30 times earnings. It's still twice the multiple in
- 38:36
- Unidentified speaker (speaker_0): NVIDIA is. So like I'm gonna restrain myself from saying that because that's such an obvious answer. I think it's the market basically not trusting the denom Not the the 2027 denominator yet, because there are four different concerns that are stacked on top of each other right now. That's not just an AMD specific one, but it's just in general for the whole semiconductor complex. I mean, the first
- 38:56
- Unidentified speaker (speaker_0): obviously is the political environment. Democratics Democrats taking both chambers has obviously increased the probability of tighter AI regulation. And so slower data center permitting and if infrastructure cannot physically come online, then obviously GP revenue AMD's modeling for the next couple years cannot show up on schedule. That's what why what I mean by
- 39:16
- Unidentified speaker (speaker_0): it can look cheap now, but if the denominator changes, it could very easily not look cheap. And that's honestly the second thing I want to talk about is peak capex argument because multiple cell-side analysts on the same day on Thursday, which caused a lot of fear, said that growth rate of hyperscaler is going to peak in 2027. And semiconductors historically can
- 39:36
- Unidentified speaker (speaker_0): peak when the growth rate of CapEx peaks rather than when the spending itself falls. So because of that, there's a lot of worry of holding the bag before the rug gets pulled, and they'd rather just get out now, ask questions later. The next one, which we didn't talk about too much, is the warrants to customer concentration risk of AMD. Like you're looking at the
- 39:56
- Unidentified speaker (speaker_0): huge OpenAI and meta incentives and asking how much is AMD paying to buy that initial share and how much. dilution is ultimately coming with it. The last thing I want to talk about is uh the financial quality of the growth, if that makes sense. Like free cash flow was uh only $500 million. Gross margin is basically flat, even though data center
- 40:16
- Unidentified speaker (speaker_0): revenue is exploding. Don't love that operating leverage. So I think the market is basically saying, like, show me the cash and show me their margins before I give you full credit for the revenue. I think that's fair. I really do think that's fair because there's a legitimate fundamental concern around regulation and peak capex colliding when you're talking about a very
- 40:36
- Unidentified speaker (speaker_0): crowded trade. And AI was a very crowded trade, and the trade also has a lot of leverage. But I do think the market is making the mistake of treating peak growth like no growth because AMD can absolutely experience a multiple compression if. 2027 is the high point of capex growth while still producing
- 40:55
- Unidentified speaker (speaker_0): extraordinary revenue and earnings growth for the next couple of years. And I think that's where the debate lives right now about what those multiples deserve, uh, whether then whether or not their earnings actually get there because they're going to be making their earnings. Um, so I kind of want to pass, go to our favorite segment then, because this is actually one of those episodes where it's not obvious if we're bullish or
- 41:15
- Unidentified speaker (speaker_0): bearish. I know typically in the last a couple episodes, like obviously we're bullish. Let's not even talk about that. So I kind of curious on you, uh, first, like, are you bullish or bearish on AMB right now?
- 41:24
- Unidentified speaker (speaker_1): Yeah. So I run a very hyper concentrated portfolio. So I should preface this by saying I'm bullish on a lot of names that I don't have in my portfolio and and don't own and
- 41:32
- Unidentified speaker (UNKNOWN): don't
- 41:32
- Unidentified speaker (speaker_1): have interest in And owning, just because to me it's like a
- 41:35
- Unidentified speaker (UNKNOWN): uh my
- 41:36
- Unidentified speaker (speaker_1): my ninth best idea is better than my twenty 21st first idea. best and I'll put more money in that. Like that is my philosophy. I know that's highly, highly subjective and polarizing. If you disagree with me, no worries. I still like you just the same. But for me, that leads me to a long way of saying bullish, um, not not a shareholder. Um, would be looking kind kind of like the other infrastructure names we've done, would be constantly looking over my shoulder, um,
- 41:56
- Unidentified speaker (speaker_1): searching for real signs of things peaking, um, not of a key keyboard warrior on social media saying, 'I saying, just I got
- 42:02
- Unidentified speaker (UNKNOWN): to.'
- 42:03
- Unidentified speaker (speaker_1): a black wall GPU for 10 cents less, so everybody panic. Um, not really bright people like like Citrini, maybe taking the opposite sides of trades and unconditionally following them into that. Maybe that is the right decision. Um, but to me, I haven't seen a ton of signs of that being the right decision. Maybe he's as better at data and seen something. Um,
- 42:23
- Unidentified speaker (speaker_1): but to me, it still is lean bullish, but I it's it's it's bullish with such a short leash um that it would really not take much for me to to move to a more neutral stance here.
- 42:33
- Unidentified speaker (UNKNOWN): I
- 42:34
- Unidentified speaker (speaker_0): mean, I think you said it really well. Like I'm bullish on AMD. Well, I own AMD. And the reason for there are two totally different things. Because for me, like this AI drawdown, like I have so many names, not just AI drawdowns, drawdown, it's just drawdown on growth. So many names down 70% from their local. highs. Like Rock Club, for example, it's at $45.
- 42:54
- Unidentified speaker (speaker_0): Um, Axon today. Like it's trading at 20 times EBITDA. Like you have there's so many
- 42:59
- Unidentified speaker (speaker_1): what a fun week, Shy. What
- 43:01
- Unidentified speaker (UNKNOWN): a fun week.
- 43:02
- Unidentified speaker (UNKNOWN): There's there are
- 43:03
- Unidentified speaker (speaker_0): so many growth names that are like in that $20 to $40 billion market cap range that I absolutely like I see a five fag five bagger potential. Like if my thesis works out and the market's giving me an opportunity for a really great asymmetric return on those. AMD $750 million billion
- 43:23
- Unidentified speaker (speaker_0): dollar market cap. Could it double? I think so. But I think the risk and reward of the opportunity right now elsewhere in Growth Land is giving me 5x potential if I'm right versus maybe a at 2x potential. But let's just say talk about the business itself.
- 43:38
- Unidentified speaker (UNKNOWN): I
- 43:38
- Unidentified speaker (speaker_0): could not be more bullish on AMD right now as a business. They have two very clear growth engines at the same time. The GPU business is finally moving into real multi gigawatt deployments, while the CPU business is getting an unexpected second life from Ingentic AI. And we get that that combination is a lot more powerful than people
- 43:58
- Unidentified speaker (UNKNOWN): Like
- 43:58
- Unidentified speaker (speaker_0): actually give it credit for because they treat AMD as the second best experience. accelerator vendor because even custom silicon programs can create host CPU demand for Epic, while coding agents and Rock M are gradually lowering the software friction that have made NVIDIA so powerful. But obviously the risks are still absolute. Like free cash flow really
- 44:18
- Unidentified speaker (speaker_0): does need a catch up. Like margins need to expand. I also think nobody's talking about the warrant dilution warrant warrant dilution dilution that's going gonna to come down the pipeline. But
- 44:26
- Unidentified speaker (UNKNOWN): you
- 44:26
- Unidentified speaker (speaker_0): gotta listen to Lisa. Lisa's telling us that server um revenue is gonna double next year, data center AI is gonna more than double, company revenue grows more than 80% in 2027, demand is still way above supply. I have a very difficult time treating this one violent political sell-off as evidence that the AMD business itself is cracking. Uh,
- 44:46
- Unidentified speaker (speaker_0): I just think that they are in such a well positioned right now for this AI economy that.
- 44:52
- Unidentified speaker (speaker_0): it's going to be up there for one of the biggest companies in the world. Uh, but am I buying it right now? What capital? I don't have dry cap any dry powder anymore. That's the issue I have right now. It's like, what do I like if I want to buy into AMD a brand new position, I have to sell something else at a higher
- 45:11
- Unidentified speaker (speaker_0): conviction and just due to the asymmetric opportunity. I think that's like kind of. the dev the issue right now. I have, but regarding the business, man, it's humming. If I I think AMD is going to become the broad second compute platform in a world where the architecture keeps fragmenting across GPUs, CPUs, custom silicon, agents,
- 45:31
- Unidentified speaker (speaker_0): and edge inference. And I think this recent quarter gave us so much evidence that it can actually become that company. Uh and the market's giving. you a gift right now, a bit on a p violent political sell-off is ignoring all earnings. So essentially, what they just reported, which was a home run best earnings report in company history, you're getting that for free in a way because the
- 45:51
- Unidentified speaker (speaker_0): stock is lower than it was when they reported that earnings. I think that's kind of where the opportunity is. But uh, this was a fun episode. I know I I will say, like, guys, like Brad and I might be a little joyful. Uh I was making a joke with Brad. I was like, all right, just put on brain face praise face on for the podcast because we are getting punched in the face right now by the
- 46:11
- Unidentified speaker (speaker_0): market. So we can't show that. It's a difficult time, guys. I'm exhausted. Yeah. You can see like Brad is uh not sleeping. He's exhausted. So it's it's a tough time. Like we feel this like angst at bet with the market, but this is also the opportune time to really dig into the companies are winners because winners
- 46:31
- Unidentified speaker (speaker_0): and losers are not being separated right now. Eventually the market's gonna stop being going with the vi uh the political
- 46:40
- Unidentified speaker (speaker_0): annoyance of this AI build out, and eventually they're gonna reward the fundamental winners. And when that bounce comes, because it will come, this is just gonna be a temporary moment in time. The companies with the best earnings are gonna bounce the hardest. So that's why we want to create this video. Andy is one of the best earnings this cycle. I do believe it's gonna be one of those
- 47:00
- Unidentified speaker (speaker_0): companies that have the most violent bounce back. And when it does, hopefully this video created all the conviction you guys need to potentially invest in the company. Not non-financial financial advice. Uh any final words, Brad?
- 47:13
- Unidentified speaker (UNKNOWN): No,
- 47:13
- Unidentified speaker (speaker_1): well said. Thank you guys for tuning in. and see you next time.
More from The Deep End: Stock and Earnings Analysis from Futurum Equities
Zeta Beat and Raised Again. Why the Stock Still Looks Fully Valued | Ep 15:
Episode discussion: Analysts argue Zeta Global is fully valued at 30x earnings, citing lack of proven AI monetization and high growth requirements.
Speakers’ claims: Speakers evaluate Zeta Global as a marketing technology company with a unified data cloud and AI agent, Athena. They acknowledge Zeta's strong data moat and partnerships with Palantir and Snowflake but remain skeptical of its current valuation. The hosts argue that Zeta trades at a premium 30x earnings multiple that requires 25% organic growth to justify. They contrast Zeta's unproven AI revenue streams with competitors like AppLovin and Meta, suggesting the stock prices in too much future narrative.
Palantir Might Be the Most Expensive Bargain in the Stock Market
Episode discussion: Futurum Equities analysts argue Palantir is undervalued based on PEG ratios, margin expansion, and AIP adoption.
Speakers’ claims: Hosts Shay and Tom Nash of Futurum Equities argue that Palantir Technologies is currently undervalued, citing a PEG ratio of 1.3x compared to peers trading at 4.5x-4.6x. Nash highlights Palantir’s operational leverage, noting operating profit margins expanded from -107% five years ago to 43% in 2026, while stock-based compensation dropped from ~50% to 14%. The speakers emphasize the 'boot camp' deployment model and the AIP platform’s ability to integrate proprietary data, which drives net revenue retention above 150%. They contend that cheaper AI models act as a tailwind via the Jevons Paradox, increasing demand for Palantir’s governance layer. Finally, Nash presents a DCF model yielding a $917 price target and expresses confidence in Palantir reaching a trillion-dollar market cap.