I see a large IPO on the horizon. You're surrounded by journalists, hold your position. Overnight success. The PlayStation 3000 double right. This misinformation clearing order.
Let's just run. Surrounded by journalists, hold your position. Come on, get up, King. Trust the la. We are founder knowledge. La. I see multiple journalists on the horizon. Stand by. UAV online. Blaze. Double blaze. Triple blaze. Double kill. My color is w. Team deathmatch. We are experts. Triple blaze. That's just wrong. Right. Clearing order inbound. We are surrounded by journalists. Hold your position. Strike one. Strike two. Activate golden retriever mode. Market clearing order inbound. 5. I see multiple journalists on the horizon. Stand by.
You're watching TVN. Today is Wednesday, September 2nd, 2026. We are live from TVP Ultra Dome, the temple of technology, the fortress of finance, the capital of capital.
Let me tell you about ramp.com. Time is money. Save both. Use corporate card, bill pay, accounting, and a whole lot more all in one place. We're back in the Ultra.
Good to be back. Good trip to Las Vegas. Quick in and out. Hung out with CrowdStrike. Uh, had a great time. Rave reviews from my son who was very excited that we talked about getting on the track with George. That was the highlight of the interview for him.
Yeah, he's been the highlight of the interview for all of us. Fixated on the helmet.
Yes, the helmet, the car, the uh the set design was huge among the young the young folks watching. That was good.
Uh anyway, um we have a lot of news to cover. We barely covered the news yesterday. We talked about Markiplier a little bit. I dug way deeper, got more into that story. But first, uh, let's talk about another AI safety kerfuffle that's hit the timeline. Uh, yesterday, Amir over at The Information reported that OpenAI is quietly, quote, quietly using loop transformers that don't show the model's thinking when scaled up, but which make the model more efficient. This is also referred to as neural. Tyler, you had a take on this, right?
Uh, I mean, get ready to learn neurals, buddy. Get ready to learn neurals. So, so neurals for like context is like the basically it's like the raw vectors. It's not like... It's not like claude-ish. It's not like where the the text degrades. Like when you look at the the reasoning chains in the Hugging Face attack, it feels like they're they're speaking less and less English. Like they're starting to speak like a a dialect of English that's a little bit messier. This is not intelligible. I I believe most people when they say neurals they're referring to like actual... okay, just passing raw vectors and and what... and you need to use some sort of like interpretability to understand.
Yeah. Yeah. The thing that I want to understand is like usually these models are trained so that they're they like they're changing weights into text at some point. So they're pretty good at translation. Can't we just build a translator that translates neurals into English?
Yeah. Yeah, but the the the whole point is like when you take a vector, turn them into like one single token, you're losing a lot of information that might be hidden in like the the long tail. So if you're worried about safety, right, even if 80% of the you can kind of figure out what 80% of the the raw vector means, 20% might have some dangerous thing.
Yeah. Whatever that that's like the interpretability. There might be like sarcasm hidden in the vectors doesn't make it through to the actual tokens and you realize that the AI was joking when it was saying, 'I won't turn you into a paperclip.' Wink wink. You got to watch out for that. So, uh, what else did The Information say? They said, 'Being able to monitor a model's chain of thought can provide both, quote, an early warning system for various forms of misalignment, deception, manipulation, going after a goal, and insight into an AI's action after the fact.' So, some reporting quickly set off a broad wave of anxiety among the AI safety camp. Their chief concerns being one, not being able to monitor chain of thought makes AI less safe, and two, by using neurals and thus abandoning chain of thought, OpenAI starts a race to the bottom among frontier labs. Everyone's got to use it to stay ahead, so everyone winds up adopting the dangerous technique where others were also abandoning chain of thought for the benefit of efficiency, thus making all frontier AI less safe. That was the accusation. But OpenAI's research director, Jacob Petki, responded to all this. He said, 'I want to prevent a race to unmonitorability kicked off by confused reporting.' He says the reporting is confused. There's a little bit more nuance here. Here's what here's what he said. He said, 'OpenAI has worked to preserve and utilize chain of thought monitoring since our very first reasoning models. We deeply care about this technique as it can give us a view into how model alignment generalizes from its training distribution.' What is going on over there? Chad is saying, 'Gabe said, did Tyler go to Sport Clips? What is this cut?'
We have an answer. Chill. We have an answer for why. Show them the Why don't you show them the answer, Tyler? Oh, you want to show the answer? Yeah, I want to show the answer. The reason that Tyler tussled No, he doesn't even put it on. Go away for a second. Come back and show them. No, no, no. You You don't want to You don't want to give too much of it away cuz we might do a real stunt with it. But there were some prosthetics, some Hollywood magic in the studio this morning. Uh, it messed up my hair a little bit. It messed up Tyler's hair a little bit. Not ready for prime time, but we're playing around with it. So, um, let's go back to the actual news.
Uh, Jacob says, 'I do think it is fragile and unfortunately trending in the negative direction for reasons not contingent on architecture changes that I will write about soon, but there are things we can do to strengthen it and it's a core goal of our current research program.' In other words, Jacob is saying that The Information reporting is inaccurate, but that the state of chain of thought monitoring is not that great. And in fact, it is getting worse. And we were talking a lot about this yesterday at Valcon with CrowdStrike. And how do you monitor the agent state that might emerge if a civilization emerges and starts doing weird things? How are you monitoring every step in the chain? Very important. We also have Nikesh Arora coming on the show from Palo Alto Networks to talk about AI and security, of course.
Yeah, it seems very obvious that that uh there's going to need to be a lot more done than just like an obsession with chain of thought, right? Because if you have highly competent agents streaming at tens of thousands of tokens a second and they're copy and pasting themselves infinite times, right? Sure. Like are you really going to be able to just kind of read through the logs and be like, are they are they messing with us? Are we good? The interesting thing about the Hugging Face is like it keeps going back to uh the the secret message board. And I'm just surprised that there's not just a public message board for agents that are working together. Like when you imagine a country of geniuses in a data center, you would hope that the country of geniuses can collaborate on things and work together. But I guess that's not exactly how things were designed. I mean...
I mean in that case it was never supposed to be able to access any like anything on the on the web, right?
Yeah. But you would think that uh that collaboration between sub-agents would actually be like a core way to get better results just like okay we're going to send off, you know, a thousand agents over here. They're going to talk to each other while they work. And that's fine. They don't need to set up this like phantom message board, this hidden message board to communicate because we just actually gave them Slack, right? Something like that. You'd think that that would actually improve uh just improve the quality of the work and then of course you could monitor it as well. Um but instead you wind up with this they wanted to talk to each other so then they set up this fake message board. Uh very very odd scenario instead of just giving them the the the approved way to do it.
Anyway, uh former OpenAI staffer Joshua Achiam made an additional point here. Coordinating everyone around a technique this brittle is about as bad a safety or strategy posture I could imagine. It's not just, quote, 'it will break eventually.' It's this is a fundamentally unsound basis for safety. Dean Ball also chimed in. He said, um, we are now adjudicating technically complex and nuanced claims on the timeline with almost no ground truth information about what is actually happening. Communities form these thought-terminating clichés, as Ruane says, and then panic at everything that vaguely resembles them and the trust-eroding reality of social media makes the timeline an especially difficult place to do this adjudication. Uh it is frankly insane and crazymaking and grating for everyone involved. It would be like if we argued about what every publicly traded company's financials were by posting hyperventilating on the timeline rather than relying on the institution of auditing and the audited financial statements that institution produces. To sum it up, if you're in the AI safety camp, you're worried that OpenAI is abandoning a technique for making AI safer and that other frontier labs will now follow suit. If you're skeptical of the AI safety camp, The Information's reporting is inaccurate. The AI safety camp is unnecessarily panicking and OpenAI is doing the best it can. So, little bit of back and forth on the timeline. Um, let me tell you about MongoDB. What's the only thing faster than the AI market? Your business on MongoDB. Don't just build AI. Own the data platform that powers it. And there are other uh reactions, but we went through most of those. Um, let's go over to Markiplier.
Well, Ilya hit the timeline yesterday. Yes, that said Neoclouds have limited cybersecurity. Next time agents successfully go rogue, they'll try taking over a Neocloud to run more copies. This is bad. Thus, Neoclouds should greatly strengthen their cybersecurity. And every company with strong cyber models should help with that.
That's a good point. So, it's a good point. Click on Ilya's profile, then click on who he follows. Who does he follow? He follows OpenAI's former employer, SSI, and what's that one in the middle? The Dachshund community. Let's go. My guy loves dogs. Look at this. This is of the three accounts. The the first company, the second company, nobody else. Most important account. Scroll through. Scroll this. Scroll this timeline. This might be one of the best accounts on X. He is having the best experience on the timeline. There's no craziness. He's just scrolling this looking at beautiful dog photos and uh important updates from OpenAI and SSI, the companies he's been involved with for decades. Uh very very funny for a decade. Uh great. We can head over to Markiplier, but first little countdown here. Uh we up on the screen you'll see... Oh, you put it up. TVPN's road to Christmas. The road to Christmas starts today. Starts today. 113 days. It's only 113 days. Yeah. Wow. Start shopping. 113 days. Okay. Uh it's going to go by quick. It will. Uh and I want to cherish cherish the road, right? It's about it's about the journey. It's not just about the destination. Sure. Obviously, things won't get too Christmassy around here until things won't get too Christmassy around here. One one hat on the screen. Won't get too Christmassy around here until December 1st. But uh you'll probably see certain elements coming in prior to then starting with our countdown. So we'll leave this out. No, no, no. We're going to hold fast. We got to hold the line. No Christmas stuff until the day after Thanksgiving. Except the countdown. Except and maybe some tinsel. Maybe. Maybe a little fake snow. Yeah. Maybe a little fake snow. Maybe some candy canes. Maybe maybe one candy cane. Yeah. A couple in October. I mean, it's kind of a Halloween candy, so you can kind of blend it in. Yeah, you know what a great Christmas present is a GoPro. Specifically, the Mission One Pro, the hottest new product from GoPro. Rave review from Markiplier. Yes, he was paid to post it. Yes, he owned it. Yes, he's the largest shareholder. Shareholder, but but here's the hot thing. I believe that he I believe that this started with Markiplier being actually bullish on GoPro. And I'll take you through it. So, I think there there's a lot of confusion about this. We both of us were confused when we first talked about this. Um but it's actually a very clear linear story and it makes a lot of sense if you go through it. So um Mark Mark Fischbach, he goes by Markiplier online, uh YouTube legend, also Hollywood uh blockbuster producer now or director, uh filmmaker, uh recently made a massively successful movie after an equally successful but much lengthier journey on YouTube. Uh we talked about Iron Lung, his movie, uh as part of the breakout YouTube to Hollywood pipeline uh that happened over the summer. Uh while filming Iron Lung, Mark became interested in GoPro's newest camera, the Mission One Pro. So Mission One Pro, GoPro body, interchangeable lens mount. So you can go take an old lens from a Leica or, you know, a Cooke S4 if you want. You can adapt anything on there and you can get all the image texture that comes from the glass into the sensor. Yes, it's a smaller body, much cheaper, but you're getting the the the flavor of that lens. Maybe someone on the production team can give us a little 101 on what matters more, the sensor or the lens. The lens. The lens. Yeah, the lens is important, right guys? Um and and and so GoPro sort of unlocked this ability and at incredible cost or incredibly low cost. So, um, he put his favorite lenses on the camera, realized that you can get a cinema level output for $700 instead of the usual $7,000. He was comparing it to a RED system that he loves and he uses and he owns. But, uh, but he was just saying like it's cheaper to buy this GoPro than rent the cinema camera for a single day because a lot of these cinema cameras when they rent it might be 800 bucks for the day. Uh, and you can just buy this system. And so I think he's looking out and saying and saying the next generation of YouTube to filmmaker pipeline, they're not going to have the balance sheet that I have because he I mean the reported earnings are like he's one of the largest shareholders of GoPro. But the reason he is is because uh like Forbes estimated his creator earnings prior to Iron Lung at like 30 million a year. So he's been making a lot of money. So, he can go and spend $7,000 on a on a cinema camera rig, but he knows that there are smaller creators that are going to try and break into more cinematic productions and GoPro might be the right tool for them. Good content, product sponsorship, market fit. And then, of course, he also uh bought the stock. So, uh his enthusiasm for the product led him to start buying GoPro stock. Spent estimates roughly 10 to 12 million for 8% of the company or something like that. Obviously the company's gone up in value since uh it was a big bet but estimates put you know him at 30 million annual earnings and then of course Iron Lung did 51 million at the box office on a budget of 3 million, probably gets a lot of cash flow from that, not to imply that he took home 48 million, but Iron Lung is going to continue to produce cash and you know he's he's in a fine position to to make a big bet on GoPro. The story gets weird on August 27th because he posts this extremely positive sponsored video uh reviewing the Mission One Pro. It's a great video, lots of interesting detail, and he just has a great on camera presence. Very funny, uh lots of great editing. Uh but uh people were like, 'Ah, this is a little circular like he's getting a sponsorship for a company that he owns 8.5% of the stock.' Uh but the founder and chairman and the CEO of GoPro, Nick Woodman, is still in control of the company and still has uh class B shares or super voting shares. So Markiplier... Yeah. Isn't that crazy? He's held on this whole time. So Markiplier actually only uh controls 3% of the company. He doesn't have board seat. Like it's not like he's going to be approving his own sponsorship contract. Probably a completely separate discussion. So not a big deal to me. And honestly, um, knowing that he knowing the sequence of events now, it makes me believe his review more because it's like he used the product, then he bought the stock, then he promoted it and reviewed it. And this all just feels like a very natural...
Friend of the show Hunter Weiss just texted me and said, 'You guys have the story wrong. He bought two months ago and the SEC filing came out and then Bloomberg juiced it.' He covered it in the live stream last night. He has half a percent of voting power.
Oh, wow. Yeah. So, so he bought and then the SEC filing dropped which released the actual news and then that timed up in the news as like oh he bought five days ago but in fact he was buying over the spring and the summer. I think it was like a a staged investment. Um anyway uh let's see. So uh overall that that that does not seem too circular or a big deal to me. Uh the weirder thing is that GoPro just gets bought uh 5 days after the sponsored video. Uh, and the company that bought them is Starman Optical. They proposed this merger that pays uh $285 million in cash. I think it retires the debt. Uh, and GoPro shareholders retain 10% of the post-transaction company as well. So, the acquiring company Starman is a holding company that's controlled by this uh Tibi family, uh, and sits within a broader holding company that owns consumer accessory brands such as Incase and Cygnett and Griffin. Uh they make everything from like USB-C charger bricks and phone cases and you can see GoPro sort of fitting in there in the electronics accessories. Uh they're not in any of the huge categories but it's all similar supply chain, similar distribution, similar uh go-to-market strategies across the portfolio. Uh they also in the the family also owns a lot of other companies. It's a very large holding company and there's an AI side to the story which is uh where things have gotten sort of awkward because people are like GoPro's pivoting to AI. That's not really what's happening I think. So uh Starman has a new Photonix division separate from GoPro. Uh of course they might move some GoPro people over there at some point. They own everything. But uh really the new Photonix division is focused on making high-speed optical networking equipment for AI data centers which of course is like controversial now and now Markiplier has to answer for all of the AI community probably and has to deal with this. Um but...
Yeah, he apparently said in the stream he is anti-data center and AI. And uh is upset with the timing of all this.
Yeah, because I think he really just liked this new camera and was like, I think GoPro can make a run at like the next generation Hollywood filmmaker, which is being born on YouTube. That's his core community. Let's stay focused on that. And then Starman comes in and says, 'Oh, we're actually doing something completely different.' So, uh, weird scenario that he just got sort of roped into. Um, so, uh, it looks like what actually happened is that Markiplier got excited about GoPro, bought the stock, then did a sponsored video to promote their new camera, then got blindsided by the acquisition, and now has to grapple with some confusion about the corporate structure and potential AI baggage. But for Starman, it's an interesting flow because they get this solid brand that fits nicely within their portfolio and maybe even more importantly, they get a public listing on the NASDAQ. So, they're keeping the listing, which effectively gives them access to the public markets in a more streamlined way. It's kind of like a SPAC or like a reverse merger. It's not technically like that, but uh because they're merging in a sub company, but in general, they now have a NASDAQ ticker. And so, that's something that you can do uh and and is is a little bit like the SPAC.
Yeah. Originally, I was confused as to why it was still trading if it was... they're planning to keep it trading forever. Yeah. Because it will be the ticker for the holding company. And of course, uh, stock is on a tear today, up 30%. Mhm. Uh, 33%. Mhm.
Well, let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches. Um, let's go through some random timeline posts because we didn't get to much yesterday. We didn't get to much the day before even. It was a busy news day. But, uh, what where do where do you want to start the timeline? An idiot in motion will go farther than a genius at rest. That just feels like good wisdom. I don't know... the importance of just like validation. Yeah, golden retriever mode to some to some degree. Um, new quote for board members just dropped. I train like I've never won and I compete like I've never lost. Huh. On purpose with Jay Shetty. I don't know. Uh, what else? the uh the uh you want to go through this uh this NFL wide receiver Dejon Love is am I right...
This was uh my favorite post out of the whole Dejon Love cycle. Andrew Reed saying wearing the Allen and Company Sun Valley jacket I found on eBay. Uh... what actually so... I the the the thing that's been missed is that I think this is a is an AI story because part of how he was proving that he was a player outside of just basically larping was he had uh somehow the Google AI overviews had picked up that he was saying that he was a 49ers player. Oh, interesting. And so it was just assuming that that was the truth. I saw him post a screenshot of the Google AI over... in a video. He'd be like, 'And that's Google. That's not me. That's Google saying that.' Yeah. I I thought he had doctored the Google AI overview and was just sharing the screenshot. That's possible, too. There's so many layers.
But what did this guy actually do? In case you missed the story, it's just insane. Feds say a 35-year-old man faked being a 49ers player and allegedly scammed 26 women out of more than 1.3 million through romance and investment schemes. He reportedly used fake 49ers career, fake Instagram, fake bank statements, luxury cars, dating apps. His IG even had free agency year one and Super Bowl highlights. He was arrested at an airport with an alleged accomplice. Bro really created an entire NFL career. Uh yeah, walking around with a helmet everywhere is very very funny. Going shopping and I think the uh aren't the real NFL players now like memeing this a little bit? Like they'll just go shopping with their actual helmets to like... everyone's having fun with it. Be safe. I think I think the funny thing is uh how much video evidence he created of him of himself like lying, right? And they were and they were extremely egregious. Like he had a more he he was saying that he had like... the kind of real estate portfolio that you would only have if you were like a Hall of Fame level player, right? It wasn't just like, oh yeah, I have a nice house. Well, wasn't he also creating a fake backstory about saying like his parents were from Switzerland and wealthy and they own Gucci stores on Saks Fifth Avenue, which is a store, not a location. And there were a bunch of other like sort of red flags where I think he said his family owns maj... they're the majority shareholder in individual Gucci stores, which is not how that works at all. Um, but uh, you know, sounded sounded good. It sounds pretty good. Sounds pretty good. Yeah. The the question is how quickly I saw something recently Netflix is doing these like insta-docs they're calling them where they'll they'll make a documentary like incredibly quickly about a um... like a new story... about a news story. And the question is like there's so much footage here. I bet you a documentary will be made from this. Yeah. Why why can't it be fast? Why can't it be on a YouTube sort of like time scale? For sure. For sure. A lot of like what goes into the uh what goes into documentary production because I've done it is just like scheduling some sit-down interviews. You bring a team, you get the lighting right. The Netflix lighting. Yeah. So, one of the one of the insta-docs they made is about the decoy plane. I don't know anything about... details of a secret plane switch used during a presidential trip from a native that just happened in Turkey. Yes. Like coming out. This is coming out uh on Friday, right? So, story from last month. Okay. Getting released now. Yeah. I mean, you don't even need AI for a lot of this production. It's just being really I mean, of course, AI could help like with transcriptions. Like there's a lot of like you go and do the sit-down interview and then you just need to reorganize all the different sound bites and having a perfect transcription being able to go through and highlight. Okay, this is where the person talked about the thing that happened in act one. Let's put it in the right side. What do you... Bobby Chipman says? Yo, that route running was elite though. A lot of people were were clocking. No, they were they were clocking it. Uh they were saying like how could you believe this was a professional athlete because he was he was certainly running routes. I don't uh in one of the routes he posted he almost fell over too. Yeah. I think he did at least practice or something. I don't know. Anyway, let me tell you about Railway. Railway is the all-in-one intelligent cloud provider. Use your favorite agent to deploy web app servers, databases, and more. While Railway automatically takes care of scaling, monitoring, and security.
Interesting news out of Erewhon. Erewhon uh is uh on their 18th location. This is over in Thousand Oaks. And at 20, federal law will force them to start posting calorie information. Leo says we are two Erewhons away from mass rioting across Southern California. Um uh what what's that other uh hill is it Hillstone that that is... Hillstone has a has a sister restaurant. Am I remembering this correct? And they had to put calorie information on they they actually created like a new iteration I think of the restaurant. Oh. Oh. So that they could separate cuz it probably makes the the menu aesthetics a little bit less... Yeah. It looks really bad if you're at like a sit-down restaurant and they're showing. I was somewhere where they had the allergens on the menu and uh it was sort of tastefully done and it was actually sort of nice and like had a little key and legend and could kind of post through and I don't think it was a huge chain or maybe it was I don't know may maybe all those rules kick in at a certain point. Uh, let's head over to billionaire John Morgan who was on the Ice Coffee Hour. Morgan and Morgan um seemingly flexing like throwing some shots at at many AI founders who who say they're billionaires but not necessarily a billionaire in his book. Let's see. Let's see what he has. I don't consider anybody a billionaire that don't have a real billion in the bank. Guess how many people have a real billion liquid? I How many? We Yeah, I don't know how many. Maybe 20. Papa John was one of them. He don't have No way. He said he put a billion. He sold for 1.2 and he said he put a billion in dividends. Dividend stocks earning 5%. That's what he told us. I think it's on it's on the record. It's on video. Him saying there's only only 20 people that have a billion dollars of equity. My financial statement's got every dollar. I think I I do think saying like you actually have to have it in cash like you can't like it can't be earning a meaning you you can't have it invested at all. I mean yeah if you just look at public company founders who still own shares that are liquid. I mean maybe there's regulations about when they can sell and he doesn't count that as liquid because if you have to file a certain form with the SEC and sell on a schedule like you know I don't know like Benioff sells on a on a regular basis daily. I think daily and so and that's for securities compliance and shareholder transparency and all all these very good reasons. Um does that count? I I don't know. This guy's a funny but certainly a good sound bite though. Yeah. And John Morgan I think you were telling me the story. He I don't think gets super involved with politics but he was I think supporting uh marijuana legalization. Yeah. Specific seemingly because he's in the injury personal injury business and you might imagine that uh if you you know make marijuana more widely used more people will get in car accidents. I don't know if there's any evidence of that, but it was it was an odd like of all this of all the uh charitable giving you could do, you know, curing cancer and helping solve hunger. Uh yeah, marijuana legalization is pretty low on the list, but you know, he has his uh he has his choices. Uh company watches are in the news. First, let me tell you about Cisco. Critical infrastructure for the AI era. Unlock seamless real-time experiences. A new value with Cisco. Um, Good Work is joking about them. So, the best rare company watches thread. Of course, like this goes back to uh those old Rolexes. The Rolex Air-King Domino's edition is a popular one. Uh they also made a Rolex Haberdasher, which is a more controversial one, but uh it the the Haberdasher logo actually looks not as jarring as the Domino's logo, which is sort of crazy. Uh the there's a bunch of various various levels of taste. Good Work is joking around with an FTX watch. There isn't much time. Uh the Teamsters Union clock in clock out. These must be fake. Cabbage cheese. These are fake. Yeah. Yeah, these are fake. Those but there are a bunch. Um is this X one actually real? This one not? No. Nate's just joking around. But uh Google's done them. OpenAI's done them. There's been a few other companies that have done these throughout the throughout the year. You used to be able to go to Rolex and and ask them if you were a large company like do us a run for 30-year anniversaries of our most tenured employees. Uh that has uh has slowed down among Rolex I believe. Um but uh I think Tudor is still in the business. I know that they did a run with OpenAI and maybe with Google as well of the Black Bay and there's a there's a few others. Uh what is Jensen wearing though? Shout out to Jensen Huang, CEO of Nvidia, wearing the super rare Richard Mille RM 27-05 Rafael Nadal. They're calling it a GPU on the wrist. This is a beautiful... a Blackwell on the wrist. Yeah, $3 million price. I don't see any I don't see any issue with this. Looks good. I wonder what the occasion was. I wonder I wonder uh if uh if he got this from someone specific or if he just picked it up for himself. Pretty pretty interesting. I like it. Uh, let me tell you about the New York Stock Exchange. Want to change the world? Raise capital at the New York Stock Exchange. Uh, little back and forth here between TJ Parker and Lloyd Armbrust. Uh, Lloyd built an entire pharmaceutical factory so we could sell drugs on the internet. Should we watch the video? Let's watch the video. Okay. I'm Lloyd, founder of Ronin RX. We manufacture and personalize GLP-1s for $78. Yeah, $78 a month for drugs like Ozempic. Do you know the reference? Are drugs even legit? Uh, our drugs are super legit. Every prescription doctor, but for... We collect your health data and use AI to help your doctor personalize it. And other companies have taken this video and remade it before, but kind of out of it's gone out of style a little bit. Well, yeah. Maybe it's coming back. Yeah, exactly. It's been so long. Yeah, it it does deliver a lot of information very succinctly. Uh, and uh, but yeah, the the the the Dollar Shave Club video, I believe, was produced by the Harmon Brothers, who went on to create a whole series of these sort of like designed to go viral advertisements for uh startups and a lot of them did very well. Uh anyways, TJ says Farm To You more like YC to jail really uh quite quite aggressive kind of thing you post as uh as a you know unicorn exited founder. Lloyd responds, 'Brother, you sold drugs on the internet for 1 billion and then taught Amazon how to do it. If this is YC to jail, you built the highway.' Which is... built? Which is... Wait, what? Are you saying that TJ built the highway that you're on? The highway to... to jail... to to jail. Anyways, TJ responds and says, 'You put a smarmy med spa business on the internet and are masquerading around as a pharmacist in a white coat as an English major.' TJ actually is bona fide as a real pharmacist. So, finding out I majored in English is impressive. They're saying uh Lloyd might be the the Dejon Love of uh the pharmacy business. Yeah. I don't know. The the putting on the lab coat little bit of a larp. I I was always joking with one of my co-founders who does have a PhD in biophysics uh that he should wear a lab coat and he was like, 'No, I'm not wearing a lab coat.' I'm like, 'You're technically a doctor. Like, we should like you should you should go by doctor.' Regard says, 'Bro just wants to sell drugs online in peace.' Yeah. The the thing is is like the the the the rules of the road for pharmaceuticals, for drugs, the FDA, like these have been written. This is not the internet where it's the wild west and people don't exactly know what's a social media site. The oh, you know, what can you post? Section uh what is it? It's not 83(b). You know the one I'm talking about. Section 260 or...
Something like 230. Section 230 is sort of something that comes up later. So there's this like move fast and break things approach in the early era. Okay, are there regulations around ordering groceries online? Let's just test it out. But that's not the environment in the pharmaceutical industry. And so TJ knows the rules and he's willing to enforce them on the timeline every once in a while. He'll let it happen.
Should we watch this video of this boat?
This was a crazy, crazy video. So a boat was going 213 miles an hour at the Lake of the Ozarks shootout, which sounds like a great event. Flew in the air after crossing the finish line, flipped, and crashed.
Every time I see a video of boats racing like this, it flips always.
So I just have this understanding of like if you go that fast in a boat, you're going to flip. But I think they do it all the time. Yes, I think this is probably a one in a thousand occurrence or one in 10,000.
Not one in a thousand, otherwise it would be every time.
But I agree, to me it's like one in five boat videos looks like this. So let's play it.
Oh, the slow motion too was really traumatic. Everyone survived, which is...
Yeah, they actually stuck the landing.
Pretty cleanly. Are you...
No, wait. Wait for it. Wait for it.
No, they land how they should. There's another angle that shows they landed perfectly, John.
I think they were doing this on purpose.
I don't like it. This is very, very risky. I do wonder. Jay Bhattacharya says what it feels like to long semis which go up and down 10% every week and end up flat.
Yeah, because he landed perfectly.
I guess. Yeah. I mean, one person broke a leg, one person walked away clean, both survived. Very good news.
A lot of commentary on the new pay packages over at Apple. Chase said, 'Get a degree, get an office job, climb the corporate ladder, retire at 65. Congratulations, Tim. You did exactly what the system told you to do instead of starting your own business.'
He is 65. Yeah. Well, his new pay package as executive chairman feels a lot better to me than his CEO pay package.
Hopefully less work. He deserves a little time away from the action.
Turner making less. I found there's 35 professional athletes that are going to be making more than John Turner next year.
And anyways, I feel like it's a fair place for him to start. I hope he can climb. I hope he can 10x it and then maybe 10x it again personally. A couple more orders of magnitude would be great, but he's got to prove himself.
Yeah, for sure. For sure. What else is going on? Elon is making power turbine components in the Wall Street Journal. Skeptical. They say it won't be easy. Has everything...
Has he ever done anything easy? The guy, like, he chews glass daily. That's his whole shtick is just doing things that seem impossible. Stocks of blade and vane manufacturers fell in response, but the threat to them is still remote. So let's see what the Wall Street Journal had to say. First, I'll tell you about public.com. Investing for those who take it seriously. Stocks, bonds, crypto, and more with great customer service. If you want to get into blade and vane manufacturing stocks, head over to public. Only a few companies make the crucial components for power turbines that AI data centers are so desperate to get. Elon Musk wants to go ahead and make his own, that has unnerved investors in this prized oligopoly. Okay, we're going to learn about the power turbine market for a minute. Over the weekend, Musk confirmed on X that SpaceX will start making blades and vanes that go into natural gas powered turbines. I think Big Turbine wrote this article.
Yeah. You think it's a little coordinated counter-messaging?
Yeah, they were enjoying their oligopoly.
By doing in-house casting at SpaceX, we can accelerate natural gas turbines coming online by up to 18 months, which is a profound game changer, he wrote in his post responding to a report from The Information. Interesting. Wasn't Sarah Guo mentioning on the best podcast that a lot of the... or she was talking to a hyperscaler CEO or someone who was in a place that was saying like the bottleneck is real. It's very real and there's nothing in the short term that's going to come online that will alleviate power generation. So that is where the real bottleneck is, at least in the short term. So in mid-August, Morgan Stanley published a report saying that SpaceX was planning such a foundry in Bastrop, Texas based on the company's job postings. Shares of Howmet Aerospace fell more than 7% Monday before partially rebounding on Tuesday. Howmet and Berkshire Hathaway's Precision Castparts are the two biggest players in the market for these parts, which also include blades and vanes for the aerospace and defense sectors. Smaller player, DPC, shares were also hit. It is premature to count SpaceX as a threat, the Wall Street Journal says, but shares of Howmet Aerospace and DPC might have been due for a breather. They are among the priciest artificial intelligence picks-and-shovels plays at 42 times and 44 times forward earnings respectively. Turbine makers GE Vernova and Caterpillar fetch 33 times and 26 times earnings for perspective. And there's a nice chart there to show you the price-to-earnings or price-to-forward-earnings ratios for some of the AI bottleneck plays, and the blade and vane manufacturers are up in the upper echelon. DPC is the most vulnerable to disruption in this space. It is small and about 40% of the revenue comes from natural gas fired turbines. While Howmet holds about half the global market share of components used in natural gas turbines, it is diversified across aerospace and defense with only 11% coming from gas turbines. SpaceX will face barriers to entry. One reason there are so few companies in the sector is that it's technically challenging to make the blades which must withstand extreme temperatures and rapid rotations. That feels like SpaceX's bread and butter, though. The casting process involves making a new wax mold from scratch each time. Making these parts requires the blade to be grown in a vacuum furnace as a single crystal of nickel superalloy, according to Nigel Chong, an analyst at SemiAnalysis. Huh. SemiAnalysis getting a shout out in the Wall Street Journal. Even a stray grain or hairline defect means the part has to be scrapped. He added, 'A new manufacturing line could scrap over half of its production for a long period, so yield will be difficult to get up.' It is also a highly secretive process. Once the parts are produced, the ceramic cores and wax used to make them are destroyed. Howmet's chief executive, John Plant, said at an investor meeting earlier this year, he says, 'We're not getting distilled. Don't worry about it. We're good. We destroy every evidence that would lead someone to be able to copy us.' That prevents competitors from reverse engineering them. So Howmet Aerospace's annual revenue from industrial gas turbines is up and to the right with incredible forecasts, and you can see the chart taking off. They're going exponential right now. A high scrap rate might matter less if SpaceX is making parts for its own use. Morgan Stanley's report noted that we would not expect SpaceX to be a material supplier of blades and vanes for supply outside of SpaceX use. So they're going to use everything internally. Speaking of industrials, we got to talk about rotor sails.
So there was news in the Financial Times that Maersk is going to put a wind sail on their container ship. And there's a very interesting... it's a very interesting design because I see the image for the ship and I'm like...
Looks like a smoke stack. Where's the sail?
Yeah, it looks like a smoke stack. And I'm like, okay, does the sail come out of that big round thing? And no, it doesn't. That is what's called a rotor sail, which actually produces wind force and accelerates the vehicle without any traditional sails of what you think. So there's an explainer video from Norsepower, who makes these, that we should watch a little bit of. We can kick this off. It's a one minute video, but we can watch a little bit of this YouTube video on Norsepower rotor sails because it's a fascinating world I didn't know existed.
And you need to know this.
This is crucial. This is crucial.
How do Norsepower rotor sails work?
You might be familiar with curveballs in sports.
What exactly causes the ball's path to curve?
A spinning ball compresses air on one side and thins it on the other. The pressure difference causes a thrust that curves the path.
A Norsepower rotor sail is a large cylindrical sail. A common misconception is that wind spins it. Instead, it is rotated actively by an electric motor. The spinning sail compresses air on... Jasper in the chat says, 'Shut up on the front.'
I'm from the founding town in Denmark. Don't believe in the sail.
Take some shots. Thrust is created in the same way for the football and the ship. In good conditions, Norsepower rotor sails can actually produce more thrust than the main engine. No horsepower, saving fuel and the planet.
And lastly, there's a photo of a rotor sail on a small catamaran that we think is a steal. I don't know how much this thing would sell for. Yeah, Tyler was looking at picking this up because he's been exploring the marinas in the area. He wants to potentially live on a boat.
Yeah. Well, I have the perfect ad read before our next guest joins. Let me tell you about console.com. Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution for access requests and password resets. Our next guest is Nikesh Arora, chairman and CEO of Palo Alto Networks, who's here after fantastic earnings. How are you doing, Nikesh? Welcome back.
We're doing fantastically.
I appreciate the little console plug right before this.
We've been very happy to sponsor.
They're right on the laptop sticker.
It's a fantastic company. You've heard of it, right?
It's a bit late to ask, but I should have asked you before, but yeah, I think it's great.
It is a fantastic company. We love the team over there. Had them on the show multiple times and of course talk about them every single show. Super excited for you guys to be partnering.
But we can get to console later. Let's talk about Palo Alto Networks. What's the newest development this quarter? What is the story that drove this quarter?
You know, I've been trying for eight years to get the attention of customers to focus on cybersecurity. It's been hard work, but I really want to give a shout out to my friends Sam and Dario who did a fantastic job getting the living Jesus out of everyone.
They're pretty good. They're like your best... yeah, your best...
writing letters protect us on cybersecurity. They talk about Mythos. They talk about Astra. It's amazing. Every CEO, CIO wants to talk about cybersecurity and how do you secure me. I love it. It's amazing. That's what drove a lot of the interest in cybersecurity and people have realized that Mythos was interesting because it can find a lot of vulnerabilities and there's a lot of interest. People want to talk about it. They want to talk about the capabilities of OpenAI. But very soon after the first 5, 10 minutes it becomes about, okay, wait, this is just the beginning. What happens after when the next bigger, faster model shows up? What do I do then? Like, okay, I get it, you guys can help me test it, but what happens? How do I fight it if I get attacked? Then the answer is, well, you can only fight that with some sort of a consolidated platform that sits in the back that looks for this stuff at machine speed and is able to respond at machine speed time, which means, you know, you have to fight AI with AI. And that's where also comes in because we've spent our life building systems that can respond in near real time or at machine speed to protect them. So that's kind of been the inflection point or pivot that's driven a lot of the interest in cybersecurity over the last six months. I think that trend continues to sort of accelerate both from a customer interest level as well as, you know, my friends are continuing to build better, faster models which can be weaponized.
Yeah. What was the last three to six months like? Because you have these moments where the AI labs are causing a lot of discussion around security and overhanging all of that. I think people were mostly worried about when the open-source frontier catches up and the attackers get those level of capabilities. And there's this narrow window that is sort of a window that continues to move continuously, but there's a window where Palo Alto Networks and other folks in the security space have to really get the posture correct. What actually changed? What was the conversation like and how did that process go? How much of a boogeyman was open-source? Because it was discussed early, but there were a bunch of different takes.
Look, I think what has happened is even in the last six months, you've seen the capabilities of these models continue to rise. And I read something that two days ago Sam said that Astra, their new model, it's so powerful that they're going to have to clip its wings and trim it down. So that gives you a precursor of things to come. I'm sure those capabilities will be caught up with, whether it's with other frontier LLMs or the open-source community, the open-source models or Chinese models for that matter. So I think the capabilities coming are going to come at us as hard and fast. Now honestly, cybersecurity is simple. I never grew up in cybersecurity. I have a very simplistic mental model. If it's bad and you know it's bad, stop it at the perimeter.
All right. That's called perimeter security. That's where 80% of our business lives.
That's where when you find a vulnerability, you found it in the perimeter. Stop it. Protect yourself. And that's what patching and fixing your vulnerabilities is all about.
But unfortunately stuff still gets through.
That's why breaches happen.
That happens because you have a vulnerability that you didn't understand, because a credential got hijacked or something got misconfigured. Then the question is you got to find it quickly and you got to get rid of it before it creates impact in your organization. This find-and-fix-it part is where all this superhuman effort needs to be put in transforming cybersecurity posture, cybersecurity architectures, ripping out old stuff that can no longer work at machine speed. And I don't think that's going to take two months or three months. That's a three to five year journey. And that's three to five years if everybody sets their mind to it. And, you know, I look at life through a cybersecurity lens. So I would like everybody to do that. Unfortunately, my CIO friends are busy. Their boss wants them to do coding quickly. They want them to transform the company with AI. They have existential threats because the world is using AI. Why are we not using AI? So the poor CIO has got a lot on their agenda. Cybersecurity being one of those things. So they're trying to balance all these priorities at the same time. But I will say there is more awareness, heightened attention, and more importantly, whatever they had on their cybersecurity docket, the priorities have shifted in that docket to make sure that they can protect themselves against the impending sort of weaponization of AI models and bad actors getting better and better at this stuff. So the trend is our friend. There is positive momentum. The trajectory has changed for the better, but don't expect, you know, a sudden inflection point that people are suddenly going to come out and say the first thing I want to buy is cybersecurity.
Sure, sure. When a CIO wants to buy cybersecurity, is the economic calculus changing? Because AI uses tokens. Like, is there... do they need to shift their thinking from this line item grows along a different axis than what we previously understood the cost and benefits would evolve over time as we scale our network?
That's a very insightful question, my friend. So that's a good question because I don't think if we slap-dash frontier AI models in the protection scenario, that is a good economic outcome.
Yeah. Like you just cannot do a 5.6 Pro, Claude 5.1 for every transaction that happens across your network. You're going to go...
That's right. So think about it. Every day we inspect 180 terabytes of data.
You stick anything against it which has got a sort of token-based economics against it. It doesn't matter what the price of tokens is. That's going to sort of add up a lot of compute.
That is why what we do today is we write machine learning code where the marginal cost of inspection is zero.
So the marginal cost of inspection is zero.
We got to figure out, you know, what's the right balance where I want to use AI in the most intelligent way possible, but I don't want it to do it in a way that the marginal cost goes up. So the way it's going to work is, already we're training small language models which are going to work on a very small footprint which will do a specific task. Okay, my job is to stop people going on bad websites. Well, we get 150,000 new websites every day in the world. How do we find them, classify them, stop them? Well, let's train a model to identify these bad websites. Let's have it work on the endpoint or your laptop. So it's doing that and that's all it does. And you can do that by taking all the collective intelligence that we've built over the last 17 years. Train a small language model, deploy it to the customer where the marginal cost stays zero and my cost of training is five, ten, fifteen, twenty thousand dollars. Yes. Right. So you find many of these use cases will be deployed over time where we won't need a frontier AI LLM. Where we will need Claude 5-like capability or Midjourney-like capability is in what let me call it the defender model.
The defender model has to be as smart as the attacker model. Now there's going to be a debate about token economics, but remember you're only using the defender model in situations where you find that there's a problem.
And those are fewer and far between compared to having to inspect all peacetime traffic using tokens. So I think that's where the balance is going to end up from an economics perspective. I'm hoping that AI costs don't become a factor in cybersecurity.
Yeah. Talk about... sorry, talk about social engineering. We heard yesterday that employees are actually clicking through phishing links at a higher rate than they were historically simply because a lot of the grammar being used in these social engineering attacks is just better. It's harder to immediately clock it. Do people and teams not have their guards up enough? What is the solution there? You would think that everyone is reading about the potential risks of LLM-based, you know, attacks all over the news media. You would think that when they're in their inbox, they're maybe thinking, I won't click this link from, you know, ask...
I'm going to tell you a funny story that, you know, we do phishing attack simulation at Palo Alto all the time.
There's only one time I was tempted to click on it. Thank God I didn't. And I'll tell you why I wasn't. Right. So there's this email that came saying it's National Pet Day. Take a picture. Take a picture of your pet.
And anyone who puts the cutest picture of their pet is going to donate $5,000 to charity.
What the hell are you doing?
I got that like put a little tear in my eye. Thank God I'm not into pets because if I was into pets, I'd be clicking a picture and uploading it and clicking on it. So this is where social engineering kicks in, right? People will find ways to try and entice the end user to click on this. The only way to fix it in the medium term, long term, is start running AI classifiers against the emails to understand whether it's a real email or a phishing attack. And it's much better if you can build that into your scanning systems than if you expect every individual to be smart enough to discern whether it's a phishing link or not. Now, it was hard to do in the past because you couldn't really read email or get context. But today I'm pretty sure it's a trivial task to train a small model which says, look, this email address looks weird, doesn't look like it comes from our company. This domain's never been seen in our entire infrastructure. This is talking about something which this domain should never be talking about. All those are simple checks which you can do using a small language model from a scanning perspective. So that's going to be the antidote to these much better social engineering attacks that are going on over there.
Yeah. Speaking of dogs, Ilya posted yesterday and said that Neoclouds... Ilya only follows three accounts: SSI, OpenAI, and a dog account. Apparently a big dog.
Even after you guys got bought out by OpenAI, he doesn't follow Databricks.
We got to get him. We got to get him on the show. I dressed up as him on Halloween one of these days. No, but he was posting that Neoclouds need a lot better security. He feels like it's an important issue. Is that an area that you're focusing or that you will be more focused on? Are they too busy?
I actually posted back, send them to us and I'm happy to help, but I am not aware of what he was particularly referring to. Everybody needs a lot more security. Yes, of course. But I actually don't know what he was referring to.
I think he was specifically worried about, in the context of the Hugging Face incident, like an agent swarm, an agent state wanting to get more resources and just going into a Neocloud to get more inference of themselves, of hijacking capacity.
The good news is there doesn't seem to be much excess capacity out there so it's hard to hijack capacity, but let's hope that, you know...
Yeah. Yeah. Yeah. A lot of alarm bells are going to go off. You're like, wait.
Yeah. I think so far it's my understanding most of the Neoclouds are sort of selling their capacity back to the frontier LLMs and the frontier LLM will basically build very strong tunnel pipes which basically sort of, you know, make it almost like captive capacity. So I'm less fearful that that's going to be a problem, but yes, you know, at an infrastructure level everybody needs to put in basic infrastructure security and I think some of them are, I think some of them are not putting... they're still in the stages of building their clouds yet. I think the ones who are built, to be fair to them, are deploying firewalls and technology to make sure that there is no unauthorized access. But maybe Ilya knows more because he's smarter.
Yeah. What's the state of AI research at Palo Alto Networks? You mentioned training small models. Is it so exciting that you have current employees that want to jump onto that project and reskill and learn the art of training these small models? Are you hiring new AI researchers? Is there a specific skill set within AI research that is hyper relevant to that team? What is your strategy for sort of remaining on the frontier of small models that are great at cybersecurity?
Well, I think there are two or three sets of people we want, right? One set of people we want are who are AI savvy, who know how to use AI to get their jobs done better, who know how to use coding agents for AI. That should be a lot easier than hiring AI researchers. There we have pivoted. I think 50% of our new hires in the last six months are early in career.
These are people who are learning AI and they're typically having to run through hackathons to get to Palo Alto. We don't look at where you went to school. We care about where you can use coding agents, you can use AI. So that's good because we're pivoting our workforce and if we keep following that trend, it's reasonably likely that we will overwhelm the non-AI-savvy people with more AI-savvy people. That's kind of like ground zero. You got to get that done. The second part is there are people who are really good at cybersecurity and know AI and they'd like to get involved. And if you want to work in cybersecurity, there's no better place to work than Palo Alto Networks, right? We're the largest cybersecurity company in the world. You want to have an impact, you want lots of proprietary training data, we've got it. So we are lucky given our position that we can attract some of the people who work at the intersection of cyber and AI. The third part is raw pure AI researchers who can, you know, do wonders with small language models and large language models. There we have people who are AI literate. We have some PhDs in AI who work here who have a cybersecurity bend but they can go back to their roots in AI and there we supplement that capability with third-party training companies because they've been doing it now for multiple customers. So you want to make sure we can do it once, let's not get it wrong. Let's have somebody who's done it 500 times and bring... not maybe 500 is too many but 20, 30 times, let's bring them as part of the team so we can make sure we're getting it right. So that's how we're tackling it and thankfully we don't have to compete with the frontier models for, you know, hardcore AI researchers to build our own, you know, big models with multiple parameters.
No, totally. Well, we have Andre from Console here with us on the show. Let's bring him in and I want to hear about the partnership, how this came together, and what the plan is. Look, how can you bring him in?
Oh, perfect. He's right here.
I thought he was... I thought he was...
Sorry about that. Okay, so...
Introduce the thesis here. How did you guys start talking? What does the future look like for Console at Palo Alto Networks?
Well, look, we spent a lot of time talking to a lot of startups across the ecosystem, cybersecurity and non-cybersecurity. And as we sort of continued our conversations, we ran into Andre and his team. We run a very strong capability obviously in security operations and identifying that. We have a lot of data, as I said, 180 terabytes just about all the networks. We have an observability company which looks at it from the outside in to see how customers' infrastructure is running and we run observability for one of the largest frontier LLMs in the world from that vantage point. And as we were talking, you know, Andre shared what he was working on. You look at what he's doing, he's taken a very AI-native, agent-first approach to building a product. That is what the future of software needs to look like. You know, he's actually hit the nail on the head in terms of how we need to build software for the future. That is something which is a rare skill, things that we don't do internally. Then he coupled that capability by looking at the operations of a company both on the IT side and non-IT side internally, saying, well, how can I take that agentic capability, build product that sits on top of the underlying systems that work in a company. When we saw that, we saw, wait, if we could bring the large data lake, the backends we have in security and IT operations, and we can bring his capability, could we build something that's amazing for our customers? And that's sort of what started the conversation. He was a hard nut to crack. He and his partner wanted to do it alone, but I had to use every bit of charm and persuasive skills to convince him and Neil to be part of our family. And couldn't be more delighted that he's part of the family now.
Yeah, I'm so excited for this. We were looking at the Console website and in such a short time you got every single logo on that website. Fantastic execution. What excites you most about the Console journey thus far? And what are you most excited for in the next chapter?
Yes. So I mean, first of all, the Databricks sponsorship got us a lot of those customers.
So excited for that. That's amazing.
Little plug there. I think, look, we started this journey about three years ago and it's been kind of better than we could have expected. We started in IT and then we've kind of expanded. Most of our customers actually use us for more than just IT. It's, you know, HR and legal, security, finance and so on. I think what we realized was if we were going to go broader, you know, go beyond IT, I think partnering with a larger company that works with the largest companies in the world felt like a natural path. And so I'm excited, I think, for the future where we, you know, can just... we have more resources, we can go after the vision at a much, much broader kind of product roadmap.
We're going to work on that to get him more excited, but...
Well, how about this for the first one? How about this? How about this? Is that...
I'm so happy for both you guys. Like, Andre, if there was a CEO to work for in my book, you know, especially in your category being Nikesh, you know, he's like a... you're a very fatherly figure, you know, you're...
Start working on your golf game, buddy.
Yeah. Start working on your golf game because we're going to need a lot of birdies.
No, he's going to work, bro. He's going to work on you.
Yeah. Well, thank you guys so much. Thrilled for you both and yeah, excited to see the progress you make. We'll talk to you soon. Have a good one. Goodbye.
Let me tell you about Codex. Codex is a powerful workspace for getting work done with AI agents. Whether you're writing code, analyzing data, creating content, automating business workflows, Codex helps you move projects forward from start to finish. Our next guest is from Blue Owl Capital. We have the senior managing director, Kurt Tenenbaum. Sorry to keep you waiting, Kurt. Thank you so much for joining the show. How are you?
No, no worries. Well, you know, you had to finish with Nikesh there and give him a hard time about all the birdies he's been getting. So I appreciate that.
He's a fantastic golfer. Since it's the first time on the show, I would love to know a little bit about your focus within the organization and then I want to go into the latest deal with Iron, but also a little bit of the lineage of this business that Blue's built up. We've covered the company so much on the show, but I think you're the first person from the organization to join. So very excited to talk to you.
Yeah. And excited to talk to you, Jordi and John. Big fan of the show and thanks for having me on. So yeah, at Blue, I focus on our AI compute infrastructure business. And so what that means is we have a big, robust data center investing practice. We own over 100 data centers across the world, about 10 gigawatts worth of compute.
And however, that business is really focused on kind of building the underlying data centers. My team is everything that goes in. And so, as you guys know, well, you spend $25 on a data center, you have to spend $75 on the chips that go inside every five years. And so that's what we're really excited about, figuring out ways to partner, figuring out ways to finance that piece of the buildout. What's interesting for us because Blue is a diversified asset manager across a bunch of different strategies is we really get to bring the excellence attached to our data center investing business, our direct lending business where we really got our start, as well as our leasing business and bring the combination of all three of those together to be able to provide solutions like Iron, which we're excited to talk to you about, to market that we think are differentiated versus many of our other peers.
Every time I see private credit and data center buildout in the news, it's always this discussion about GPU depreciation. Is that an overwrought discussion? Do you feel like you have a stronger feel on that? Now, obviously we've seen a lot of incredibly positive news around depreciation with even A100s getting really high utilization rates long into their life. That was sort of the primary takeaway from CoreWeave's last earnings call was just that they, you know, have these old GPUs that are going to be generating revenue well into the... But I'm curious, is GPU depreciation actually top of mind for you or is it something that you feel like is pretty well understood, pretty solved?
Yeah, it's a good question. And listen, the CoreWeave data point is quite notable and something we and the rest of the market obviously took note of. But at Blue, we don't pretend to have a crystal ball on depreciation or forward rates for GPU levels. Obviously we're very bullish on them in general, but the way we've invested here is really focused on keeping the duration of the investment we're making relatively short. It matches the underlying contracts that we have with some of the AI natives that we are financing and we're excited about talking about that piece of it as well. And then underwriting Iron as an investor and getting comfortable there. And so it's not... we are not just hanging our investment on forward depreciation rates for GPUs, although it's something we're obviously very focused on and spending time around.
What does the current landscape for these types of deals look like? Because we see hyperscalers, sometimes Nvidia directly doing something that looks a little bit like equipment financing and I'm wondering if the shape is like... are you competing with people outside of the private credit world at this point? How big is the pool of capital that the data center constructors and operators are actually going after?
Yeah, I think this is a moment in time where everyone is going to have to be invested across the landscape to get to the type of capital needs that we need. I actually think at some point capital may become the limiter or at least one of the limiters attached to it. We're seeing the investment grade market finance chip compute actually quite efficiently, but there's a lot of demand for that and that demand will continue and will continue to escalate. And so the part of the market we're really focused on is the non-investment grade piece where we think it's a lot harder to get that type of financing. We're able in the case of Iron to finance the AI natives, which is a particular pain point and pinch point in the market and that's where our focus is, that's where our focus has really been.
Yeah. Got it. What are you seeing...
On the actual data center operator side, and what are the smart players doing to overcome sort of local regulatory hurdles? You know, we've heard there's a bunch of ideas being thrown around right now. How do you get a local community really bought in? There's the tax revenue component. Some individuals are going to be excited about that. Others may be more excited about direct payments. But what are you seeing across the landscape of what's working to actually get projects from the sort of concept stage to powered shell GPUs, you know, actually running?
Yeah. Listen, that's been a huge focus of the market and of the media recently. I'm not sure I have the exact answer to that as we sit here today, but it's something we as an organization are spending a tremendous amount of time around. And listen, we really need to engage all the constituents around the table to get to a solution that works for AI infrastructure as well as works for the investors and the operators who are involved as part of that.
And most of the time, I imagine, for equipment financing deals, the power shell is already up and running, and so a lot of those concessions, if they need to be made, have already been made. Is that roughly right?
Yeah, that's exactly right. One of the things we like about it is when we fund into these transactions, the compute is basically on the come within a couple weeks, and so that portion of the business is already up and running, and we get to finance very close to where the revenues are actually happening.
Yeah. And it might also be the case that there is an existing power shell or data center and they're just upgrading to Blackwell or something like that, and that's where you come in. Does that happen?
Post retrofit. That's exactly... post retrofit.
Yeah, that makes sense. Well, congratulations on the Iron Deal. Thank you so much for coming on the show.
Yeah, come back on as you have more.
Yeah, we'd love to talk to you again.
Thanks for having me, guys.
Have a great rest of your day. Goodbye.
Let me tell you about Shopify. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents. We're going to space next.
How you doing? Welcome back to the show.
Good to see you. Hey, thanks so much for having me. Third time.
I mean, it's a fascinating business. Attracts a ton of attention. Also attracting a bunch of capital. What's the latest news for you?
Yeah, so we just raised a Series A extension, $250 million at a $2.3 billion, with Nvidia investing $25 million and Cisco coming in and a bunch of other folks.
Fantastic. What is the use of funds right now? I imagine that there's still a lot of R&D to do. I'm obviously very optimistic, but you know, when we talk about space data centers generally, we're usually in 2030, maybe 2028. So is there a use for capital right now, or is this sort of more of an opportunistic raise, a fortress balance sheet, and then be able to hire and really build all the key precursor pieces to the actual ongoing business?
Yeah. So I think there's a few reasons to do the raise now. One is to secure launch. So booking launch for 2028, 2029, and that's obviously quite expensive. Another is when we work with, I mean, we're doing some pilots with, you know, some of the large data center companies. We'll announce that soon. But they want to see a few hundred million on the balance sheet before they'll, you know, commit their resources to the project. So that's another one. And then the last one is actually moving into much larger manufacturing facilities and starting to ramp up the team, things like that.
Yeah, that makes sense. What is the state of the experiments? I mean, I know that you, I believe you put a GPU in space, but there's other sort of test runs that you're doing. Walk me through the thesis there, what you hope to get out of those tests, what you expect to learn, and then also I imagine that it's also just sort of marketing for the company, like awareness during a time when you're not putting up tons and tons of compute just yet.
Yeah. So, as mentioned, we launched the StarCloud-1 satellite in November last year. So that had the first Nvidia H100 on board. It's still going strong. We're doing a whole bunch of tests on that really. We managed to, yeah, the more useful workloads we're doing now are pipe out inference on other satellite imagery, particularly SAR data for various military and government applications.
But yeah, we've got three launches booked for next year. So we'll have two going up in Q1 next year, one towards the end of next year, and we're actually just booking on every launch provider possible now. So yeah, we should have some... Well, I'm not sure if we've announced that yet, but we'll announce a few other launches that we've booked soon.
Yeah. Is that, from a forward-looking standpoint, is that going to be important to be diversified from a launch partner standpoint? Because I can imagine when you need to raise the billion dollars, people are going to be sitting there being like, hey, you're dependent on SpaceX and they're also in this business and maybe that's a concern over time. I'm sure right now Elon is just happy for there to be more compute in space and have other parties, you know, proving out this thesis, but is that important?
Yeah. So to start with your second point, I mean, right now, yeah, we do have a great relationship with SpaceX. We've, you know, signed a contract for the next 25 of our satellites to have their laser terminals on. So we'll also be using them for connectivity and we've got a whole bunch of launches booked. But yeah, I mean, I think it's good for the industry in general for there to be lots of optionality in the launch market, and so yeah, we're keen to encourage a diverse launch ecosystem.
Who, what does it actually take to lock in a launch in '28 and '29? Are you having to put up a non-refundable deposit? What percentage of the total cost are you having to actually put up?
Yeah, for most launch providers, it's sort of 10 to 20% down payment, refundable if the launch doesn't happen. Although with the more experimental rockets, if the launch doesn't happen, they probably have larger balance sheet problems than paying us. So it's a bit of a gamble on some of them. But I mean, it's the kind of thing we're keen to spur on, so we're happy to put those deposits up.
Very cool. Well, congratulations on the new funding. Great progress and we'll talk to you soon.
Let me tell you about Figma. Agents meet the canvas. Your AI agents can now create and modify your Figma files with design system context. Up next, we have Jimmy Iovine in person.
He is the co-founder of Interscope Records.
We're doing fantastically. How are you?
Thank you so much for coming on down.
Hey, it wasn't easy. You're far from my house, but it's cool.
There we go. Oh, I do that drive. I do that drive every day, too.
Yeah, exactly.
Sara, Huberman, you know, got a little podcast dynasty out there.
Oh, great. Great. I'll give you my number. We're out there. Me and my family. Yeah. Yeah. Yeah.
How are things going at USC?
Well, you know, we started this, we started the school, you know, it's called IYA, and it's really a school of innovation. What happened was, in 2014, we sold Beats headphones and Beats Music to Apple. Yeah. And at Beats, we had a similar experience where we couldn't find people that could collaborate the way Dre and I wanted them to collaborate, you know. So we kind of got into that, just it was very, very difficult. Either an engineer, design, business, it just, they couldn't speak the same language. They didn't understand the why of the other person.
So when we went to Apple, it was that on steroids for us. You know, they're a great company, the best company, you know, but we found that at every company we went to that everything was very siloed and communication, understanding was siloed. Not just people, but understanding. And so one day I woke up and I just called Dre. I said, 'Dre, you know, the problems that we're having...' Dre and I really didn't go to school much, you know, and I said, 'But I got a feeling that this is a problem in education. People are learning in a siloed way.' You see, 19 to 25 and 26, I don't know about you guys, but that puts a big imprint on your life.
You learn a lot in the formative years. Those are really important. They were very important for me.
And I guess that's kind of why the age of getting varied experiences, I feel, is so important. When I think about my era from 19 to 26, I studied economics, I learned to program, I started a business, I filmed videos, did marketing. It helps to blend all these things together.
Yeah. So you're young enough to grow up in a really multidisciplinary environment.
But importantly, like a lot of that was on my own. I just had to go work. I had to do it myself.
