OpenAI’s Windsurf Crash, Grok’s Wild Week, Replace Tim Cook? — With Aaron Levie

Box CEO Aaron Levie is back for our weekly discussion of the latest tech news. We cover: 1) OpenAi's Windsurf deal falls through 2) Is OpenAI okay? 3) What percentage of all AI spend goes to coding? 4) Google's AI code play 5) Grok 4 is out 6) Does Grok show the scaling laws are still in effect? 7) Would Box work with Grok? 8) NVIDIA hits $4 trillion 9) Are we in an AI bubble? 10) Should Tim Cook step down? 11) Could Apple merge with OpenAI? --- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. Want a discount for Big Technology on Substack + Discord? Here’s 25% off for the first year: https://www.bigtechnology.com/subscribe?coupon=0843016b Questions? Feedback? Write to: [email protected] Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Alex Kantrowitz?

  OpenAI's windsurf deal is off and the executive team is going to DeepMind? Elon Musk's Grok had one hell of a week. Nvidia becomes the first $4 trillion company. And should Apple replace Tim Cook? As some analysts are suggesting. That's coming up right after this. This episode is brought to you by Genesis. What does it actually take to put a Gentic AI to work across an entire enterprise? I recently spoke with Genesis chairman and CEO Tony Bates about why orchestration is emerging as a competitive advantage in customer experience and what companies need to get right in the AI era. Then I sat down with Adam Mitchell, head of enterprise business solutions at Voya Financial,

• 0:41Alex Kantrowitz?

  to go inside a large scale agentic CX transformation, including the business decisions, trade offs, and lessons along the way. You can watch both conversations now on my YouTube channel today. It's never been more challenging to get frontier compute. How do you know where to look? Who to trust? Or how much to pay? Akash is an open cloud that provides easy access, flexibility, and peace of mind. That's how companies like Razor were able to cut their inference cost by 15 times. Visit akash.network/alex for a free consultation. That's akash.network/alex for a free consultation. Welcome to Big Technology Podcast Friday Edition, where we break down the news in our traditional,

• 1:25Alex Kantrowitz?

  cool-headed, and nuanced format. Boy, do we have a treat for you because today, Box CEO Aaron Levie is joining us to break down this week's news. And we have a full slate, a more than full slate, because OpenAI's windsurf deal is off and the team there is going to Google DeepMind. We can also talk a little bit about Grok, the ups and the downs, big ups, big downs, Nvidia hitting $4 trillion, and then, of course, these rumors that Apple wants to replace Tim Cook. So, great to see you as always, Aaron. Welcome back to the show. Good to be here. What a week in tech. Absolutely crazy. Let's just start with the big headline first. This just dropped right before we started recording. OpenAI's deal to buy Windsurf is off and Google will instead hire

• 2:07Alex Kantrowitz?

  the Windsurf CEO Varun Mohan, co-founder Douglas Chen, and some of Windsurf's R and D employees from the company to join DeepMind. Google and Windsurf announced Friday. So Aaron, can you tell us about the significance of this Windsurf? What is Windsurf? What deal were they were going to have with OpenAI? And what is the significance that that deal is off and they are moving instead to Google?

• 2:33Speaker 01

  Yeah, in this industry at this point, you never get even just like one piece of news. It's always multiple pieces of news embedded in one major thing. So this is this is sort of a multi, I mean a multi-part announcement, I guess. So Windsurf is, has been one of the faster growing AI coding platforms. It's an IDE that is built off of VS Code that lets you have agents that automate your coding. And quite successful, particularly in the enterprise. They were, they were one of the first to really kind of nail the enterprise oriented sales motion, lots of protections for data and your code base that they had a good fit on. Varun's a fantastic kind of founder entrepreneur.

• 3:18Speaker 01

  And the expectation, I think, was that they were going to be acquired by OpenAI, help OpenAI really kind of boot up their coding efforts. And clearly that's now off. Obviously, the rumor in that process was there was some structural issues with maybe the Microsoft terms and different parts of that deal. It's, you know, no one has ever kind of explained exactly the problems there. But now with that deal off and Varun and team going over to Google, it's a recalibration of the market. It's actually interesting. So the thing that everybody should have been thinking this entire time actually was, where is Google in AI coding? Because right now you have, you know, Anthropic from a model standpoint tends to be seen as the leading model for coding.

• 4:03Speaker 01

  And then they also now launched Claude Code. OpenAI launched Codex, which is a very strong kind of offering in agentic coding experience. And so the odd man out there is Google, where Google has, you know, is a very deep engineering centric organization. And so one would have imagined that they would want to be front and center with AI coding. The Gemini 2.5 model is seen as very good at coding. But again, it's in this a little bit of no man's land because they neither have an IDE nor do they have like, you know, what tends to be the best coding model from Anthropic. So they had to do something in this space.

• 4:45Speaker 01

  And this is a pretty exciting move to launch into.

• 4:49Alex Kantrowitz?

  So the politics that you talk about between OpenAI and Microsoft, I am just going to imagine that Microsoft has GitHub Copilot, which allows you to do a lot of this AI-generated code thing. And the fact that it's invested all this money in OpenAI and has access, proprietary access to OpenAI's models, probably not such a fan of OpenAI going out and building a competing product.

• 5:10Speaker 01

  Yes, although it's not obvious to me what leverage they have in that in that dynamic. So I think that, I think, you know, everything at this point is basically just rumors and conjecture. It's very clear that

• 5:23Alex Kantrowitz?

  That's what we do best on this show.

• 5:24Speaker 01

  Yeah, exactly. Sure. Or basically our entire industry at this point. But I think, I don't perceive that OpenAI is constrained by anything strategically at this point, vis a vis the Microsoft relationship. So I doubt, you know, I think the rumor was more there was things like, you know, IP issues and other dynamics with the acquisition. But again, it's all rumors, so impossible to know. These, you know, deals fall through for a variety of reasons. But I would not be surprised if if OpenAI continued their motivation for needing to be in this space more aggressively. And so I doubt this is the last that we hear from OpenAI on either IDEs or coding in general.

• 6:07Speaker 01

  And certainly, they're very committed on the Codex side. And people have had great, you know, experiences with Codex, their AI agent.

• 6:15Alex Kantrowitz?

  Now, people might look at this and be like, well, this is just continuing a pattern of OpenAI running into drama anywhere it goes. Is it getting concerning at this point? From the outside, it looks like it is.

• 6:26Speaker 01

  I think it's fine. They are somehow juggling, you know, building some of the world's largest data centers, you know, massive, massive energy, you know, needed for that, massive GPUs. They're, you know, acquiring Johnny Ive's, you know, company. They are, you know, releasing models at an incredible cadence. The rumor next is, is an open source model. So, so I think they have probably 50 different things going on. This being only one of those, those activities.

• 6:57Alex Kantrowitz?

  So, speaking of speculation and conjecture, what percentage of all AI spend right now, all generative AI spend, do you think is going to coding? Because the way that it's talked about, I mean, if you think about just Anthropic's growth over the past year, I would be stunned if it wasn't more than 50%.

• 7:15Speaker 01

  Oh, yeah, for all, all AI tokens in general? Yeah, I think that's, I think it'd be fun to look at a graph of this. I mean, it tends to be one of the highest volume, you know, kind of, if you look at a hu, like, what's the relationship between a human and the amount of AI that they can consume? Coding absolutely would be the, the peak use case right now. There's no other, there's no other human task that, that one person could cause so many, you know, kind of tokens to be produced. Like, deep research is, is great, but you do it maybe once or twice a day, and it's relatively confined. Summarizing information, very efficient, not very token-heavy. So, coding is definitely the one that is like, you know, this incredible, like, you

• 7:56Speaker 01

  know, one person could cost thousands of dollars per day of GPU expense if they, if they really wanted. So, I think this is going to be the killer app for, for the foreseeable future in terms of just sheer volume of tokens. And so, this is why it's such a big prize. And Google, again, you know, it's funny, actually, just, you know, timed well with, with your Sergey and Demas interview, I mean, With Sergey back at, like, there's these little small nuances that, you know, you probably don't want to overly extrapolate from, but, or anecdotes that you don't want to overly extrapolate from, but I think Sergey being back at Google is a very interesting thing to consider around how, you know, this is a company that has a great operator

• 8:37Speaker 01

  in Sundar, an incredible, you know, AI innovator in Demis, Jeff Dean, you know, deep on research and science, and then now this kind of hardcore founder in Sergey, like, this is not a company that is going to lose the coding battle. Like, they, like, there's no way that Sergey is sitting around being like, "Oh, I'm going to use Anthropic to code the next version of a feature I'm building." Like, he has to make sure that they're using Google's technology. Like, that is a, that is, you know, obviously a point of pride for any founder, is to make sure that you're building the technology that you're using for, you know, the domains you're going after. And so, I just, you have to imagine how committed

• 9:19Speaker 01

  they are to solving this problem, and Varun now and team are going to be now one more way to accelerate that.

• 9:27Alex Kantrowitz?

  Maybe that's the reason why Mark Zuckerberg is deciding to spend billions of dollars on AI talent, seemingly, is because they started using Sonnet, I think, for coding, so he realized there was a problem there.

• 9:37Speaker 01

  Well, it's not crazy, right? Yeah, I mean, think about, think about all of these founders, right? Greg Brockman at OpenAI, Sergey Zuck, I mean, they don't want to walk around their office and find out that the thing that everybody's really excited about is somebody else's model. So, like, that is like, that would be like if, you know, you worked at Facebook and everybody was on X all day long and not using, you know, Facebook. So, like, these things are very major points of pride for these founders, which makes it the race so exciting to be watching.

• 10:12Alex Kantrowitz?

  Yeah, back when I was reporting on social media, whenever there was a trend on Facebook, when they had their trending column that originated on Twitter, they would never say, "People are talking about this on Twitter." They'd say, "They're talking about it on social media," and then point back to Facebook posts of people talking about the Twitter thing. So I think that really goes to the hubris of these companies. And just to, you know, put a finer point on what you were saying, you know, for those who are listening and are maybe more on the financial side, not that technical, you're paying for every - it's generative AI, so you're paying for tokens, the characters that these machines generate. And so when you say, "Build me a web app," it's just tens of thousands, if not hundreds of thousands of tokens. And that's why we're seeing people spend this much money in coding.

• 10:55Speaker 01

  So I'll give you an example of how crazy this gets. I was talking to a founder this week where, I mean, like, I, you know, every day I see something that I'm just like, like, I have to completely reassess my estimation of the future. This founder is, right now, a solo founder. He has many different, I don't know if it's five or 10, whatever the right number is, many different agents in the background going off doing individual parts of his codebase, as well as the marketing kind of website that he has to build for this product that he's working on. And so he is effectively, as a solo founder, a manager

• 11:39Speaker 01

  of multiple agents doing all of this work. And then his job, and basically the new form of engineering work out there is to come up with incredibly precise prompts that are super tuned for his use case, and then kick off all these agents in the background that are going off and doing work. And then he goes and reviews their code, and he integrates that code into the broader codebase, and then, you know, effectively is there, you know, reviewing and auditing all of their work. The reason why that's so impactful is, or meaningful, is that one person could literally be causing tens of thousands of dollars a month in AI consumption because of just the single actions that he is doing.

• 12:22Speaker 01

  So while that's not going to be the behavior of everybody on the planet, that is this, that is a massive force multiplier of human-to-compute ratio that we've just never seen in computer history.

• 12:33Alex Kantrowitz?

  All right, so OpenAI isn't the only one making headlines this week. There's been some crazy stuff happening with Grok. Both a new model, Grok 4, and the behavior of Grok has been disappointing, shall we say. So let's start with the actual new model first, and then we'll talk about the alignment issues. So Elon Musk builds this massive GPU cluster in Memphis called Project Memphis. He calls it Colossus, I think that was the name of this GPU cluster. And we finally see, I think, the first model that's built on top of it, Grok 4. This is from Tom's Guide, Grok 4 is live. He says that it's going to be expected to rival OpenAI's GPT-5, which we still don't know when it's coming.

• 13:18Alex Kantrowitz?

  Claude's for Opus, you have artificial analysis. This is a benchmarking firm. They basically say that Grok is blowing away all these different benchmarks. And then, of course, in the ARC AGI test, it outperforms every model by a significant margin. Some have said maybe these benchmarks are, you know, maybe Grok is just benchmark hacked, or you can't believe them. But it seems like there's enough evidence here that there's a chance that making this GPU cluster massive has worked for Elon Musk. What's your read on it?

• 13:49Speaker 01

  Yeah, I mean, I think it is working empirically. Obviously, you can, you can, and we saw this with Meta a little bit, you can sort of train your models to perform better at some of the evals or the benchmarks, which then, you know, somewhat can delude you into thinking that the model is better than it is, where it's really just better at these kinds of tests. However, right now, I think most of the evidence is that this is a very high-performing model kind of across the board. It continues to align with the theory that more compute, more data, generally is going to produce better models. And then they're doing some novel things that I think are emerging across the industry, but maybe this will be the first. Kind of real commercial model at scale that does this, but they have a model called Grok

• 14:34Speaker 01

  four heavy that has multiple agents go off and execute basically the same task and then go, they go and review their answer for which answer they think is, these agents think are the best result. And so this is a great example of how you can have a lot of compute in the training process, but then also have lots of compute in the inference process where, where you just have the model working harder and harder and harder to produce better answers, which, which is clearly producing, you know, great results. They show the, the scores of what Grok four heavy can produce. And I think that will become a standard across the board. So I think it's, it's absolutely a continued improvement on, in, in, in model quality and

• 15:19Speaker 01

  model performance. And we're, we're super excited that, that this, the scaling laws, you know, are continuing to play out. And this is just more evidence of that.

• 15:29Alex Kantrowitz?

  Well, it's interesting. So I want to talk with you about the scaling laws because we've had a number of folks come on this show and say, yeah, we're seeing diminishing returns. And it's not nobody's. I mean, Thomas Kurian, CEO of Google Cloud said it pretty much straight up a couple of weeks ago. And now it seems like it's been tested where Elon said, I'm just going to win on scale. And he makes what is I think the biggest GPU cluster in the world. And it looks like it is producing. One of his engineers, a guy named Uday Rudjaru, is he's actually, he's left and he's going to OpenAI to work on Greg Brockman's scale team. And I messaged him after he left and I said, do you believe in the scaling laws after what you've seen? And he says, yeah, the more GPUs, the better. And it looks like

• 16:14Alex Kantrowitz?

  that's what they're showing. So what makes for this disconnect between everybody yelling diminishing returns and what we're seeing now, which is like, maybe that's not the case?

• 16:25Speaker 01

  Well, I, I would, I would kind of say that both can be true. Diminishing returns is first of all, it's a relative concept. So diminishing relative to, to what rate. But, but I think that the way to think about it is if you think about a curve that eventually Sort of asymptotes, all that matters is where are you in that in that curve? So if if a curve is sort of like this and it's asymptoting right here, well, if we're right here, that's bad. But if we're right here, it will be quote unquote diminishing returns, but you haven't asymptoted or plateaued yet. And so all that matters is is where you are on that on that curve and trajectory. And you can see based on some of the evals, it's not as if it's not as if there's a going to be

• 17:08Speaker 01

  a 10x improvement in intelligence anytime soon, simply because some of these evals were already at 80 or 90% of of the of where the evals at. And so there there isn't even room for the model to be to be 10x better. And so that might mean, though, that you have to apply five or 10x more compute to get to that next final, that last mile of of of intelligence, which again would be both diminishing returns, but also something we would still continue to drive as an industry, because you're still just going to get you're going to appreciate that quality difference. And so that that I think is totally fine. You know, in general, talking to enterprises, we're already for the most part for, you know, with with many, many exceptions, we're already for the most part in a position where the

• 17:52Speaker 01

  technology well exceeds anybody's ability to adopt all of these benefits so far. So simultaneously, we want a we want the progress to continue at this exact rate. And there's most use cases on the planet still could be benefited just by even what today's models can do. So we want more innovation, we want more compute, we want more intelligence. But even if you stopped right now, you'd still have, you know, that massive amounts of economic gain get delivered from from what we've already created.

• 18:23Alex Kantrowitz?

  Right. But I guess the question really is for those, and I don't think you've said this, but there's many in the AI industry saying, well, the scaling laws are a straight shot to AGI if we keep making things bigger. So I guess I'm trying to test what we're seeing with Grok, test that statement based off of what we're seeing with Grok.

• 18:42Speaker 01

  I I I'm not going to make any predictions on that front.

• 18:46Alex Kantrowitz?

  Okay.

• 18:47Speaker 01

  Because the the the.

• 18:49Alex Kantrowitz?

  But do you think this is evidence for or against?

• 18:51Speaker 01

  I think the smartest people on the planet have two totally different views. And so I am, I'm not going to get in the middle of that one. I mean, clearly you have people like Ilya where, you know, it's rumored that he's working on a different architecture or and and and maybe a different path. And then obviously you have other people that are, you know, let's just throw more compute and data at the problem. I think you can start to sense actually as an industry that the the AGI term has actually kind of gone into the backseat. And obviously more of the conversation is around superintelligence. And I think there's more and more comfort around this idea that actually the race really is just how do we build intelligence that far exceeds a human? And what will the economic and and, you know, kind of societal benefits be of just even

• 19:36Speaker 01

  accomplishing that, which are massive? And I have always sort of found the AGI thing to be, you know, particularly squishy as a concept. I I in the B two B world, I deal way more with just like utilitarian concepts. And so superintelligence and this idea of we have AI that will far exceed a human, like that that alone is enough of a breakthrough to be shooting for. And I think what you're seeing with scaling is we will be able to certainly accomplish our collective definition of superintelligence with the the current the current path around with scaling laws.

• 20:10Alex Kantrowitz?

  Okay, so you would say that there's two camps. One is keep scaling and the other is we need new techniques.

• 20:17Speaker 01

  Well, if you if you.

• 20:18Alex Kantrowitz?

  Is that right?

• 20:19Speaker 01

  Yeah, if you have, if you put Jan, Jan LeCun, Ilya and.

• 20:24Alex Kantrowitz?

  Demis would be in that category of we need new techniques.

• 20:28Speaker 01

  And Demis in in one category. And then you put a bunch of sort of, you know, scale max, you know, maximalists, you know.

• 20:37Alex Kantrowitz?

  Who would that be? Dario?

• 20:38Speaker 01

  Maybe Mario, maybe anybody just running one of these current clusters. I don't know where Sam is these days. So, so, you know, probably Sam.

• 20:45Alex Kantrowitz?

  Well, he said, we know what to do. We know what to do. And he's investing in Stargate. So it seems like he's a scale max.

• 20:51Speaker 01

  That's a scale maximalist. So, um, but, but what's interesting is actually, I think that, that you'd be able to get them all to say the same thing, which is, which is this category that says we need a new idea are probably AGI maximalists. And then there's another category, which is like, actually, it's already proving out the economic and societal advantages of, of even our current approach to AI. So just let's keep running that for as long as possible and we'll just keep eking out more and more benefit. Like you could already dramatically improve every healthcare experience on the planet. Just by using whatever the latest, you know, state of the art model is in every, in every area of healthcare.

• 21:31Speaker 01

  Like, everybody will absolutely get better doctor diagnosis. They'll get better healthcare. The, the doctors will be happier when they transcribe all of their, uh, their, their conversations with patients with AI like, like, and that's just like today's state of the art. We don't need any new breakthrough just to have that ripple through everything that we do. If every engineer on the planet had background agents that were, you know, checking for bugs or writing, writing new code for them or updating their libraries, all that long tail work that's really, really inefficient and not enjoyable. Already the economic advantage of just today's architecture would be massive. So, so I think that, that the, I think you can basically be happy about both outcomes.

• 22:11Speaker 01

  Like the super intelligence track with more scale is a great track to be on and we're just going to get more and more benefits. And the sort of like, we need a new idea, AGI maximalist, that's fine too. And that's just upside if, if and when we discover whatever that thing is.

• 22:27Alex Kantrowitz?

  Okay. So I want to poke at this a little bit because we did see something this week that is concerning and really goes to the stability of these models. Which is that, uh, Grok became, I don't know, a neo-Nazi. It seems like half the time these buckets become neo-Nazis, but none of the big ones, none of the big.

• 22:44Speaker 01

  I don't know if it was Neo. I think it was, uh, it was like OG.

• 22:46Alex Kantrowitz?

  Straight up Nazi.

• 22:47Speaker 01

  Yeah.

• 22:48Alex Kantrowitz?

  All right, I was giving it too much credit. So, so this from the BBC Musk says Grok chatbot was manipulated into praising Hitler. Grok was too compliant to use her prompts, too eager to please, to be manipulated essentially. This is being addressed in response to a question asking which 20th century historical figure would be best suited to deal with, I think it was the Texas floods. Grok said to deal with such anti vile, such vile anti-white hate Adolf Hitler, no question. All right, so that definitely a Nazi full blast. Also, someone who insulted President Recep Tayyip Erdogan of Turkey. And so he got the Grok got blocked in Turkey. So just really off the reservation here messing with Erdogan.

• 23:32Alex Kantrowitz?

  So I want to ask you, we got a, one of our listeners dropped a question and we're going to get some Discord questions. But basically asked me, what does it say about the stability of these models? That with a little tweak Grok turned into Mecha Hitler? That doesn't sound like a tight system or architecture. It sounds really wobbly.

• 23:52Speaker 01

  That's a question for me. I mean, unfortunately, I don't know if there's been a full postmortem as to whether that was a training issue. All of it's in the weights to be Mecha Hitler, or if that was a system prompt issue. In which case you can do quite a bit with a system prompt to effectively, you know, change the direction or path of what you want the AI to respond to. So to the extent that it was as simple as they used to have a system prompt that said, you know, please be politically correct and be thoughtful. And, and make sure that to, you know, not say anything offensive. If they used to have that and then they basically said, actually, no, you know, say anything you want.

• 24:32Speaker 01

  Then, you know, in that latter mode, users could certainly kind of, you know, cajole it into, into then doing Mecha Hitler stuff. And.

• 24:43Speaker 01

  And so, so I think, I think it's, it's sort of unknown how they train that model, how much of this was system prompt, I, you know, for, for being able to remove that as a risk factor, I think it's sort of well understood what you need to do post-training and, and what you need to be, you know, doing from a safety standpoint. And then it's really just a decision of the model provider and the application layer of, of how to implement those things. But I, I thought it was obviously a ridiculous, you know, ridiculously bad situation, deeply, obviously, you know, offensive and dangerous, but also not really that much of a meta story about AI, simply because you can get these models to do anything you want. And the whole, the whole thing is as an industry, you're kind of working toward,

• 25:26Speaker 01

  you know, trying to keep these things confined within a particular pattern of behavior and, and, and sort of, you know, level of, of, of communication style.

• 25:38Alex Kantrowitz?

  This is the next iteration, Grok Force from TechCrunch. Grok Force seems to consult Elon Musk to answer controversial questions. So they decided, I guess, to try the next version where if you ask a controversial, controversial question, let's say about the Israel Palestine conflict, abortion and immigration laws, Grok will reference Musk's stance on these subjects through news article written about the billionaire founder and the face of X. And TechCrunch tried to do this and was able to replicate multiple times in its testing. Is this the, the answer to the alignment problem? Just follow what Ivan believes?

• 26:12Speaker 01

  It's, it's an experiment of, of how to, how to achieve it. So, he, listen, he's always claimed he's, he's fairly centrist. So, so that would, that would make it pretty aligned. Yeah. I mean, I, I mean, they clearly keep stepping on the rake and, and the rake, you know, keeps hitting themselves in the face. But I, I, I have faith that they will, they will find a way to work through some of these kind of ridiculous situations.

• 26:40Alex Kantrowitz?

  Right, and we've talked about where the money is in AI today. I mean, I'd say we both said majority coding, and then probably comes enterprise use cases.

• 26:48Speaker 01

  Yeah.

• 26:49Alex Kantrowitz?

  And as this is all unfolding, we have the big technology Discord server. One of our users says, "Oh, you're speaking with Aaron Levie this week? Why don't you ask him this question?" This is the question: "Given what we just saw, that Elon is willing to do with Grok, would you really, in your heart of hearts, consider this model for use at Box, or even extending it a little bit more?" Why in their right mind would an enterprise consider integrating Grok, given this pattern of behavior?

• 27:17Speaker 01

  Well, I think it's a fantastic question, and it's absolutely, you know, worth thinking about. Do you remember, like, 10 years ago, Microsoft had an AI chatbot, I think, called Tay or something?

• 27:29Alex Kantrowitz?

  Tay.

• 27:30Speaker 01

  Yeah, Tay.

• 27:31Alex Kantrowitz?

  So I remember it well, because I had the exclusive, so Microsoft came to me to break that news at BuzzFeed.

• 27:36Speaker 01

  Oh, wow.

• 27:37Alex Kantrowitz?

  And I wrote, "Microsoft has this fun chatbot called Tay. It will, you know, be your friend." I pinned it to my Twitter profile, went to sleep in San Francisco, woke up that morning. Overnight, Europe and the East Coast had figured out that Tay had been a Nazi, and I woke up to many concerned messages telling me, "Please take the pin down."

• 27:54Speaker 01

  Okay, so I'm glad that...I didn't know you caused this problem, so that's actually what...

• 28:00Alex Kantrowitz?

  I didn't cause it, but I might have inadvertently supported it, so I took the pin down eventually.

• 28:07Speaker 01

  I think this space is a, you know, always this process of figuring out where these models, you know, kind of go a bit crazy, produce either the wrong information, or hallucinate, or have accuracy issues, and it's all about continuing to iterate on how to improve the system prompt, the model, the alignment of these models.

• 28:33Speaker 01

  And so I, you know, just judging by both how they responded, they took it down, you know, almost immediately as, as these examples were coming out, the fact that they acknowledged kind of why this was occurring and, and what they're, what they're working on about it. I, I think that they will continue to improve their, their model and, and the AI system. And then it's really up to individual customers to decide which, which model do you, do you trust? You know, what do you want to use? And I think everybody should take into all of the factors of, that they, that they would want to consider. So I, you know, we're, we're certainly not in the business of, of, you know, telling our customers which, which type of AI model to use. There's going to be some that have really perfected a use case. And so thus you're going to want to use a particular AI model.

• 29:18Speaker 01

  But, but, but, you know, I think everybody has to make their own decision of which, which AI to use.

• 29:24Alex Kantrowitz?

  Yeah, I guess that their point was if you're an enterprise, I think this is one of the examples given. If you're an enterprise and you're using like Grok to write emails, you don't want it to like in the middle of responding to a sales request to be like, and by the way, you know, who was great? Hitler, but you know.

• 29:39Speaker 01

  My, my guess, my guess, I haven't seen, I haven't, I haven't read all of the, I haven't read all of the, if they've done a postmortem or anything, my guess is that that's not built into the model as much as.

• 29:51Alex Kantrowitz?

  Right.

• 29:52Speaker 01

  It was a, it was more of a Grok kind of specific application issue that, that caused that. But let's, let's see what they, what they, you know, how they respond.

• 30:01Alex Kantrowitz?

  I just want to quickly agree with you here because Elon, we had talked about this actually on the Monday show with M G Siegler, that Elon had repeatedly, you know, talked about how he lost control of Grok and it was citing Media Matters to try to take down Cat Dirt, which we know is a capital punishment worthy crime in the Elon universe. And he kept saying, you know, Grok's getting a rewrite. So this is clearly a post-training snafu where they took it from something that was politically correct. They wanted to make it less politically correct. And this is sort of where you get on the internet when you want to go there.

• 30:31Speaker 01

  I think to, to respond to that initial question from that person, I do think that anybody who wants to have an enterprise business does have to ensure that they are building basically purely utilitarian AI systems that are generally considered to be very safe and, and trustworthy. So, so if you want to be in the, in the B to B game, which will be most of the volume of, of AI usage and APIs over time, because that, that's how you will show up in every other product, then, then this matters a ton. I just haven't seen evidence that, that they don't. Want to go fix those problems and but we'll see.

• 31:06Alex Kantrowitz?

  Okay, yeah, I hear you. I think you're probably right here. All right. I want to go to break. Before we go to break, if you are on techmeme.com this weekend, you probably see that this podcast is showing up as the top podcast in a list of shows. It's reverse chronological, so we posted it, I think, most recently before the weekend, but it's a great placement and I want to thank Techmeme for it. If you're not familiar with Techmeme, it's read by tech industry leaders, executives, VC founders, key product people. It has info-dense headlines summarizing the news and enabling leaders to absorb what happened in tech as quickly as possible. I use it all the time for the show and it provides unique and valuable context, including related news, tweets, blue skies, threads, when people are still threading.

• 31:50Alex Kantrowitz?

  Highly recommend Techmeme. Thank you, Techmeme. It's really great to be partnering with them. All right, we're going to go to a quick break and then we're going to talk about Nvidia hitting $4 trillion. This episode is brought to you by Avpoint. Everyone's racing to roll out AI right now. Copilots, chatbots, agents doing real work. But here's the part nobody loves talking about. All that AI runs on your data and most teams have no single way to see it, secure it, and prove it's under control. That's exactly what Avpoint does. For 25 years, they've been the trusted layer beneath the world's most demanding data, now extended across your entire AI estate, your data, your cloud,

• 32:31Alex Kantrowitz?

  and the agents acting on your behalf. It's how more than 28,000 organizations deploy AI with confidence, so innovation scales without scaling risk. Avpoint, the unifying trust layer for AI. Head to avpoint.com to see how enterprises deploy AI with confidence. Learn more at avpt.co/bigtechnologypodcast. That's avpt.co/bigtechnologypodcast.

• 33:00Alex Kantrowitz?

  Violent threats against executives are growing at an alarming rate. When a company experiences backlash, executives are the first people blamed. When their address and other personal details are sitting online, this backlash can lead right to their front door. The team at Ironwell knows this better than anyone. They've protected some of the most targeted executives and individuals on the planet for almost two decades. As a wide-glove enterprise security service, they protect your people with continuous personal data removal, proactive prevention tools, and human-led emergency support. So when someone goes looking for your executives, Ironwell ensures they hit a dead end. So here's what to do. Go to ironwell.com/bigtechnology, fill in the quick form, and request your free risk assessment. Their team will show you just how exposed your executives are

• 33:45Alex Kantrowitz?

  and how to lock it down before a threat reaches their front door. That's ironwell.com/bigtechnology. Learn how to stop online threats before they become real-world attacks. The link's in the show notes.

• 34:00Alex Kantrowitz?

  And we're back here on Big Technology Podcast with Box CEO Aaron Levy. Aaron, the money for an industry that is majority-enabling coding use cases keeps pouring in, and we now have our first $4 trillion company. I think M.G. Siegler pointed out that the first trillion-dollar company was Apple. The second, the first $2 trillion company was Apple. The first $3 trillion company was Apple. The first $4 trillion company is Nvidia. So it just goes to show you all those decades of working to sell computers and iPhones. Now, the GPUs are the hotness. It's been, yeah, this is from the Times. Nvidia spent three decades building a business worth $1 trillion,

• 34:41Alex Kantrowitz?

  two years turning itself into a $4 trillion company. Is it just another number, or is there something significant about this?

• 34:48Speaker 01

  I think it's, I think it's fun. It's a, it's a fun milestone. There's obviously nothing magical about four versus 3.999. So, so to some extent, it's mostly symbolic. But, but I think what, what it being the largest company in the world is as an embedded message in it, which is just the point of leverage that Nvidia has relative to essentially what everybody is betting on is the future of the economy, which is an AI-powered economy. With robots and self-driving cars and AI systems that we chat with and agents that do work for us, you know, you would expect that a meaningful portion of the profits of that economy will accrue to the infrastructure providers of that economy. And as you go through

• 35:31Speaker 01

  the stack, you've got the hyperscalers, you have the model providers, you have then the chip providers, and Nvidia is in the pole position, you know, on the chip front. So I think it's well deserved. Jensen's a beast. You know, he's, he's, you know, just worked, obviously insanely hard for decades to get to this point, right? And it's a right place, right time with many, many decades of building up to be able to be in that position. And so I think it's, I think it's an important milestone for sure.

• 36:01Alex Kantrowitz?

  Speaking of these big numbers, though, I mean, eventually they have to be tied to reality. And it's not just Nvidia, right, which is going to have to justify four trillion now. I mean, I guess the scale hypothesis, sorry, the scaling laws news is good for Nvidia. Maybe that's part of the reason why it's up today. But you also have CoreWeave, which has had a 4X jump in its share price after like a pretty underwhelming IPO. And then you have Meta spending all this money on talent. Newcomer says this AI data center media conjures up the B word, which it means bubble. Is it something to fear? What do you think about this term bubble?

• 36:41Speaker 01

  I, I, I think it's all well-placed at the moment. So we're, you know, if you, if you think about, let's just fast forward 20 years. One cool thing, actually, maybe just small anecdote, one cool thing is Waymo just arrived in where I live in Silicon Valley in the peninsula. And, and, you know, everybody in downtown SF has already kind of had their religious moment on this, but we've never had it in the suburbs. And and you just you get into one of these things and it takes you to a meeting or it takes you to dinner and it's a completely life altering experience of just imagining

• 37:25Speaker 01

  in 20 years from now, what if every car you get into is is autonomous? What if every factory you go to has, you know, is like 80% robots just running around? What if every computer you use is augmenting your work by a factor of 10 to just to work on your behalf to do way more? What if every time you, I mean, maybe people won't like this, but every time you have a sniffle, there's an AI, you know, doctor that's like doing diagnostics on you, just like the future is going to be going to be so many of these autonomous systems around

• 38:06Speaker 01

  us, helping us with education, healthcare, transportation, commerce, just basic productivity? And so if you think out 20 years and that's the world that we live in, who would you want to invest in other than other than that architecture and that infrastructure stack right now? And Nvidia would be at the at the center of that, but then many of these other players and platforms would would would would obviously be in that investment case. So I know I don't I don't think it's I don't think it's crazy. And I think it's 100% sort of directionally aligned with where the economy is going.

• 38:42Alex Kantrowitz?

  I thought three trillion was surely the end of it, but we reached four so quickly and I'm just like, oh no, like, is it gonna hit five soon? At this point, there's nothing that's out of the imagination for me.

• 38:55Speaker 01

  I mean, we probably need to start talking about the 10s. So like, what will that?

• 38:59Alex Kantrowitz?

  Really?

• 38:59Speaker 01

  Yeah, why not? Yeah, sure.

• 39:01Alex Kantrowitz?

  That's true. Okay. How long and all right, over under Nvidia 10 trillion by 2028?

• 39:08Speaker 01

  Oh, I maybe I would I'd give it a little bit more time, but but, you know, to be to be worth $10 trillion, you probably want, you know, 300 billion in profit, let's say. And so with their margins, that means that, you know, they're doing 400 billion in revenue. Like, that's totally not crazy. 400, 500 billion in revenue for Nvidia, that's that is a totally realistic scenario to to imagine.

• 39:35Alex Kantrowitz?

  Okay, this isn't investment advice, but I'm starting to scratch my head here. All right, so of course, there's a company that they've replaced as like the one that's been setting these bars with, which is Apple.

• 39:44Speaker 01

  Yeah.

• 39:45Alex Kantrowitz?

  This is a report in Bloomberg, "Apple Should Consider Replacing Tim Cook As CEO," Lightshed says. So the story says, "Apple should consider replacing Tim Cook as the iPhone maker's struggles with artificial intelligence raise significant risks for the company. Apple needs a product-focused CEO, not one centered on logistics," the two analysts said. "Missing AI could fundamentally alter the company's long-term trajectory and ability to grow at all. AI will reshape industries across the global economy, and Apple risks becoming one of its casualties." You know, it's great setting this up, right? The, you know, could Nvidia hit $10 trillion? Because if AI is going to be as transformative as you suggested with all these various use cases, it is true, Apple has been flat-footed.

• 40:30Alex Kantrowitz?

  Is this the craziest suggestion that the Lightshed guys are making?

• 40:35Speaker 01

  Well, I think the thing I would say, so maybe a couple of things. First of all, I think, I think Tim's great. And so I have a, I have a bias toward him for a number of reasons. But the, but, but I, I, you know, the thing that is worth noting is how strong Apple's position is in. And that, what that, what that then equates to is their ability to watch the space and figure out the right move to make and when to make it. Because whether, whether some people like it or not, you know, this is still the best device, handheld device on the planet. And it has the best set of apps on the planet, and it has your whole life kind of tied to it.

• 41:17Speaker 01

  So given they own that platform, their ability to lodge an AI into that at any point in the future remains very strong. And so I, I look at, I look at this as, you know, if you have, you have basically three, three options as a company. You could be a first mover and then totally sort of have a debacle and it not work. And we've actually seen plenty of examples in AI where the first mover is no longer the relevant player. You could have a scenario where you are a first mover that has a compounding advantage that continues to persist. Let's say OpenAI is in that category. Incredible execution and absolutely amazing.

• 41:58Speaker 01

  And then you have another category, which is you enter this space at a time when the architecture has sort of been figured out, when we understand the economics of the model, when you're not having to, you know, you're able to have step function levels of improvement by the time that you launch into it. And I think maybe Apple didn't purposely make that choice, but they are clearly in the position where they can actually have that choice now. And so I think you can just look at this as if this was 2004, we could have easily said, why has Apple not released a phone? And yet by 2006, like that wouldn't have mattered and they had the dominant platform that would, you know, continue to exist.

• 42:42Speaker 01

  I mean, Microsoft had a tablet computer in 2002 or something. I own one, or my co-founder owned one and I owned one of their Windows smartphones made by Compaq or HP. And so think about that they had the smartphone and they had the tablet computer first and neither of those things mattered to the long-term dominance in the space. And so Apple has a position and a potential of basically when the time is right to jump in, they still have the devices that we're using, they still have the OS that we're using, and they'll be able to have learned from all of the mistakes of, you know, various companies along the way. So I wouldn't count them out and I think they are clearly sitting around saying, when

• 43:27Speaker 01

  is the right time to pull a trigger on a much bigger move? And so I think we have to just wait for that.

• 43:32Alex Kantrowitz?

  What do you mean much bigger move?

• 43:34Speaker 01

  Well, they either have to make the decision of either train a model that is, that gives them a state-of-the-art AI model, or do some...

• 43:45Speaker 01

  Substantial partnership or acquisition move, all of what we've seen with these, you know, kind of founder or CEO hires. Obviously, the acquisition environment is complicated because of DOJ and FTC. But I would, I would certainly be astonished if in two years from now, there wasn't one of those, those choices being made. But I'm not, I'm not sort of that worried that it hasn't been made yet.

• 44:08Alex Kantrowitz?

  All right. Here is my galaxy galaxy brain idea. It's one step further from the typical galaxy brain. So I've been on the show advocating for perplexity. Maybe I've been thinking too small. Let me put it this way. Apple just lost its COO this week, Jeff Williams. And everybody thought Jeff Williams was going to be the successor to Tim Cook. Are we now in a moment of setup where Sam Altman and Johnny Ive have teamed up on a device? Tim Cook is getting ready to retire in the next couple of years without a clear successor?

• 44:47Alex Kantrowitz?

  Now that Williams is gone. Do we see the ultimate tech merger where OpenAI becomes for profit and Tim Cook says, Sam, Johnny, pick up the legacy. They did the picture. I think they want this. Can it happen?

• 45:05Speaker 01

  That is a, that is a wild, that is some wild fanfiction. Anything could happen. I think that is, that should be totally in the, in the category of, of options. You know, if you're being realistic, by the time that that moment would, would likely occur, you know, OpenAI should be much bigger. That would be much more complicated than as a deal. But I like the, you know, certainly the, as a brainstorm, it's a, it's a great way to brainstorm.

• 45:31Alex Kantrowitz?

  Okay. That's a very nice way to let me down. And yeah, I said merger for a reason. I wouldn't call it an acquisition. It might have to come at this point where the two just come together that way.

• 45:41Speaker 01

  No, no, fair point. And I've seen crazier things in my life now in tech. So I can't write anything out at this point. So, so let's, let's see what happens.

• 45:50Alex Kantrowitz?

  Let me put this, put it this way as we end. I think that type of deal is far more likely than Apple buying Anthropic.

• 45:58Speaker 01

  Just because it's going to require something so much more substantial? Or why would that be more likely?

• 46:05Alex Kantrowitz?

  Because I think it's a better cultural fit. I think the Anthropic team and Apple would clash, but I think OpenAI going into Apple, you know, could potentially work. Although OpenAI is much leakier than Apple. Although Apple leaks everything to Gherman these days.

• 46:19Speaker 01

  You know, the only thing I would just suggest or posit is, is, you know, this, it'll be fascinating to watch what Meta does with, obviously, its new superintelligence org, because, because, we actually already saw it with Grok, to be clear, but Meta will be a second round of this. If, from a more or less standing start, you know, they're able to accomplish, let's say, some new breakthrough state-of-the-art model in six months, 12 months, 18 months, or whatnot. I think what that will prove is basically, it still remains largely a talent and compute and data game, which means that you don't really need to buy an existing incumbent. You mostly just need to decide to go big on, on, on the compute and on the training, and, and obviously have the right talent to do that.

• 47:05Speaker 01

  And the day that, that, like, it doesn't really matter whether you had, you know, all of the other prior versions to, like, before that moment, like, like, you're, you're doing a reset no matter what. So I would, I would just argue that, like, we, we get all excited about this idea of some big mega acquisition, but it's not really, it's not a problem that requires that kind of scale, except for when you're just doing the capital expenditure of the GPUs. You really just need the right talent, the right training data, and the right compute. So I would, I would more bet not on one of these very large, multi-tens of billions of dollar deals, simply because there's other paths to get there that are not as complicated.

• 47:48Alex Kantrowitz?

  That's a great point. I mean, it's less about, you know, an individual company's IP because everyone's effectively sharing the IP. It's about productizing it right now.

• 47:56Speaker 01

  Well, that's exactly right. So if you imagine this industry, within one year, every single breakthrough idea eventually gets discovered by everybody else. Like, there's never, nobody has kept an advantage for more than a year on some secret idea that only they have. And so, Apple's ultimate, Apple's ultimately advantage is they have a distribution model that nobody else has. And they have a form factor of where AI could show up that nobody has. So they don't need, they don't necessarily need to have the best model relative to, you know, one or two months being ahead of anybody else. They just need to have, like, a good enough model that any one of our non-tech friends would just be like, this is fantastic.

• 48:36Speaker 01

  I love this thing. Which is just not, again, does not require that scale of acquisition or whatnot.

• 48:43Alex Kantrowitz?

  All right, everybody. The website is box.com. You could also find Aaron's very insightful posts about AI on LinkedIn, Aaron Levie, and on X. His handle is Levy. Aaron, this was so fun. It's always great to speak with you. I appreciate the time. Thank you.

• 48:59Speaker 01

  Good to see you, man. Take care.

• 49:00Alex Kantrowitz?

  You too. Thank you, everybody, for listening, and we'll see you next time on Big Technology Podcast.

• 49:06Speaker 02

  Now, more than ever, we need someone on our side. But Rob Whitman? Whitman's votes drove up costs, cut Medicaid, and risked 350,000 Virginians' health care to pay for tax breaks for himself and his donors. He's not on our side. Shannon Taylor will fight corruption, cut taxes, and bring down costs. This November, vote for Shannon Taylor. She's for us. Paid for by HMP, the HouseMajorityPac.com. Not authorized by any candidate or candidate's committee.

• 49:35Speaker 03

  In this life, you can't win.

• 49:37Speaker 04

  When the world gets you down.

• 49:38Speaker 03

  No!

• 49:39Speaker 04

  Escape to Pluto TV.

• 49:41Speaker 03

  This is objectively fantastic.

• 49:42Speaker 04

  Relax into the movies and shows you already know and love, all for free.

• 49:46Speaker 03

  Shwing!

• 49:47Speaker 04

  Even our news is uplifting.

• 49:48Alex Kantrowitz?

  You stay classy, San Diego.

• 49:50Speaker 04

  Watch School of Rock, Crazy Ex Girlfriend, Anchorman: The Legend of Ron Burgundy, The X-Files, Wayne's World, and so much more.

• 49:56Speaker 03

  We're not worthy!

• 49:58Speaker 04

  The best things on TV really are free.

• 50:00Speaker 03

  You just have to know where to look.

• 50:01Speaker 04

  Escape to Pluto TV. Stream now, pay never.