Speaker 0:  spacex is selling grok 4.5 as a claude killer, while spacex rents an entire supercomputer to the company behind claude. elon musk may be competing in the ai model race while quietly becoming the landlord for everyone else's race cars. so grok competes while spacex rents. so grok 4.5 is lower priced model aimed squarely at coding autonomous agents, knowledge work. meanwhile, anthropic has secured the full capacity of spacex colossus one data center with more than 220,000 nvidia gpus and 300 megawatts to expand claude's capacity. the aor is becoming less like apple versus android and more like an arms dealer supplying both sides. spacex can potentially profit whether developers choose grok or claude. what do you think about this, john?

Speaker 1:  sorry. the arms dealer supplying both sides. it's such a good analogy. i love it. so i'm not surprised here. elon has been all over the place, and, like, it's everything's been so convoluted. it's become so convoluted with him. so ai when elon musk founded xai in 2023, its mission was relatively straightforward. build, ai frontier model capable of competing with all the other big dots, you know, openai, entropic, google, all that. and so the first grok model was integrated directly into xpremium. but since then, the company has expanded dramatically. so now xai encompasses frontier, language models, vision models, coding assistance, enterprise apis, ai infrastructure, social media integration, massive gpu clusters, potential robotics applications, scientific reasoning models. and so it's i would argue, it's this rapid diversification has made it increasingly difficult for outsiders and even insiders to define what business xai is actually in. and maybe that's what elon's trying to do. maybe he's trying to make things so confusing. who knows? but, like, yeah, he's putting his hands into everything, and he's he's combining companies and all that sort of stuff. but, what do you think? he's a genius in doing this, or do you think he's kind of all over the place still? or do you think he has a plan? i think he has a plan. but

Speaker 0:  well, it's hard to say that he's not a genius.

Speaker 1:  yeah. no. i mean, in this particular case.

Speaker 0:  so i think overall, there is a plan, but at the same time, elon changes things fast based on whatever's presented at the time. that's not meaning that it goes with the flow. i'm saying that things change. meaning, we build these colossus data center. we were hoping rxeye would be great. we solved a bunch of these problems related to, these data centers, but people aren't using it. so let's rent out, the cpu. they were losing, the ai battle when they realized coding like, i don't know, six months ago, everyone realized the best coding agents was the way to go and all of a sudden switched focus. they realized they were losing the race, and that's why they bought cursor. and immediately after, built that to build grok 4.5. i've been using grok 4.5 by quite a bit the last couple of weeks, and it's, working great. i'm using it because i have cursor right now, and i'm mute building with it. and they give you a 50 extra tokens or something for using crack 4.5, and it's working pretty good. it's doing a really good job. now are they all over the place? i think a little bit, but that's because spacex, xai, tesla, robotics, space travel, satellites, are all over the place. so, yeah, neuralink and, boring and the tunneling company and stuff. and so, you just take a step back and say, he wants ai to be able to power his robots, to be able to help with the space, transition for shifts and whatnot, as well as coding and running his company. so i think there is a direction in all this, but i also think that he's real he realized that he's got all his cpu. why not get the $2,000,000,000 a month off anthropic, for renting it out? and, you know, i think that's gonna go a long way to making them profitable. again, it's very incestuous potentially. are they potentially being able to, you know, i don't get some data out of that cloud anthropic deal because they're running in their data center or whether it's a pristine hands off situation.

Speaker 1:  yeah.

Speaker 0:  i'm i'm not sure. i would imagine anthropic would make sure it's a pristine hands on situation because that's the way dario, plays it. i'm just curious. i we'll see how it goes because i think they're all kind of floundering at this. believe, and i've been saying this for a long time, the only companies really making money in the ai space is hardware companies. so now he's got a big hardware data center. he can rent it out even so that's a function where he'll start making money. the software company, except for the specific coding ones and the very specific, you know, website builders like global or whatever, the other ones aren't making money. so as we increase token pricing, as we, go through all of this, they're desperately trying to, increase how much everything costs, right

Speaker 1:  yeah.

Speaker 0:  to, get to profitability. when i say that, for example, you know, we talk about spacex and grok point, 4.5 at $6 per million output tokens. opus 4.7 is $25. 4.8 is 25. fable five is 50. gpt 5.4 was 15. $5.05 and $5.06 sol is 30. gemini is $10.12, and nine. deepseek is cheaper. version, four pro is 87¢. but you get the idea that even if it's not a 100% superior than 4.8, based on benchmarks or what you're using it, even if it's inferior, that $6 is attractive to people because, we had, like, six months of everyone was token maxing and all this stuff, and then the scenario switched because people went from $6 to $10 to $15 to $25 to $30 to $50 per million output tokens in the last year. so all of a sudden, these bills start piling up, and people like, oh my gosh. we have to have budget controls. we have to have all these things. so, we're gonna have to wait and see where these things all shake out. i don't know if cloud four point five is in chaos or their deal with cursor and using, all their data to build 4.5 has really jumped them ahead. i'm interesting to see how they embed it in tesla, they embed it in the robots, and they start, building these harnesses for the robots. this will become very, very interesting. any final thoughts on this, john?

Speaker 1:  no. i just yeah. you know, you're right. xai appears to be building an ai ecosystem. so it's got data centers, social media, robotics, autonomous vehicles. and so i think that's that's where the play is. it's the ecosystem versus openai is building ai software. so yeah.

Speaker 0:  so spacex may not need grok to defeat claude if it can make billions powering claude. is renting your supercomputer to your biggest rival a brilliant platform strategy or mission that their model is winning? and what do you think around this? is there room for everybody to play? or in now, two years from now, is xai gonna be relished to, something below and clawed higher? and where's openai fit this picture two years from now? let us know what your thoughts are in the comments below.