Musk v Altman Takeaways, Apple’s iPhone Revenue Surge, Shopify’s Increased Fintech Ambitions

0:13Hi, everyone. Welcome to the Informations TI TV. My name is Rocket Drew. It is Friday, May 1st, and we have a great show lined up for you today. First up, we're going to break down Apple's latest earnings with an analyst from UBS. I'll also chat with the popular author of the Marcus on AI Substack, Gary Marcus himself, about the Musk-Altman trial, why OpenAI's models are obsessed with goblins, and the perils of coding agents. We'll wrap up with some exclusive reporting from the information about Shopify's big push into financial services. It's going to be a great show, so let's get into it. apple's iphone business rebounded in the march quarter growing nearly 22 percent from a year earlier revenue in china also rebounded jumping 28 for the quarter investors are now looking ahead to a major leadership transition with longtime executive john turnus set to take over as ceo later this year here to discuss all those results is david vote a u.s it hardware analyst at ubs Welcome on the show, David.

1:17Great. Thanks for having me. What was your absolute biggest takeaway from the earnings call? Yeah, I think two things. One, the iPhone portfolio and the Mac portfolio has been incredibly resilient in the face of economic challenges across the globe. So I think that surprised investors to the upside, both on iPhone demand and Mac demand. And the second most important point is the company generated relatively strong profitability. Again, despite sort of the macro considerations of higher memory costs across the globe. So net net, a very positive quarter for the company, and we need to check more to continue as we go into June.

1:51Yeah. Let's talk a little bit more about the iPhones. So there were a few years of slow growth there, and in the last couple quarters, we've seen some renewed growth. Why is demand so high lately? Yeah, I think demand is being driven by two factors. One, a revamped portfolio. So the specifications of the iPhone 17 have clearly resonated with consumers. Better camera, better battery life, and other sort of features that I think customers want. The second consideration, I think, is really a little bit more company-specific. Their ability to get plenty of components when needed. So unlike other companies that are struggling to get memory, Apple is winning in market share, and we would expect that to continue.

2:30So it's a combination of company-specific development as well as macro challenges for other competitors. And how is it that they're winning in memory? What set them up to do that? Yeah, if you look over the last decade or so, Apple has built an incredibly resilient supply chain. They're one of the leaders in terms of the Southeast Asia supply chain, North American supply chain. So their expertise that they've built up over a long period of time has allowed them to really have strong relationships with the big memory providers. And quite frankly, they've just been at it longer, and they're better at it than most of their peers, including Samsung.

3:01Got it. But Apple says it's still working through some supply constraints. for Macs. So what is that telling us about demand and their supply chain? Yeah, so Macs, actually, it's a little bit of a different story. The demand for Macs have been incredibly strong given by the Neo and the Mac Mini. So the supply chains are really a function of the company not really preparing for the overwhelming strength that they've seen. So it's not a supply chain resiliency issue. It's more of a, you know, basically a demand environment that was much stronger than anyone had anticipated. On iPhones, it's more memory and advanced node considerations, but max the strength is just really the issue at this point.

3:36Interesting. What about in terms of China? I think that market has been a little bit of a concern in the past, but we saw some solid growth this quarter. Is there a turnaround that's happening there? Yeah, in China, Apple has done a really good job of winning back share, particularly in iPhone. This quarter, China, the Greater China area grew revenue at 28%. What you've seen in that market is, again, the customer proposition has really resonated. The product's good, the pricing is appropriate for the market, and they've taken back some share that they've lost to the local Chinese vendors. We would expect that to continue as well, given the memory dynamics that we talked about earlier.

4:10I see. On the earnings, we also heard a little bit about this net cash neutral target that they have. Can you explain that to us? Yeah, it's a good question. So over the last decade or so, the companies had a focus on reducing their cash balance and really using that to draw shareholder value in the form of buybacks. We hit that sort of net cash neutral break even point effectively this quarter. So what the company I think is going to do going forward is really still be aggressive on buybacks, but really giving themselves the flexibility to think about capital structure in a much more tactical way.

4:42So for example, if the company has debt coming due in the next couple of quarters, they don't have to be dogmatically stuck to that framework. They could think about appropriately using their capital structure, whether it's refinancing debt, issuing debt, whatever the case may be. So I think it just gives them more flexibility and optionality. it doesn't change our view that they're going to buy back stock aggressively, and the company's still going to generate significant returns in cash flow going forward. Yeah, I mean, the company's been a cash-generating machine for years, but do you think this could also signal a move into higher business spend, potentially on AI or M &A?

5:14Is that something you're watching out for? Yeah, M &A is probably a low-priority opportunity for Apple. The biggest deal they've ever done in their history is the Beats deal many years ago. That was the predecessor to Apple Music. When you think about investing in AI, though, you can see it in the numbers today. They spent a lot of money on R &D. So this quarter and the June quarter, for example, R &D spending is going to be up over 30 % year over year. So that's where you're seeing the spend today. It's in the R &D line. CapEx and other investments, I think, are going to remain relatively modest compared to history.

5:45So the company is investing for the future, but just through the R &D line as we sit here today. Got it. Well, strategically, what kind of AI investment should Apple be making right now? Is there anything more specific? to say about that? Yeah, look, they have the dominant edge AI platform potentially, right? There's over 1.5 billion iPhones in the installed base. So we think ultimately that's going to have a tremendous amount of value in the consumer's hand. The question I think most investors have or we have is what does that actually look like? So there's partnerships that they currently have in place with OpenAI.

6:15There's the expectation that the Gemini relationship will expand fairly meaningfully at the developer conference in June. It's still unclear what the consumer-facing applications are outside of large language models that win. But given the install base in Apple's engineering talent, we think they have a very good shot of being very successful. So TBD, but we'll know more at the developer conference in June. Yeah, absolutely. I mean, I could see so many potential applications for it. The other day, I was parking my car on the street and I looked at the parking sign and said, I wish I could just get a notification on my phone when it's time for me to move my car, you know, but Siri is not really equipped for that.

6:50Not quite there yet. Yeah, Yeah, not quite there yet. But you see the potential for sure. Exactly. We got to talk about the leadership transition. So how are you thinking about that? Are there things you think could change under John Ternus? Yeah, look, I mean, I think John is going to be a capable leader. He's done an amazing job currently in his role within Apple. He's been a long time sort of engineering talent. I don't think much changes from the company's perspective, right? This is a company that is very thoughtful, very forward looking. They set plans out one, two, three, five years in advance.

7:21So I think John will be a continuation of what Tim and the board of directors had put forth over the next couple of years. And remember, Tim will be with the company for quite some time. He's still not stepping down as CEO until September. He will be executive chairman. So I think the strategy is unchanged. I would also point to when they changed the CFO fairly recently, the system of financial controls, the communication from a financial perspective has not changed one bit under the new CFO, relatively new CFO. So I think we're going to see more of the same. Over the long term, does John want to put his imprint on the business?

7:52Absolutely. It's a little bit too early to make a speculative call on what that might be. I think right now, the status quo is probably the appropriate course of action for the company. You expect a lot of continuity. It's possible that in the long run, he will have some different priorities. What else do we know about his differences in terms of personality or how he has helped run the company in the past? Wow, that's a tough question. I would say engineering-wise, he's a little bit different than Tim Cook. Tim is much more of an operational supply chain expert, and he was what the company needed at that point in time when Tim took over.

8:25And let's recall, Tim Cook has been the CEO for about 15 years now, so an incredibly long tenure and an incredibly successful tenure in that of itself. John is a little bit more of an engineering background person. So obviously, he's been very heavy-handed in sort of the product roadmap of the company. So we would expect to see more of that. But that's not to say that he doesn't have software ability and software expertise. So I think you're going to see kind of the same kind of view. I think John potentially is the right person at this time. We're going to give him a lot of rope to kind of take the company to the next level.

8:55But obviously, AI should be the real focus of what John and team are going to focus on. And we'll see how that develops over the next, you know, call it six to 12 months. Perfect. Well, thanks so much, David. That was David Vogt from UBS. Elon Musk has been testifying this week in his legal battle against OpenAI, claiming that OpenAI breached a charitable trust by operating as a for-profit instead of a non-profit. I've been in the courtroom all week, and I want to bring on someone else who has been closely following the trial. That's Gary Marcus, NYU professor emeritus and author of Marcus on AI on Substack.

9:27Welcome on the show. Thanks for having me. I was not in the courtroom like you, so I've only watched secondhand, but I've watched as closely as I can. Ken from the second. Yeah, it seems to me you've been following as closely as anyone. So we have to start by talking about the trial. We're one week in. You had a few takes on the trial that you published the other day, which is great. You're doing more than following. You have some serious opinions about what's going on. So can you walk us through these three takes of yours that you wrote? I don't remember them numerically, but I guess the biggest take is that this trial is really about whether the company was founded as a nonprofit and what that means.

10:07But you also have Elon Musk in the room, and he will win the trial or might win the trial if he makes it about what OpenAI did. But if the trial becomes a referendum on Elon, he might lose the trial. And I'm afraid that from his perspective, his ego may be standing in the way of that. He should put himself less in the picture. It should really be about what Sam and Greg did, not what Elon did. And we'll see how that goes. Yeah, absolutely. And I think OpenAI's lawyers are pretty good at getting under his skin, getting a rise out of him when he's being cross-examined on the stand. So you can imagine they're able to throw focus away from OpenAI's early history and paying more attention to Musk himself personally.

10:52I mean, I think he's right that they're trying to trick him. I guess that's what their job is, right? And he's getting agitated by that, and that's calling more attention to him. I think the facts are the company was founded as a nonprofit for the benefit of humanity. I just read something that as recently as 2020, they were writing to Elon saying, can you give us money for the nonprofit? They used that nonprofit status to recruit people. They got a tax exemption and so forth. That's what Elon should be talking about. He shouldn't so much be talking about himself and how he helped build it and so forth.

11:27But that's not really relevant and not really helping him. Is there anything that you're expecting to see next in the trial? For example, we haven't seen Altman testify yet. Are there arguments on OpenAI's side that you expect them to advance? What is OpenAI's case here? Well, I mean, I can sort of speak from experience having testified next to Altman and having had Altman testify next to me at the U.S. Senate in 2023 at the first AI oversight committee. And what I will tell you is that he is a master of projecting in sincerity, a master at telling the room what it wants to hear and not always truthful.

12:03And sometimes he can get fact-checked in real time and sometimes not. So, for example, in what somebody the other day called the most important or famous lie in history, Altman sitting right next to me said that he had no equity in OpenAI. But in reality, he owned at that time a venture fund associated with OpenAI. He did all kinds of deals with companies that he owned with OpenAI. And he owned indirect equity in OpenAI through Y Combinator, which owned equity in OpenAI. And he didn't disclose any of that. And the senators were like, wow, you're doing this just for health care? And he let that slide and let a story go through that wasn't true.

12:40He also said that he supported artists getting compensation. And then he told the British government, we shouldn't have to pay copyright for any of this stuff. or pay licensing deals for any of this stuff. So, I mean, he said he supported regulation. He said he supported my plans personally. And now he's, you know, supporting lobbying against any regulation. So Altman will tell people what they want to hear. He will come off well, and it will be up to Musk's lawyers to come back on Altman and try to expose whatever might not be true. Right. Well, it's a very adversarial environment to be testifying like that while you're being cross-examined.

13:15And the lawyers have access to all of his past statements, his deposition, and they can point out to him his past statements as well. I guess one of the ways that OpenAI might defend Altman from some of those claims is to say that even if Altman had a stake in some companies that did business with OpenAI, that those deals, those transactions were still in the best interests of OpenAI, that maybe they were done in an arm's length way, but there was no real conflict of interest at play. I imagine that's how - I mean, if you believe that, I can find a bridge for you already installed. Goes from Manhattan to Brooklyn.

13:48It's really beautiful. It's built over a century ago. It's a really great bridge. I've been on it many times. I like biking on it. Would you like to buy it? I'll give you a discount. That's very kind of you. You've also said you don't really like either side in this case, but you think Musk has a point. I've heard that from a number of people. Can you explain that to us? Why don't you like either side? Why do you think Musk means? I think they've both been dishonest. I think they've both traded on promises they don't really know how to deliver. I think Elon, what he did with USAID alone cost many lives.

14:24I really don't like him. I'll be honest about that. I don't like Altman either. He lied literally next to me. And he did it so effectively that I believed him at the time. I feel betrayed in a certain sense. I said a nice thing about him to the Senate. And I realized that in hindsight that he tricked me. I don't like that. So neither of these people are are you know going to get christmas presents from me let's let's be honest um but i call them like i see them i've criticized elon many times on twitter um she happens to own i've criticized all of them many times but i think right here on substance i can't really speak to the legal matters but at least kind of the the spiritual question of did open ai sell itself as a non-profit for the public good and then change that well clearly the answer to that is yes so you know i understand why elon is perturbed about that i have to leave the legal questions about what one can do with a non-profit legally and and so forth um and what the nature of their contract was to other people um but in terms of the underlying question did they take this non-profit and turn it into a for-profit exploiting a lot of people on the way in various ways like people who went there to work for a non-profit and then wound up in this other thing.

15:43I think there the matters are clear. And unfortunately, that leads me to kind of sound like I'm an ally at Elon Musk. I certainly am not. Sure. Well, you're not alone in your concerns about his political actions. I think a number of the members of the jury share those concerns, and they said so while they were being selected for the jury. They should. Some people had very vocal concerns and got weeded out, but ultimately people were able to say, I have concerns about his political actions, but I can set those aside that we're focused on business here, not politics. So I think that's where they landed on that.

16:16And if I can just drop in one thing, I actually wrote an open letter to Elon yesterday on Twitter, challenging him. So I'm not like all Elon all the way. I said, if you really care about AI safety, like you're telling this jury, then you need to tell the Trump administration to regulate AI. We're not going to have AI safety if AI is unregulated. So I mean, even as people think, oh, here's Gary Schilling for Elon. I'm challenging him, telling him he's not doing the right thing. So the reality is the world is complicated and different people have different positions on different things. Elon is right about the nonprofit and he's wrong to no longer take AI safety as seriously as he once did.

16:51Yeah. Well, OpenAI's lawyer has really taken him to task over that. Musk is making these sweeping declarations that AI could kill everyone to the jury and to the judge, but then he can't answer some very basic questions about AI safety. He's unfamiliar with the right. Yeah, that was embarrassing. Yeah. Yeah, that's embarrassing. I agree. So when you say you're sort of rooting for Musk, it sounds like, you know, not as a legal matter, but in terms of the spirit of the agreements as you understand them, you think he has - I'm not rooting for Musk, just to be clear. I am rooting for the redress that he wants.

17:24He's not even in it for the money directly, although we could talk about his very complicated motives. But he wants the transition from nonprofit to profit to be voided. it. That is what I am rooting for. If open AI, I mean, just think about it in terms of global good rather than the law, which again, I'm not really speaking to. In global good, if open AI was a firm that was working as a nonprofit in the interest of humanity, they would do very different things than what they're doing right now. For example, their product has many negative consequences. There's non-consensual deepfake porn. Maybe that's them.

18:00Maybe that's not a maybe that's not perfect example there's delusions is a perfect example right there's been a bunch of cases um of delusions that seem to be tied to their products and they don't want liability for that they're campaigning against liability if they were really a non-profit they would say hey we've got a problem here we should actually wind this down until this really is not causing serious harm to humanity really is doing uh benefit for humanity but instead they're driving a race forward with products that are premature and unreliable and so forth. And so the world would certainly be better off if Elon got that part of it.

18:34I don't care about the money, like these guys are all rich or whatever. But I would like to see OpenAI forced to be the nonprofit benefiting humanity that it was built on, that it told California, and that it told the world and its recruits. Got it. So you're saying setting aside sort of the legal merits of the case, just in terms of the remedies you would like to see, your argument is you think the world would be better off if Musk gets some of the remedies he's asking for. And in particular, you have in mind unwinding the restructuring that OpenAI did last year? Exactly. Yeah. I mean, I don't know if the law actually supports that.

19:08That's not my bailiwick. But I know that the world will be better off if that happened. I see. Well, it's certainly something that he's asking for in the case, in addition to a whole lot of money. And it's still unclear where that money would go to the nonprofit part that remains, not to his pocket, which is something I think people are not totally clear about. I mean, that's complicated because Elon would still stand the game, right? Because he owns an AI company that competes with OpenAI. So even if he got not one nickel directly, he would indirectly benefit from that. And I don't think we should get ourselves about that.

19:40Right, right. That is what he's asking for, is he wants the money to come from the for-profit arm and go into the nonprofit arm. That's right. OpenAI has some questions about that. Certainly the judge has some questions as well. all. Let's switch gears here for a second. You also wrote this week about an incident where an AI agent deleted a company's code base. Can you give us a little rundown on what happened there in that incident? I don't think it's the first time we've heard of something. I think it was a database, not the code base, if we're talking about the same one. I don't know of any incident where the whole code base has been deleted, but there have been actually multiple incidents where entire databases have been deleted.

20:16In this particular incident, The guy was a very, let's say, big fan of Vibe coding, had some competence. He was not straight off the street, had some understanding of what's going on. And the system deleted his whole database. Eventually, though he doesn't say it in the main text, he was able to get the company that operated the database to retrieve a backup. But it initially looked like there was no backup at all. And, you know, it's only because probably because he had a tweet that was viewed by 20 million people. that he was able to get it back. Like I think the average person probably would have been stuck on customer service, which we all know is terrible these days.

20:54I have a tweet about that too, but we won't go there. So that side of it was most interesting to me is it was a failure of guardrails. So like the system guardrails said, I believe, don't delete databases or things to that effect. And certainly they should have, these kind of coding agents. But the reality of LLMs is that they can't follow instructions very well. And that is a serious problem. That's why having automated agents is a dangerous thing. They can't really follow instructions. You know, earlier we saw, for example, Apple had in its news thing when they did quick headline, don't hallucinate.

21:31And then they got in trouble with the BBC because it was hallucinating anyway. It was there in the system prompt, don't hallucinate. But the reality is the language comprehension on these systems is limited and it failed. Here is another example. The agents are told, keep your code secure, don't delete stuff, but they don't really understand that. And so they're not really that reliable. So there actually been multiple incidents, but this one was particularly well documented. My tweet about it, I think got like 2 million views or something like that. I think it was something like total disaster, totally predictable.

22:03It got 2.3 million views, I think, by the last check. And that was about this particular incident, but it's not the only incident. And it was totally predictable. I did predict it. I wrote an essay in August after I went to the Black Hat Conference and learned how serious all these problems were. And I'd written a little bit before. It was called Coding Agents Plus LLMs Equals Security Nightmare. And it is kind of a security nightmare. This one was not about security, but database existence at all. But it's just a complete mess. These systems are not reliable enough to chain them together to do these things without real expert oversight.

22:40Even there, the problem is that experts can tune out. With a driverless car, not a true driverless car, but a driver assisted car, people tune out after a while. They trust the car and that's when they get in trouble. It's the same thing with these automated coding agents. Right. I think we've seen that with dangerously skipped permissions, which is the flag that you provide Claude code to get it to just take actions autonomously. People get tired of just approve, approve, approve manually. They're going to approve no matter what Claude is proposing. So at that point, they say, you know, I might as well opt out of the loop.

23:10I guess you see this incident as kind of a failure of instruction following. It's that the model didn't sort of comprehend what the instructions were that it was giving. Instruction following and also understanding the consequences of its actions, not really understanding that people might not want to have databases deleted, not really understanding that backups are always essential. I mean, stuff that a good engineer would know, it's just not really doing. I mean, these sound like capabilities issues. Do you expect that these issues will get resolved just naturally with scaling as the models improve?

23:41Well, you're a younger man. You'll see that. I mean, it will happen eventually. It might happen in the next several years. They're obviously actively working on it. But the problem is at the core, you have an LLM. What we're moving to now is actually what I've argued for, which is neurosymbolic AI, where you have LLMs and some symbolic code. But the way they're doing it is still a mess, and I don't think anybody has a good principled way to do it. And I think we actually have to work on our reasoning has to be much better. The system actually has to build a world model of what the user is trying to do, which I think is a very hard question.