Elon Musk Addresses Cursor Staff, Nebius’ Nvidia Groq Chip Adoption, Tim Cook’s Farewell Party

0:13Gil Luria:Welcome everyone to the information's TITV. My name is Akash Pasricha. It is Tuesday, August 25th. We have got a jam-packed show for you today. First up, we have an exclusive story that Meta is planning to launch an AI agent. We then have the inside details of Elon Musk's first appearance at Cursor's All Hands after its SpaceX deal closed. We're also talking with a top Nebius executive about its decision to move forward with NVIDIA's Grok chip. We'll also chat with an analyst about NVIDIA's results coming out tomorrow. and to close out the show, it is Tim Cook's last week as Apple CEO. We'll take you inside the farewell party that even had a performance from OneRepublic.

1:00Gil Luria:Our Apple reporter has those details. It's going to be a great show, so let's get right on into it. Meta has been moving fast and furious with its AI product releases. A few months ago, it released its Muse Spark model. A few weeks ago, it released Muse Code. And this week, my colleague Jyothi Mann had exclusive reporting that the company is working on a new AI agent. I want to bring on Jyothi Mann, our meta reporter who wrote that story with more details. Jyothi, welcome to the show. It's great to have you here. Thanks for having me. From our San Francisco bureau, Jyothi Mann is in San Francisco, folks.

1:34Gil Luria:So this is your notice to any potential sources that you live in the big city now. That's right. I'm here, ready to meet people. Great. Well, I was excited to get you on this morning to talk about this story. So you've reported previously on this AI agent that Meta is working on, but you had some new details in the story out this week. What did we learn? Right. Yeah. I mean, so Mark Zuckerberg has dropped hints about this earlier this month, that we're going to see them release personal capable agents that understands you, your goals, and everything you care about. We have previously reported on some of the things it can do, but we've just recently learned that they're planning to, or at least targeting, a release of later this month or potentially early September.

2:24Gil Luria:And tell me about some of the integrations here with this AI agent platform that Meta is working on here, because it looks like it's not just going to be on its own. It's very much pulling from other online apps and stuff like that, too. That's right. So we know it's been trained to access websites in a secure environment like DoorDash or Etsy or Reddit or even Outlook. And the idea is for people to be able to instruct their agent to perform actions on their behalf, like schedule emails or order food on their behalf. So we're seeing that based on early prototypes that will sort of have all these capabilities and skills where agents can perform tasks on users' behalf.

3:08Gil Luria:How closely are they modeling it after OpenClaw, after the uptake of that product? Very closely, it seems like from the early prototype that I've seen, it seems to be quite similar in the sense that there's a dashboard and you'll be able to see other skills listed. So for example, if you've built a fitness tracker, you'll be able to access that again and you might be able to use that skill to create another fitness tracker or create a travel itinerary for a trip you've planned. So yeah, very similar interface, I'd have to say. And they're going to charge for this tool? Yeah. So we kind of reported earlier this year that they are considering a tiered pricing system.

3:53Now, one of the kind of premium tiers was priced at around$199.99, and that would include higher usage limits.

4:02Gil Luria:I like that you're very specific. It's not$200, it's$199.99. That one cent makes all the difference. But then you reported today, so there is sort of a, I think it was 20 bucks a month. There's another, I guess that would just be the monthly cost, I guess. The lowest year, yeah, that's right. So it's all kind of being considered a few months back. So it's unclear what they've kind of set themselves, what they've decided upon yet. But I mean, this is kind of interesting, Jyoti, because I mean, at least from the consumer angle, I mean, Meta has always been a free company as far as its social media products go.

4:42Gil Luria:You know, this strikes me as one of the first times where they would be charging for a consumer product if we sort of leave aside the wearables business. I mean, is this sort of a new paradigm for Meta in terms of charging consumers for its tools or am I missing some context here? No, you're absolutely right. I mean, earlier this year, we saw them release paid subscriptions to the Meta AI on Instagram and WhatsApp. So it is a big change. And we've sort of seen Meta start to see this as a product that people might actually pay for rather than just a feature that makes Facebook or Instagram better.

5:20And that's important because, you know, Meta has historically made money from monetizing attention and advertising. But with AI, it's a different model. You have a product that people, if they're using that directly, repeatedly, that's potentially enough value for people to pay for. So it's sort of hedging its bets against whether AI can be a really meaningful subscription business model alongside the ad business. And it also allows them to recoup some of their investment in AI.

5:47Gil Luria:Right. So the other half of the story that was really interesting to me is it sounds like Meta is coming out with a new model. what do we know about the the latest details of its of its latest uh post spark or post muse model i should say right so we are seeing that they're planning to release a model which is internally called watermelon in october so that's the date they're targeting of course that that could change um but we don't know yet if it's going to be part of the muse spark family or if it's a standalone model but i suspect it will belong to muse spark so we probably see muse block 1.3.

6:28Gil Luria:But maybe going back to the agent, as far as we know, this agent will be powered by Meadow's own models, right? I mean, because previously, per your reporting, I mean, there was some discussion about they were leaning heavily on Anthropics models while its own models are still under development. But it sounds like they feel pretty comfortable now with their own suite of models to power these agents. That's right. We understand that when it is released, it will be powered by Meta's new Spark models. Right. So, Jyoti, I mean, last question for you, you know, as we consider these new agents, these new apps, it now has the models now to power all this.

7:10Gil Luria:I mean, we've talked on this show previously about Meta's new strategy to unveil loads of new apps and not just bake everything into one app. Why do you think the company is pursuing that strategy? I mean, it's got loads of apps. It's going to charge for these consumer apps. It's a markedly different strategy than what we're used to. What do you think is behind all this? Well, I mean, obviously different products serve different habits and audiences. And it kind of speaks to a point that's worth mentioning where CEO Mark Zuckerberg said earlier in the year in an earnings call that AI is really allowing them to move faster and build things with fewer people.

7:50And so they're able to kind of take those risks and experiment a lot with products without kind of putting the entire ad business at risk. So they have that advantage of moving quickly and being able to create new apps just through bi-coding alone. So the strategy is sort of don't put everything on one interface. We can build multiple products quickly, see what catches on, and then put more resources into ones that are sort of winning. Great.

8:17Gil Luria:Well, Jyoti, I want to thank you for coming on. It's great to have you in San Francisco now permanently. I look forward to catching up more. That is Jyoti Mann, our meta reporter here at The Information. Cursor is officially an Elon Musk company now, as SpaceX's acquisition of the coding business has completed. Our Elon Musk reporter, Grace Kay, has the scoop on Musk's first address to cursor staff at its all-hands meeting. I want to bring her on for the details. Grace, welcome back to the show. It's great to have you here. Thanks. Okay, so the cursor deal is done. Your story starts with this all-hands meeting that the company had.

8:57Gil Luria:Elon Musk seems to have made an appearance. What do we know about the meeting? Yeah, so the meeting took place the same day the deal closed, and Elon Musk gathered. he called in via video across the offices and he basically welcomed them and kind of set out the mission you know of grok is falling behind we need you guys to come in and help us he outlined the stakes basically that anthropic is leaving right now he had a lot of complimentary things to say about anthropic and he said you know that ai might someday you know be more powerful than humans and uncontrolled by them. So, you know, he wants to be the first one to build it and control it.

9:36Gil Luria:I like how you say that, like, it's just an everyday all hands call. I mean, there's a lot in there, Grace. Okay. There's, there's the candidness that we are behind, you know, we need to catch up. There's the praise for Anthropic, which, you know, I don't know that it's, I don't know that every CEO would, would, you know, praise rivals so seriously. And then there's, of course, the idea that AI will replace humans, which we can put aside for the moment. But I mean, if we take a step back here, I mean, what were the expectations that he seemed to lay out for this Cursor transition? Was he seeming to put a lot of pressure on the company?

10:18Gil Luria:What was his tone? Yeah, I think there's definitely a lot of pressure on the company coming in. I think the expectation is that Cursor is going to help them, you know, help fix a lot of problems that XAI has been having, especially when it comes to their coding model. But Cursor staff is being spread out throughout the company. So they're going to help in a lot of different avenues, whether it's, you know, Grok Imagine, their image generation model, or, you know, their voice model, they're going to come in across the company. The other thing that's interesting is Cursor has all these enterprise relationships.

10:49And that's something that XAI doesn't have as much. And SpaceX leadership has actually been encouraging Cursor, you know, to keep hiring more people to their enterprise teams. So I think a huge part of this relationship is going to be Cursor coming in with the sales staff and these established relationships that can really help XAI.

11:07Gil Luria:Do we have a sense for how much Elon is encouraging Cursor to the enterprise relationships here? I mean, conceivably, they'll be selling space XAI products too, but the Grok model just isn't that good right now, right? So I wonder what the messaging is around like what they should actually be selling. Yeah, I think that's something they're kind of figuring out as they go. What I had understood is a lot of people coming in thought that, you know, SpaceX was acquiring Cursor because of their brand. And, you know, the brand was going to be something that they were going to push. And, you know, more recently, leading up to the acquisition, people were told that, no, the Grok brand is going to be a priority and potentially maybe we'll even, you know, come up with this new brand and that kind of cursor the brand is going to take a back seat and might eventually be phased out well that's kind of interesting yeah well and the brand the i mean the branding story here is interesting because i mean part of the grok brand is that it's very so closely affiliated with x and we know that businesses are sometimes shy from partnering with X or XAI or GROC because of how, you know, uncontrolled the platform can be at times.

12:27Gil Luria:So, I mean, if enterprises like Cursor, then, I mean, maybe it makes sense to leave that brand intact. Yeah, yeah, that's the interesting thing because GROC is definitely the edgy brand. And I think Cursor has a really good, you know, established relationship with software engineers and is very professional. Right. Yeah. Let's go back to the hiring piece. So Cursor is a thousand people, which honestly, I did not know until you pointed that out in the story. You mentioned that Elon has encouraged them to keep hiring. Sounds like there have been some departures, though, since the acquisition as well.

13:01Yeah. So there's been more than 45 departures. The general counsel left, some leadership on the GTM team, many of which had just joined a few months prior. right after I wrote the story, actually, within an hour of writing the story, Sasha Rush, who is, you know, was kind of a celebrity hire for Cursor and has been leading the composer team left also. So, you know, there's definitely a lot of people leaving at the same time that they're bringing in all these GTM people and hiring. So I think it's like part of it is, you know, acquisitions. You understand that some people might not align with the mission or they might not agree with it.

13:42So there's that. And then also, I think there's small cuts to before the two companies joined.

13:49Gil Luria:Right. What about the overlap in products here? I mean, do you have a sense for what's going to happen to the overlapping tools that they have? Yeah. I think it's interesting because they have served like two very different markets. So there's not a ton of overlap besides the composer model and the Grok model. But I think the focus is definitely going to be on Grok products going forward. Okay. And Grace, as you are watching this acquisition unfold and the integration happen, I mean, there's a lot of opportunity here, clearly, with the enterprise relationships, the brand that SpaceX itself has, the Elon Musk leadership.

14:33Gil Luria:What are the risks here that you are watching for and that we should be watching for as the story unfolds. Yeah, I mean, I think number one is to see what happens to the cursor brand and, you know, what this potential new brand will look like and how, you know, that might impact enterprise relationships. I'm also really curious about, you know, if people are leaving now, what is that going to look like in the future? Because XAI, you know, famously in February had a massive talent exodus. And in the AI talent wars, you know, that's a key part of the AI race. So what is going to happen to Cursor Talent?

15:09Is this going to impact their ability to bring new people on? How is that shakeout? Yeah.

15:15Gil Luria:We should point out that per your reporting, so Michael Truel, the CEO of Cursor, he's going to be reporting directly to Elon. So it sounds like he's going to be a close advisor to him. Yeah. Yeah, he's definitely going to be up there. And then the other co-founders have kind of spread out too across the organization, which is interesting. Right. But they're sticking around for now, as far as we know, which there's a lot of talent there. So, Grace, a lot to watch for and a lot of great reporting. I want to thank you for coming on. That is Grace Kay, our Elon Musk reporter, here at The Information.

15:50Gil Luria:Nebius had a newsy 24 hours. The company announced it has closed a convertible debt offering totaling$5.75 billion. And also, it is adopting NVIDIA's Grok chip, which is now in full production. I want to bring on Nebius' Chief Revenue Officer, Mark Boroditsky, for a conversation about those announcements. Mark, welcome back to the show. It's great to have you here. Kosh, always a pleasure. Happy to be here. So let's talk about the new Grok chip that Nebius has latched onto from NVIDIA. I mean, let's get to the basics here first. What are the problems that this NVIDIA Grok chip is solving? And then we'll talk about your decision to adopt it.

16:30I mean, in the greater scheme of things, we all know that as AI workloads become revenue producing, inference is everything. And we all know that the workloads are generating more and more tokens. And that ultimately requires us to be able to process a greater and greater sequence of tokens. The LPUs that are provided by the Grok LPX allows us to create this tandem processing capability to produce more tokens faster. So ultimately increasing the efficiency that we're able to operate and the performance that customers and the end users are expecting. So we're very excited about this new offering in the market.

17:16We know that there's going to be a lot of customer demand, given the discussions that we're having with our token factory inferencing platform.

17:24Gil Luria:Right. And how much is Nebius spending on adopting these new Grok chips? We're not actually sharing those details at this stage of the game, but we are taking delivery of a substantial order, and we'll be rolling them out in the coming months. Right. I want to ask you about some language that NVIDIA used in the announcement. So they said it's extending the Vera Rubin platform. So, I mean, what, this is the idea that all of the compute you get from Vera Rubin, it just puts it on steroids? Is that the idea? Basically. I mean, so it's built as a partner to the Vera Rubin chip. So the GPU in the Vera Rubin is partnering with the LPX chip to be able to create that on steroids.

18:07I'll use your term. Performance.

18:09Gil Luria:So, yeah, 100%. Did you consider any other inference chips? I mean, Cerebrus has a pretty good inference-focused chip, right? You know, our teams look at everything. You know, that's part of being a competent and highly motivated partner to our customers and the market. And we're also deeply involved with NVIDIA and all the innovation that they're doing there. And we also are seeing what customers are looking for. Customers are asking for NVIDIA at every turn. So net-net, we see this as the best combination of the market and the best opportunity with regard to overall customer expectations. Right.

18:44Gil Luria:So Nebius is the first AI native cloud, I guess, to adopt the Grok chip here from NVIDIA. I mean, we've talked on this show about how NVIDIA, the scale with which they are rolling things out, I mean, it's the first in a lot of ways. And so there's sort of, it's hard to know what to expect sometimes. And so given that you are the first major cloud customer here focused on AI to adopt the Grok chip, I mean, what are the risks with the integration and with the rollout that you are aware of, cognizant of? And how do you think about mitigating them during the rollout? Well, we're already working with the chips.

19:26So we've taken delivery of test racks of both the Vera Rubens and the LPXs. So we're already working through the combination, and we understand the dynamic of new chips in the market. We have a lot of experience when we've rolled out previous generations, GB300s, GB200s, B300s, B200s, H200s, et cetera. So our team is very versed in implementing new chips, and we know the challenges, and we're prepared for them. Our expectation is we'll be working with the most innovative customers that are also prepared for the beginnings of a new chip run. Those will also get the advantage of figuring out how to improve the performance they're offering.

20:07And there's an interesting dynamic in the market today. People are afraid to fall behind. And once they see the capabilities confirmed, we oftentimes are seeing what we have seen across the entire range of our chips. Multiple customers demanding the same. Right.

20:26Gil Luria:Mark, I want to ask you about the convertible offering that you guys just closed, so$5.75 billion. What's that going towards? A bunch of the Grok chips, a bunch of Rubens? What are you going to spend it on? We're building more data centers and we're filling them with more GPU capacity and they'll be coupled with LPXs and CPUs and all of the requirements that folks have. We're capitalizing the business. I mean, the bottom line, this is continued confirmation of the market's confidence in what we're building and how we're building it. That's why we were able to raise the biggest convertible debt offering ever.

21:05I didn't know that.

21:07Gil Luria:It's the biggest ever? That is correct. That is correct. Wow, that's got to be a pretty big data center you're building. A pretty big suite of data centers. We're building many. We're building many. Our model is a portfolio approach. I mean, we're at any given time probably delivering a half a dozen data centers, and that requires a substantial amount of capital. No question about it. Let me ask you one last question before you go, Mark. So you guys reported your quarterly results a couple weeks ago. I think it was 452 % top line growth, I think, which means that it's a pretty exciting time for your team.

21:44Just tell me a little bit about how you are positioning the company to look ahead beyond this era where it's all demand.

21:53Gil Luria:There's no supply. Everyone just wants all the chips, all the compute they can get. I mean, how are you positioning the team to think three, four years out when it might not be 450 percent? Maybe it's, you know, half that, a quarter of that, whatever it is. How do you take that long term view? Well, first of all, it's a really tough dynamic and you're completely correct. If I could get my team to think three, four years out, that would be magic. Right now, we're just trying to get them to think 12 to 18 months out. So to shift the conversation from the transaction that's in front of us and to partner with the customer around their broader demand requirements.