SpaceX Signs $6.3B Reflection AI Deal, Nvidia Eliminate Data Center Water artwork

SpaceX Signs $6.3B Reflection AI Deal, Nvidia Eliminate Data Center Water

The Elon Musk Update

June 23, 2026

In this episode, we cover SpaceX’s reported $6.3 billion compute deal with Reflection AI and why access to Nvidia GB300 chips could matter for open-source AI competition.
**SPEAKER_1** (0:00)
Welcome to the podcast. SpaceX has just signed a $6.3 billion compute deal with the open source lab Reflection AI. Nvidia's new Rubin AI server runs at 45 degrees Celsius for the coolant, and that basically kills data centers' water use, something I am super excited about. Apple is weaving AI into iOS 27 apps, and they're basically betting against the chatbot framing that mostly everyone else has decided is gonna be the winner.
EcoWave is power. A company called EcoWave Power is gonna be using Nvidia AI to run data centers on ocean waves, and AI has out persuaded human debaters in a study done by Oxford and AISI. If you wanna get all of these stories into your inbox every single morning, I have a newsletter on aichatdaily.com. That's my website that kinda goes along with this podcast. If you go over to the subscribe tab on that button, and I'll leave a link in the show notes, you can join the newsletter and get all of these stories. I did a deep dive and a link to the full article about all of these. So it's kinda like the show notes for the podcast. If you want more information, a deep dive, or in your newsletter, maybe you don't have time to listen to the full podcast, but you wanna get a rundown of each of the top five stories of the day, go subscribe to the newsletter and I'll send that to you. The link is aichatdaily.com/subscribe if you want the newsletter. Alright, let's get into this first story with SpaceX. They just signed a $6.3 billion compute deal with Reflection AI. And I haven't talked about Reflection AI in quite a long time. This is one that I did a number of podcasts ago when it was kind of announced and launched. We haven't seen a lot coming out of it. But it's, you know, evidently they have a lot going on if they're signing $6 billion compute deals. This one in particularly is going to give them access. They're an open source lab, but they're going to get access to Nvidia Gb300 chips at the Colossus II data center. This is the one near Memphis and they get that through 2029 So it is the third major company that is going to be on site. Anthropic and Google have both signed deals as well. And I think this shows that there is a massive growing investment in a lot of these open source AI development companies. I'm really excited about this one in particular. Reflection is going to pay about $150 million a month starting in July of this year. That's less than Anthropic, who's paying $1.25 billion a month. And Google is paying $920 million a month at the same site. But I mean, again, this is just another one of these mega customers that SpaceX has been able to pull in to paying for this data center that they've built. If you don't remember, the last episode I dropped on this Reflection was founded in 2024 It was actually two former Google DeepMind researchers. One thing on these deals that we're seeing at this data center that I will note is that all of them have a 90-day cancellation clause after the first three months. So I mean, I'm not saying SpaceX is trying to juice the numbers, but you could imagine, let's say, Reflection's like, hey, look, I need three months of training, compute done at your data center, and SpaceX is like, sweet, why don't you sign a three-year contract with us? You can cancel after three months if you want, but we want to make sure we got a three-year contract. Okay, I don't think this is what they're doing, but I think this is what some people's criticism has been on these particular deals, because Anthropic, Google DeepMind, and Reflection all have that. Now, I mean, when you're talking about these kind of deals, paying 1.25 billion a month or 920 million a month, those are massive sums of money, and so it kind of makes sense that if a company couldn't keep paying that, then they might need a way out. But I don't think Anthropic and Google are going to be the ones that are dropping the ball on those payments. Reflection, I'll be excited to see how they grow, but that is still a ton of money, $150 million a month. The reason why I think this matters is because SpaceX basically has accidentally become the world's biggest compute landlord for all of the top labs. They rented all this infrastructure, and originally XAI, I think, built all of those to train their Grok model, but they keep just building bigger and bigger models, and then they had SpaceX's backing to help them roll out Colossus 2, their brand new data center, and I think they have 2.3 billion in monthly revenue flowing through Colossus 2 right now, and actually I think they have an even bigger one that they're working on now. So I think this is interesting because their model isn't really just Grok, their kind of XAI models. It's really the hardware that they got for that model, and then they're now renting it out to some of the other hyperscalers. Okay, Nvidia had a big conference, and they unveiled a whole bunch of stuff. I have two of the big stories that I want to cover in the podcast today that I'm most excited about, but there's a ton of cool Nvidia news. If you care to look it up, you can go check out more of that on aichatdaily.com. I have like 10 different articles. But one of the big things I'm excited about is that Nvidia's Rubin AI servers, they're running on 45 degrees Celsius coolant, and that basically kills the data center water usage issue that we've had, that a lot of people talk about, right? There's a lot of negative PR about data centers, and maybe you're in that camp. I'm not really sure.

17 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000773910373