WEBVTT

1
00:00:00.120 --> 00:00:13.500
Y'all welcome back to Blockspace Live. Back at it for another week of beautiful news. Don't look at the market though. Everything's puking except for industrials. And Bitcoin, we will have ours. We will be vindicated.

2
00:00:14.040 --> 00:00:29.160
But we're not gonna talk about prices at the top of the show today. We are going to lead with a story of Mara being sued for allegedly not paying its chief architect, a French national, in its Exion acquisition.

3
00:00:29.930 --> 00:00:44.640
And also, there are questions, Charlie, as to whether or not they will try to invest more in the French energy giant's data center firm. That's our first story. More on that in just a second.

4
00:00:45.360 --> 00:00:58.360
Following that, we have Lagos Finance CEO Jay Patel on just to get a checkup on the market. Things are screwy. Google beat earnings last week, but then it fell 8%.

5
00:00:58.900 --> 00:01:09.720
No one really knows what's going on, but hopefully Jay can obfu- uh, de-obfuscate the landscape for us there a little bit. For our second and third stories, we have a note on Nvidia

6
00:01:10.840 --> 00:01:22.760
pledging $250 billion, backstop 250 billion, for a 10-gigawatt planned data center by OpenAI in Ohio with SB Energy, a division of SoftBank, as the developer.

7
00:01:23.230 --> 00:01:34.560
And then we will close on China's latest foray in dismantling the AI tech sector's momentum in the US and elsewhere with a memory company coming out of the shadows.

8
00:01:35.080 --> 00:01:50.150
The black swan you didn't know that you didn't want, and no one really needs. Uh, we're gonna have to argue about that last line, Colin, but I'll break that down towards the end of the show, so make sure you stay tuned.

9
00:01:50.700 --> 00:02:08.400
Blockspace goes live every weekday at 1:00 PM Eastern, featuring quick hits on the latest in AI data centers, emerging tech, and markets. We are Compute's daily show every single weekday, right around lunchtime.

10
00:02:09.479 --> 00:02:18.660
If you like what you hear, this live stream goes on to be a podcast shortly after we wrap up. And if you like the podcast or the live stream, you'll love the rest of our content.

11
00:02:18.720 --> 00:02:28.800
Make sure to head to our website, blockspace.media, to get the full suite of Blockspace content. We do a lot of written stuff and news stories.

12
00:02:29.560 --> 00:02:43.280
If you weren't aware, I don't usually call this out, we have a Telegram channel. You can find a link to that on our website and social media. This show is brought to you by CleanSpark, NASDAQ-listed ticker CLSK.

13
00:02:43.520 --> 00:02:59.300
More on CleanSpark later on in the show. Colin, this s- this first story about Mara is fun. It's spicy. There's a lot of-- We're gonna use a lot of alleged- allegedlys here to CYA, but what's, what's the TLDR here?

14
00:02:59.640 --> 00:03:11.950
Everything is alleged except for the lawsuit itself. Yeah. So we will hedge our bets accordingly at the start. All statements are assumed to be allegedly from here on out- Correct... if we don't explicitly say.

15
00:03:11.950 --> 00:03:19.720
If we don't hedge correctly. But the fact of the matter is, this is a very spicy lawsuit, Charlie. I'm gonna throw it up, our coverage from Blockspace here.

16
00:03:20.360 --> 00:03:33.880
Headline reads, "Mara's former French advisory team sues Bitcoin miner for unpaid fees. Strategy pivot." This is a legal dispute that stems from Mara's acquisition of Exion.

17
00:03:34.920 --> 00:03:46.079
So Mara purchased a controlling stake in Exion, which is the data center subsidiary of France's national grid operator, the EDF. Untangle that.

18
00:03:46.440 --> 00:03:56.480
It's like, you know, my father's sister's boyfriend's cousin's nephew's company. There's-- It's a very, uh, it's a very tangled web of association here.

19
00:03:56.580 --> 00:04:03.300
That being said, though, [lips smack] Mara closed on this controlling stake in Exion in February.

20
00:04:03.340 --> 00:04:14.040
It was rolled out with great fanfare as their entrance into the European data center market as a key pillar of their AI pivot.

21
00:04:15.280 --> 00:04:31.500
Now, the chief advisor of this acquisition, Francois Garcin, not to be confused with garcon, which means boy, and three affiliate entities are suing Mara, alleging that they weren't paid what they were contractually obligated to be paid.

22
00:04:31.530 --> 00:04:41.220
And the [laughs] funniest part about this is they're saying they allegedly weren't paid because there was a dispute about the VAT tax for the transaction.

23
00:04:41.580 --> 00:04:52.500
So for Americans who don't know what a VAT tax is, it may be the single most vexing piece of Europe's ever-burdensome regulatory and taxation environment.

24
00:04:53.160 --> 00:05:04.120
[sniffs] A value-added tax is a flat tax that is added to a number of different transactions, depending on whether it's buying food and drinks or engaging in business transactions like this one.

25
00:05:04.810 --> 00:05:15.080
[sniffs] And in this case, it was 20% of what was being paid out to these advisors, which m- I, I'm, I cannot comprehend this, Charlie.

26
00:05:15.160 --> 00:05:26.620
My American mind cannot comprehend this, just like I can't comprehend why they don't have AC in Europe. But [sniffs] going through the list of accusations here really quickly.

27
00:05:27.450 --> 00:05:49.660
[sniffs] So Francois Garcin, the founder of Argenthal, his advisory firm, and its affiliates sued Mara, alleging it concealed the true nature of its project during the Exion acquisition, and it's seeking 11.32 million, uh, euros in unpaid fees on a 2025 mandate to develop Mara's French data center interests.

28
00:05:50.180 --> 00:06:03.300
So the timeline here. Mara hired Garcin, according to the lawsuit, and his firm to spearhead the European expansion, specifically in France, including the politically sensitive acquisition of a majority stake in Exion.

29
00:06:03.680 --> 00:06:14.408
That's, again, Fr- the, uh, French AI HPC company that is a subsidiary of the EDF The state-owned utility in France. Garcin, in the lawsuit, says he delivered.

30
00:06:14.448 --> 00:06:28.838
He built relationships with EDF, TotalEnergies, Engie, and RTE, helped set up Mara France and European entities for the company, and according to his recounting, navigated a fraught French regulatory political approval process.

31
00:06:29.468 --> 00:06:37.068
And if y'all recall, we covered this story pretty in-depth earlier in the year and at the end of 2025. There was a kind of stop and start.

32
00:06:37.128 --> 00:06:46.968
There were-- There's really good reporting from a French crypto firm called The Big Whale that came out saying that the project actually might be killed because there were sovereignty concerns from French officials.

33
00:06:47.008 --> 00:07:10.128
They didn't want an American company coming in and buying up a stake in one of their national companies, and they were worried that the American company was gonna come in and loot or whatever, you know, uh, pillage the French system, the, you know, the world-renowned sy- French system for building businesses, and they weren't going to have control over what could be critical infrastructure for the age of AI.

34
00:07:11.808 --> 00:07:22.188
[sniffs] So there were all these sovereignty concerns. There was a full government investigation, and there were even, uh, criminal referrals to prosecutors over the deal. Garcin allegedly helped smooth that over.

35
00:07:22.508 --> 00:07:39.148
The deal was pushed through. It was finalized February 20th, 2026. [sniffs] Now, here's the interesting part. [sniffs] According to the lawsuit, [sniffs] Garcin and his affiliates were contractually obligated to receive

36
00:07:40.328 --> 00:07:55.348
€200,000 per month in a payment plan for a total of 2.4 million euros for an advisory fee. [sniffs] They also were subject to a 4% closing fee,

37
00:07:56.708 --> 00:08:03.988
4% of the entire acquisition cost of Xion, which currently, this-- that's called the success fee.

38
00:08:04.628 --> 00:08:12.868
And when you take all of those together, there is 11.32 million in total payments that they are saying they did not receive.

39
00:08:13.288 --> 00:08:20.828
That's exclusive of the interest and, uh, exclusive of any interest and the VAT tax, and they are also asking for $600,000 in legal fees.

40
00:08:21.488 --> 00:08:35.788
The other thing that's really interesting here, if Mara decides to exercise another tranche of investment into Xion, another €110 million, the total consideration could exceed €16 million.

41
00:08:36.898 --> 00:08:40.058
[lip smack] A few more notes here, Charlie, before I toss it to you for second take.

42
00:08:40.648 --> 00:08:58.208
What's interesting about this to me is, according to the lawsuit, Mara paid about, uh, one, uh, 1.4 million in those service fees to Garcin and his affiliates, which included VAT tax with the invoices that he sent.

43
00:08:58.248 --> 00:09:11.328
So Garc- G- Garcin added the VAT tax into the invoices on top of it, and they made about seven payments, according to the lawsuit, without dist- without disputing the VAT tax at all.

44
00:09:11.338 --> 00:09:20.528
[sniffs] But in January 2026, according to the lawsuit, their accountant, John Ellis, raised doubts, claiming other EU service providers didn't add VAT tax.

45
00:09:20.568 --> 00:09:29.748
And later, their CFO, Mara CFO, Salman Khan, asked Garcin for proof that they were actually dispersing the VAT tax to French authorities.

46
00:09:30.248 --> 00:09:44.648
[sniffs] The other interesting part about this is that in November 26-- uh, November 26th, 2025, Mara's own investment bank, Stifel, told one of their executive, uh, members [lip smack]

47
00:09:45.368 --> 00:09:59.458
that, that they also had to charge a 20% VAT tax on the Xion deal. And this executive's response, one Duncan Dickerson, who does corporate development, said, quote, "You need to pay a 20% VAT on top of banking fees?"

48
00:09:59.648 --> 00:10:11.268
End quote. Allegedly showing that Mara was aware that... was kind of unaware that VAT tax applied generally to any of this. So that's-- The VAT tax is at the core of this.

49
00:10:11.628 --> 00:10:24.498
To-- From my reading of it, from what is alleged in the lawsuit, it's almost like Mara didn't quite understand that this was going to be a consideration, and that once the bill was paid, [chuckles]

50
00:10:24.928 --> 00:10:32.668
they allegedly were like, "Well, screw that. I don't wanna pay these taxes. We got the deal through. This is crazy." [sniffs] Yeah.

51
00:10:32.838 --> 00:10:43.808
A 20%, uh, tax that you didn't account for on w- what could be a substantial amount of the project, [chuckles] like... Uh, I mean, I know these projects are ballooning in cost, but holy smokes.

52
00:10:44.288 --> 00:10:55.408
You know, uh, there are some gems in the filing here, and I'm gonna share them. We didn't include them in the article 'cause they're not really, like, good, like, uh, you know, to lead with an article in.

53
00:10:55.488 --> 00:11:06.058
But they are fun fodder to pull up and chat about. I've got one here in Section B of this lawsuit. Again, this is alleged, but the lawsuit says,

54
00:11:07.468 --> 00:11:21.418
"Thiel led Garsin to believe that Mara was retaining him and the advisor entities to pivot towards a transatlantic pro-France, pro-European strategy, and personally endorsed the vision of a France-centered partnership between Mara and France's leading energy companies.

55
00:11:22.048 --> 00:11:41.388
Thiel," again, allegedly, according to the lawsuit, "directed Garsin to close the door with, quote, 'a very loud bang' on others seeking to partner with France and to convince Macron," President Macron, "other French stakeholders that an American company like Mara's entry into France was not a Trojan horse, but represented a strategic advantage."

56
00:11:41.898 --> 00:11:50.338
I'll also remark that maybe... that I think Mara might be the easiest company name to pronounce of all the prodi- different Neoclouds. [inhales] Another one in here

57
00:11:51.288 --> 00:12:01.288
is, um, a lot of people are gonna, uh, Control + F and find that, um, Garsin confirmed that Thiel had already arranged his private jet for the trip.

58
00:12:01.348 --> 00:12:18.220
So we do have confirmation that there's a private jet flying around on Mar- on Mara's behalf. And then the last one, which I think, Colin Is one of our favorites here at the bottom of page 25 [laughs] in a, uh,

59
00:12:19.300 --> 00:12:20.680
i- in a reference here.

60
00:12:21.680 --> 00:12:35.610
Um, again, allegedly, despite the America First corporate culture at Mara, Thiel criticized the Trump administration in conversations with Garcin after this meeting, accusing it of having a Bitcoin agenda that prioritized the Trump family.

61
00:12:36.480 --> 00:12:49.020
In addition to these accusations of purported self-dealing, Thiel also expressed a low opinion of Eric Trump in derogatory terms, often using the term, quote, "dumb" when referring to him.

62
00:12:50.300 --> 00:12:57.970
I will not offer my own comments or opinion, uh, but spicy. Spicy. Spicy. If true, spicy.

63
00:12:58.160 --> 00:13:07.220
I mean, but that doesn't stop, you know, the likes of Thiel and other executives from bending the knee to Trump when it serves their interests, right?

64
00:13:07.380 --> 00:13:18.600
A- all of these Bitcoin mining executives were throwing themselves at the feet of that ticket because, quite frankly, they didn't have a good viable option.

65
00:13:18.660 --> 00:13:30.360
The other party was actively trying to dismantle their industry. Yeah, I'm not gonna lie, I'm also gonna stand this a little bit too because in as much as it looks in retrospect pretty bad, like very sycophantic.

66
00:13:30.440 --> 00:13:43.010
Again, Trumps, Trumps and his family have enriched themselves very, like, extractively. Um- So much so that the new draft of the CLARITY Act has a provision that bars public officials- [laughs] Oh, my God...

67
00:13:43.020 --> 00:13:52.940
from engaging in promoting or issuing cryptocurrencies. Oh, and you could, you know they just be like, "Okay, we've already profited. You can put that in there." Maybe that's why CLARITY's taking so long.

68
00:13:53.180 --> 00:14:03.580
Well, anyway, but, you know, those of us who've been in this be like, well, you know, there was a time when we would do just about anything to, uh, at least be able to have bank accounts and do business.

69
00:14:03.680 --> 00:14:15.300
So, um, you know, as much as I like to, in a little way, like mock it and joke about this, like this was the reality, like we kinda had our backs up against the wall. Um-

70
00:14:16.260 --> 00:14:27.600
We had our backs up against the wall, and then we levied the president to be to help us get off the wall. Yeah. And he did that, and then he peed on the wall by- Yeah... issuing a meme coin.

71
00:14:27.720 --> 00:14:38.680
Like the, like the Calvin- And- Yeah. Yeah... yeah, exactly. Uh, one, one last really important thing to note here, Charlie, before we sign off and have Jay on, or sign off on this topic and have Jay on.

72
00:14:40.080 --> 00:14:41.860
F- firstly, kind of a fun aside,

73
00:14:42.840 --> 00:14:58.420
according to the lawsuit, you know, Garcin's trying to build this case that they're only a pro-America culture, and the lawsuit alleges internal mes- messages show an America First culture at Mara, including a Make Mara Great Again WhatsApp [laughs] message from Thiel's Chi- Chief of Staff.

74
00:14:58.460 --> 00:15:07.850
Not gonna lie, good, that's a good title for a group text. Yeah. I mean, you know what? Yeah. O- out of all the derivatives of Make America Great Again, of course they're gonna do that. Yeah. You know?

75
00:15:07.920 --> 00:15:19.580
And who, who- They could sell Mara, you know, Mara MAGA, whatever hats. I don't know. Also, like yeah, a, an American company built in America by Americans is America First. I, I mean, color me shocked, Mr. Garcin.

76
00:15:19.720 --> 00:15:22.720
I think you got taken for a ride there if you're surprised by this.

77
00:15:23.320 --> 00:15:53.840
But the last thing that I find really, really interesting about this is on February 7th, 2026, according to the lawsuit, again, this, they're alleging this, Thiel sent a message to Duncan Dickerson, the guy in corporate development we covered earlier, and Zabi Nouei, General Counsel and Corporate Secretary, stating, quote, "Between the cost of advisors, the various members of the management team who are all asking for pieces of this business, and the complexities of doing deals in France, I think it may be time for us to pack up our tent and focus on other areas."

78
00:15:54.400 --> 00:16:06.350
Salman Khan, the CFO of Mara, then seized on this, encouraging Thiel to abandon the Exion transaction altogether. Quote, "The best deals are the ones we don't do. I wholeheartedly agree with this assessment." End quote.

79
00:16:07.310 --> 00:16:19.940
That was, that was 13 days before the transaction closed. Now, clearly they didn't back out. They ended up spending 120 million euros, and then an additional 33 million for a 64% stake.

80
00:16:19.980 --> 00:16:31.300
But I do think this kind of casts into doubt, if those texts are true, whether or not they will exercise a third transaction for another $120 million investment into Exion.

81
00:16:32.019 --> 00:16:44.420
And that would include, at 110 million, a 4.4 million euro fee plus VAT to Garcin and his affiliates for the success fee according to the original contract.

82
00:16:44.940 --> 00:16:55.660
Uh, you know, if they were thinking about pulling out of this thing back then, I do question whether or not they will continue to push forward, especially when you consider, I was asking Claude about some specifics with Exion.

83
00:16:56.000 --> 00:17:09.480
Uh, reportedly, they only had, like, 5 million euro in revenue last year. I mean, w- what are we doing? I don't know. Maybe for a European company, that's, that's huge. I, heck if I know. [laughs] Yeah.

84
00:17:09.520 --> 00:17:19.180
[laughs] That goes entirely to- If we're a European company, that's enough to take eight weeks off, and you can have Aperol Spritzes on the coast of France in Saint-Tropez or something like that. Yeah.

85
00:17:19.480 --> 00:17:30.100
But, you know, I, so all that being said, the whole thing is really messy. I think the Exion acquisition was really a head-scratcher for a lot of us when we were covering it.

86
00:17:30.120 --> 00:17:32.120
And I'll, I'll just editorialize here a little bit.

87
00:17:32.190 --> 00:17:47.550
When all of your peers are building these massive data centers, inking deals with Mag 7 clients or with clients that are associated with Mag 7 companies, what are, why are you, why are you doing this acquisition in Europe to begin with?

88
00:17:48.589 --> 00:17:51.050
The, the, it's not a friendly regulatory environment.

89
00:17:51.180 --> 00:18:01.900
Clearly, the taxation was so much so that Mara was allegedly blindsided by that, and they, they just don't have a culture of building over there in the same way we do here.

90
00:18:02.620 --> 00:18:08.830
And, and, and to me- They got a bunch of nuclear power. I mean, we know that They do have a bunch of nuclear power, and that is the strongest angle for this.

91
00:18:08.840 --> 00:18:18.620
France actually is the dominant nuclear power in Europe on an energy standpoint. I believe it's, like, 75% to 80% of their energy comes from nuclear, so much so that they're a net exporter of it.

92
00:18:19.340 --> 00:18:26.829
And that was part of the sovereignty concerns with this. They wanna find ways to actually capitalize that on a h- at, that at home. And so that's, that's fine.

93
00:18:27.360 --> 00:18:39.230
And perhaps Mera takes Axion and builds a very lucrative AI business out of it. It just seemed like they felt like they needed to do something to get their foot in the door, and they decided to do this.

94
00:18:39.230 --> 00:18:46.810
And instead they put their foot in their mouth. I don't know. I'm just, I'm surgical. No, I... That's a good point. We- I, I think overall though,

95
00:18:47.820 --> 00:18:52.260
there's a chance that we look back at this and say, "Why did they ever think this was a good idea in the first place?"

96
00:18:52.340 --> 00:19:04.140
Because th- you know, all these other companies that you're competing with are doing deals in the US, and clearly there's a premium on that. 'Cause this didn't really do very much for Mera stock at the time.

97
00:19:04.220 --> 00:19:18.660
Last thing, sorry that you just joined L- ELZ- [laughs]... X9G. Lawsuit TLDR is one of Mera's chief partners for this Axion acquisition is suing them because they allege they were not paid.

98
00:19:19.040 --> 00:19:29.680
We will have a clip on this up on YouTube after the show. You can also rewind the live stream to hear this, so if you'd like to learn more about it in depth, please do so. More dramatic than an episode of Love Island.

99
00:19:29.800 --> 00:19:41.640
Call it Love Iberian Peninsula. Let's roll an ad from our sponsor, CleanSpark, and then we'll go to Jay Patel. [gentle music] We are CleanSpark, America's Bitcoin miner.

100
00:19:42.240 --> 00:19:59.580
A publicly traded company with the largest operating hash rate. Powered entirely by self-operated infrastructure across four states. This is our proof of work. We are setting the standard for what's next.

101
00:20:00.500 --> 00:20:17.320
Learn more about the intersection of energy and Bitcoin at cleanspark.com. All right, we've got Ligos Finance CEO, Jay Patel, in the wings. Let's bring him on up here.

102
00:20:17.420 --> 00:20:31.080
Jay, welcome back to the show. How you guys doing? Great. Uh, let's get a quick temperature, like, market, uh, temperature check, 'cause it's, we've seen a little bit of pullback today. It's just Monday.

103
00:20:31.440 --> 00:20:40.840
Nothing major happened over the weekend that I could tell. Something about Iran, but again, how much does that matter anymore? Uh, drop us in hot, Jay. What do you think about the markets right now?

104
00:20:42.040 --> 00:20:50.900
Uh, honestly, I'm, um... A- and I don't wanna spend too much time on the MicroStrategy piece, but I actually think the m- market, Bitcoin at least, is in a pretty good spot.

105
00:20:51.080 --> 00:21:01.010
I think there's probably not too much you can read into Iran because the, the war is simultaneously over and ongoing at all times. Um, the fact- There is no war with East Asia. Yeah. Yeah.

106
00:21:01.020 --> 00:21:04.620
We are always at war with East Asia. There... We were always at war with Persia. Yeah. Yeah.

107
00:21:05.220 --> 00:21:17.340
The fact that there's still kind of, I wouldn't say strong performance in Bitcoin, but like, um, you know, it's been fairly decent knowing that Sailor's not gonna be buying now or probably anytime soon.

108
00:21:17.380 --> 00:21:26.780
And you know, for everyone who was like, "Oh, the only reason Bitcoin's still above 55K is because Sailor was buying for so long," um, like clearly there's room for us to rally without him.

109
00:21:26.920 --> 00:21:31.220
Um, and so I dunno, from that perspective, it makes me a little more confident.

110
00:21:31.260 --> 00:21:45.980
I'm also, I don't know if it's me personally tired or just the market tired of it, but it's like, I think we might be able to move on to the next part of this cycle with or without this whole STRC thing resolved, and it, you know, if we can do that, that's even better.

111
00:21:46.120 --> 00:21:56.060
Um, but I, like, I don't think it's the overhang where, you know, the doomers were like, "Sailor's gonna bring Bitcoin down with him." Like- I agree with that... I don't think anyone can do that, not even him.

112
00:21:56.120 --> 00:22:08.220
I'm really tired of Sailor being the main character. I mean, when your main character is posting AI slop videos of some babe in a bikini sipping a piña colada on the, you know, pool bar because she bought Stretch.

113
00:22:08.560 --> 00:22:17.640
You deserve to live comfortably. And, yeah. I mean- [laughs]... you know, you deserve to live comfortably, you know. I take on all the risk. You get all the reward. I mean, I'm just, I'm tired of it.

114
00:22:17.720 --> 00:22:19.140
I want him, I want him to go away.

115
00:22:19.820 --> 00:22:31.140
Uh, so we won't spend too much time on strategy here then, Jay, but I do wanna ask you, one of the more puzzling features of last week, there was a, a generally a big route in tech stocks.

116
00:22:31.500 --> 00:22:45.060
Tesla fell like 15, or like tw- somewhere between 10 and 15% on a single trading day, just absolutely hemorrhaging. But one of the more puzzling ones is Alphabet also had a horrible week last week.

117
00:22:45.080 --> 00:23:01.860
It was down almost 9% on the week, and this was after it beat earnings. And it rep- it beat its, it beat revenue expectations. But two things worth noting is while the earnings per share came in at $9.11,

118
00:23:03.040 --> 00:23:13.440
the actual adjusted earnings per share when you took out paper gains from its investments in Anthropic and SpaceX were about $2.85 versus an estimate of $2.89.

119
00:23:14.120 --> 00:23:23.880
Couple this with the fact that they up- they revised upwards their 2026 CapEx guidance from 180 to 190 billion, to 195 and 205 billion.

120
00:23:23.960 --> 00:23:32.680
Seemed like people were just, investors were not really happy with that quarter, despite the fact that their cloud revenue was up 82% year over year. Square this for me.

121
00:23:32.720 --> 00:23:44.680
What are investors saying about where we are in the current AI CapEx cycle, given what was presented during Google's call and the fact that the stock tanked afterwards? Yeah. I think, um, well, I'll say two things.

122
00:23:44.720 --> 00:23:57.210
I'll say one, we're kind of in this like, it seems like the last 12 months we've went into this kind of hyper-accelerated AI bearishness to bullishness kind of oscillation.

123
00:23:57.240 --> 00:24:05.260
Even though I'll say the overlying, the underlying theme is definitely that market is bullish on AI. You know, there's still a bunch of CapEx and financing going into the space.

124
00:24:05.280 --> 00:24:11.720
But, you know, if you think back to when, uh, DeepSeek V3 and R1 dropped, there was the, the, the comedown.

125
00:24:11.980 --> 00:24:19.800
Um, and then, you know, a few months later, I think there was that Goldman paper about like, will there actually be an ROI on AI and, you know, all the stocks dropped again.

126
00:24:20.340 --> 00:24:30.546
And it seems like we go between that and, like, there will be infinite returns, kind of like- Every few months, and so maybe we're just on that trough. The one thing I will say on the Alphabet side of things, though, is,

127
00:24:31.676 --> 00:24:43.156
um... Well, I'll say two points. One is obviously it's hard to really ascribe too much value to the whole, like, SpaceX and Anthropic investments in the sense that, like, these were investments they made a long time ago.

128
00:24:43.216 --> 00:24:52.196
I don't think investors are gonna give them additional credit for making them a long time ago. Um, on, on the, on the revenue side of things, I think one thing is very clear.

129
00:24:52.316 --> 00:25:02.136
Like, these businesses went from generating a ton of free cash flow to almost all being kind of flat or cash flow negative. Um, and we can get into the hyperscaler cash flow story.

130
00:25:02.236 --> 00:25:12.456
The other, though, is that if you think about GCP cloud revenue, I-- And look, I think these are impressive businesses. You'll hear all of these, like, venture investors be like, "Oh my God, here's an a, a billion...

131
00:25:12.476 --> 00:25:19.316
You know, a multi-hundred billion dollar market that can skill-- still grow, you know, forty percent plus year over year. We have, like, infinite TAM."

132
00:25:19.916 --> 00:25:31.816
I think one thing that you have to remember with GCP, though, is that GCP growth used to be very low marginal cost in terms of, you know, the, the, the, the marginal revenue that they brought in, right?

133
00:25:31.876 --> 00:25:36.196
If you're providing, you know, on-demand CPU and storage, like, that's not very expensive.

134
00:25:36.296 --> 00:25:44.416
I think a lot of the new hyperscaler demand and revenue is around GPUs, which is obviously much more expensive, and there's a bunch of build-out.

135
00:25:44.826 --> 00:25:54.596
Um, I think one thing that's interesting, though, if you think about this, is, um, you know, in my head, I'm almost thinking, like, it's clear that the market doesn't believe that GPU demand is going away.

136
00:25:54.736 --> 00:26:03.496
Because, like, if you look at H100 or B200 prices, they're still in contango. Like, people are paying a significant premium to lock up demand for the future.

137
00:26:03.556 --> 00:26:12.936
So I believe that there's something else that the market is pricing in, which is why they're not giving Alphabet and others the full credit for this expected future revenue.

138
00:26:12.956 --> 00:26:27.196
And you know, my mental model for that is basically, like, if you have all this contracted revenue in the future, the risk from, from Alphabet's perspective is basically, you know, if you're looking at a GPU deal, it's whatever your expected realized GPU price is times utilization.

139
00:26:27.236 --> 00:26:33.036
Let's assume utilization is high 'cause all of us are gonna be generating memes on Groq and, you know, doing whatever we do.

140
00:26:33.235 --> 00:26:43.636
Um, there's, there's power, which I think power is expensive, but there's a bunch of folks coming online providing new, potentially diversified sources of power. There's operations.

141
00:26:43.936 --> 00:26:51.516
You know, let's assume that the hyperscalers have good OpEx costs, because if they don't, then I'm guessing all the Neo clouds are probably in a tough spot, too.

142
00:26:51.576 --> 00:26:58.096
Depreciation, I think, you know, if H100s are still cl- kicking around, I don't think depreciation's a very g- big risk. And then I think the last is the financing cost.

143
00:26:58.276 --> 00:27:04.796
So maybe financing cost is what the market's clinging onto, which is like they burn through their free cash, their cash pile.

144
00:27:04.836 --> 00:27:09.446
If they're gonna go cash flow negative and have to start borrowing a lot, what is the rate they're gonna be able to borrow at?

145
00:27:09.456 --> 00:27:18.676
You know, the 10-year is basically making new highs last week, so maybe that's part of the story there is, like, even if GPU demand is there, how expensive is it gonna be for them to continue to build out?

146
00:27:19.456 --> 00:27:22.416
That's a really good point. I mean, rates have been getting blown out recently.

147
00:27:22.516 --> 00:27:33.936
If you look actually at the curve, shorter-term rates have been moving more violently because investors are pricing in short-term or near-term inflation, um- Mm-hmm... vis-a-vis the Iran war and more oil disruptions.

148
00:27:34.476 --> 00:27:44.936
WTI is back up at, like, 90, I think, or thereabouts. And so the spreads for a lot of these bonds are actually getting blown out as well for these corporate bonds for the, the data center CapEx.

149
00:27:45.736 --> 00:27:58.276
And my unsophisticated take on it is I'm starting to get really anxious when I just look at how much is being financed or committed, and my thought is where the hell is all of this [chuckles] liquidity going to come from?

150
00:27:58.336 --> 00:28:09.256
You know, it's all fine and good to say the Mag Seven's gonna spend, you know, just south of a trillion dollars in CapEx, you know, in, in 2026. That's great,

151
00:28:10.616 --> 00:28:19.036
but I mean, do we even have enough plumbing in the system for that as it stands without something either seriously breaking or someone getting seriously overextended?

152
00:28:19.956 --> 00:28:27.336
Yeah, I think that's probably the right way to think about it. I hate dumbing it down to such a simple model, because it's obviously much more complex.

153
00:28:27.476 --> 00:28:40.656
But if you think about there's some fixed supply of dollars that's willing to lend into this market, and both the US government is probably borrowing at all-time highs. All of the hyperscalers are tapping the market.

154
00:28:41.156 --> 00:28:51.336
Even NVIDIA, which is, like, the furthest along this chain, you know, printing cash are-- You know, they're doing bond issuances. So, you know, clearly there's immense demand for capital.

155
00:28:51.816 --> 00:28:59.866
I think you have to think, like, "Okay, maybe we can satisfy this year's." But, like, are we gonna be able to do this plus twenty percent next year and the year after and the year after? Like, it seems hard.

156
00:29:00.596 --> 00:29:13.176
Um, and then on the other side, I'm thinking from these hyperscalers, like, yes, this contracted revenue is great. Um, but a GPU is not gonna get more than 100% utilization, right?

157
00:29:13.276 --> 00:29:20.616
So you kind of have a cap in terms of, like, how much can you monetize a single GPU. You can't, like, sell more than the capacity of the GPU that you have.

158
00:29:20.656 --> 00:29:28.976
But you could get screwed on financing costs, and I think you could still also get screwed on power and build-out costs for everything outside of the GPU.

159
00:29:29.076 --> 00:29:41.506
Like we talked about it a few weeks ago, there's almost, like, you know, all of the less efficient components in standing up these data centers might inflate the cost, and then you get kind of hammered on both sides where it's more expensive to finance all of this.

160
00:29:41.516 --> 00:29:51.976
It's more expensive to build it out. And if you've contracted, um, if you have contracted compute obligations in the future, it's not like OpenAI or Anthropic or whoever else is gonna be paying up more.

161
00:29:52.016 --> 00:30:05.536
I guess, you know, they're going to keep contracting into the future, and maybe you can charge them further out. But, um, you know, that is a risk. Um, you know, again, triple tapping on free cash flow.

162
00:30:05.616 --> 00:30:22.516
Epoch AI put out this article last month kind of showing operating cash flow, showing that the, the inflection point for the hyperscalers happens this year. Um, and y- we- you've touched on this previously.

163
00:30:22.536 --> 00:30:37.312
We talked about, you know, the memory stocks, Samsung, SK Hynix. But- Is there a, a free cash flow trend, maybe short-term, in other sectors in the AI market that are not, say, the Mag 7 or hyperscalers?

164
00:30:37.892 --> 00:30:44.052
Would you, do you see the market prior that wants free cash flow to go other places? Where do you see that?

165
00:30:45.252 --> 00:30:58.292
I, I'm, I'm skeptical of, you know, a, a broad-based market rotation into chips or even energy infrastructure for that matter. 'Cause I think it's a completely different business to underwrite.

166
00:30:58.332 --> 00:31:10.812
Like a, you know, if you were historically an investor in the Mag 7, it's not just that they were free cash flow positive, high margin businesses, it's that they had what you, you know, at one point thought were immense moats, right?

167
00:31:10.992 --> 00:31:21.552
You know, Meta basically had, you know, captive audience of customers. You know, Google, the same. Um, I think maybe Apple's the only one left that's really like that, right?

168
00:31:22.072 --> 00:31:28.892
Um, I, I think if you think about, like, the memory manufacturers, you know, I think the reason that they trade at such low multiples...

169
00:31:29.052 --> 00:31:40.952
Granted, the stocks have rallied like crazy 'cause they're just printing money, but they trade at low multiples to earnings, is because I think the market knows that SK Hynix and Samsung are not going to keep printing, you know, tens, hundreds of billions of dollars a year for very long.

170
00:31:41.052 --> 00:31:48.912
Someone will come in and compete. You know, there's CXMT. I'm sure there's others. Um, and so I don't think that's as defensible and I, so I don't think you'll see, like, a rotation there.

171
00:31:50.212 --> 00:32:01.132
Y- [laughs] you know, like, in a weird world, do you see a capital rotation where, like, now the memory manu- manufacturers are investing their excess profits back into the other end of the supply chain and, like, you really get the circular financing [laughs] going?

172
00:32:01.172 --> 00:32:15.412
I don't know. Um, but I- I- I don't see, like, rotation out of Mag 7 straight into chip stocks or anything like that. So yeah. Do you see, uh, do you see rotation back into Bitcoin?

173
00:32:15.772 --> 00:32:33.652
[laughs] I mean, you know, what, at what point do we go full circle with this or are we just gonna be stuck in limbo for the next year or so until this- Well, I mean, I think, so, so for Bitcoin, I'd say the, uh, maybe one of the near term catalysts is, like, the US economy is incredibly, incredibly tied to AI now.

174
00:32:34.212 --> 00:32:44.152
And I don't think we can afford to let this flywheel slow down. So if at some point it's, can the US government print a little bit more money to keep this wheel going?

175
00:32:44.232 --> 00:32:47.032
Like, I think that might be the catalyst for Bitcoin.

176
00:32:47.112 --> 00:33:01.571
Obviously, you know, there's the short-term inflation fears with Iran and, and supply chain stuff, but I think you might get that inflation either way because you don't want the kind of build-out and CapEx to slow down, especially when that's, like, most of the economic growth.

177
00:33:02.192 --> 00:33:03.832
I'll say one interesting thing though is, like,

178
00:33:04.852 --> 00:33:17.321
you would think y- uh, I guess one counter to this thesis is, like, you would think that the companies in the kind of energy supply chain, you know, the ones that own the power, maybe even to some extent some of the Bitcoin miners, would be doing better.

179
00:33:17.672 --> 00:33:26.952
Because at least if you're a hyperscaler, that becomes more attractive for you to hedge out one of the components, right? You know, you've got depreciation under control, ops under control.

180
00:33:27.112 --> 00:33:29.932
Let's say you assume that you can, like, you know, manage your financing costs.

181
00:33:30.072 --> 00:33:38.772
If you can make sure that you have reliable access to power at a good cost, like, maybe that should be even at more a premium, but I think even most of those stocks have just gotten dragged down with the market.

182
00:33:38.872 --> 00:33:53.252
Um, although maybe it's also, you know, the market doesn't love their other businesses. But yeah. We will see. It could, it, we could, yeah. Let's try to avoid a snake financing its own tail scenario.

183
00:33:53.342 --> 00:34:05.032
[laughs] And whatever can, whatever we can construct, which is a future bullish scenario for Bitcoin, I'm, I'm here for. Jay, thank you. Yeah. You thought, you thought, uh, Nvidia investing in the labs was circular.

184
00:34:05.252 --> 00:34:18.772
Imagine SK Hynix financing Nvidia's debt. Oh. [laughs] Oh, man. [laughs] Okay. Jay, thank you so much for your insights. Appreciate it. And, uh, yeah, the markets are always interesting. AI is super fun to talk about.

185
00:34:19.212 --> 00:34:36.812
We'll have you back on soon. All right. See you guys. Adios. Thank you, Jay. It's like a human centipede of circular finance. [laughs] We, we want the show to get big. We gotta avoid spicy references like that. Okay.

186
00:34:37.352 --> 00:34:49.312
Uh, we're gonna keep on rolling. We're gonna talk, again, financing, Nvidia backstopping OpenAI for a, a casual quarter trillion. And then we're gonna go to memory again.

187
00:34:49.512 --> 00:34:58.092
CXMT becomes the largest publicly traded company in mainland China. Before that, a word from our sponsor, Luxor.

188
00:34:58.952 --> 00:35:10.592
[static] This episode of Block Space Live is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations.

189
00:35:11.272 --> 00:35:16.912
Luxor's Commander gives you real-time fleet monitoring, bold remote commands across your fleet, and Intelligent Miner.

190
00:35:17.412 --> 00:35:24.852
That's an automated profitability engine that runs every five minutes and adjusts your fleet's power settings to live hash rate in energy markets.

191
00:35:25.512 --> 00:35:39.572
Our cutback tests show 10% improved profitability with intelligent mining over binary mining. Commander Pro is $100 per megawatt or a 25 basis point pool fee adder. But you could also try it for free for 60 days.

192
00:35:39.632 --> 00:35:51.872
So if you'd like to learn more, go to luxor.tech/commander. Yeah, speaking of circular financing, Charlie, we're gonna go ahead and hop on over to Nvidia. Nvidia, excuse me,

193
00:35:53.332 --> 00:36:01.572
and talk about their reported $250 billion backstop for OpenAI's planned 10 gigawatt data center in Ohio.

194
00:36:02.312 --> 00:36:14.392
Now, this is coming to us by way of Data Center Dynamics, which is r- reporting from Wall Street Journal, who I believe had the inside scoop on this. But the TLDR is as such.

195
00:36:14.832 --> 00:36:27.952
Nvidia, Nvidia plans, they're in talks to provide a 250 billion, yes, B, billion, financial backstop for a massive OpenAI/SoftBank data center planned in southern Ohio.

196
00:36:28.412 --> 00:36:38.992
We actually covered this on the June 11th live stream. This is a proposed 10 gigawatt data center on Department of Energy land, which used to be a uranium enrichment site.

197
00:36:39.032 --> 00:36:47.552
The first phase is a much smaller chunk of that 10 gigawatts, 800 megawatts, and the developer is SB Energy, which is a division of SoftBank.

198
00:36:48.612 --> 00:37:04.372
SoftBank has been one of the biggest movers with regards to AI in the US from a foreign actor. The, uh, the bank was reportedly a part of the Project Stargate initiative, which has somewhat stalled.

199
00:37:04.451 --> 00:37:18.072
But the Japanese, uh, multinational company is one of the foremost companies involved in the development specifically of some of these data centers and the energy pipelines for them.

200
00:37:18.612 --> 00:37:32.032
The total cost estimate, Charlie, is $500 billion for this site. I'm not sure if that's inclusive of GPUs. I assume not, which is kinda crazy.

201
00:37:32.052 --> 00:37:42.832
But the first phase is expected in 2028 at around 800 megawatts. Like I said earlier, the project was first amount- announced in March, and again, SB Energy is developing it.

202
00:37:43.432 --> 00:37:55.712
This will include 10 gigawatts of new power generation to feed the site, including 9.2 gigawatts of natural gas generation. That is a lot of methane, my guy. That is huge.

203
00:37:56.292 --> 00:38:06.422
Japan, the country itself, has agreed to invest $33 billion in the natural gas power project as part of its broader trade deal with the United States.

204
00:38:07.052 --> 00:38:19.332
The US government will reportedly pay SB Energy to operate the project. There will be a revenue split until Japan recoups its $33 billion investment, after which point the US will get 90% of the revenue.

205
00:38:20.112 --> 00:38:27.992
Now, as for what NVIDIA's backstop actually does, the details are still under negotiation. Seems like this is still in the early phase.

206
00:38:28.012 --> 00:38:39.712
But the idea is that NVIDIA would guarantee financial, financing vehicles behind the project, specifically for the data center build-out and lease, not for the GPUs themselves. That's an important distinction.

207
00:38:41.132 --> 00:38:55.252
OpenAI will be operating those GPUs. SB Energy will be owning the data center, the powered shell, and Nvidia will, NVIDIA will, will, will supply the financing or the backstop for the financing and the GPUs.

208
00:38:55.312 --> 00:39:17.972
Now, importantly, just to untangle some of OpenAI and NVIDIA's financial relationships to date, this is separate from NVIDIA and OpenAI's $30 billion investment or N- NVIDIA's investment into OpenAI for $30 billion, and is also separate from talks to finance up to 350 billion in chip sales to the company.

209
00:39:18.672 --> 00:39:31.772
So if this comes to pass, you're talking about from NVIDIA a total of $600 billion worth of either direct financing or promises for backstops for these AI data centers.

210
00:39:32.532 --> 00:39:43.712
Once again, Charlie, I keep saying it, the numbers kind of stop making sense at a certain point. You can't even really begin to fathom how much money is sloshing around in this CapEx cycle.

211
00:39:43.792 --> 00:39:54.072
So maybe instead of looking at the money, let's kinda look at the actual physical footprint. This is kinda where I'm gonna try to add a little bit of color outside of the normal dollar figure for things.

212
00:39:54.152 --> 00:39:56.892
Um, where is this? Why is this being built in Ohio?

213
00:39:57.072 --> 00:40:17.772
Well, because, like Neel Kashkari said about the Federal Reserve, the United States has effectively infinite natural gas, and a lot of the natural gas is in what's called the Appalachian Basin, which goes right over the eastern half of Ohio.

214
00:40:17.792 --> 00:40:32.952
And there's two, uh, plays, shale plays there, the Utica and Marcellus. They're at kind of different depths along the strata, created by different geological events. And these are very mature oil and gas fields.

215
00:40:33.032 --> 00:40:38.932
They're actually primarily, they're almost more gas fields than they are oil, and so they've been developed for a long time.

216
00:40:39.732 --> 00:40:54.392
And due to very path-dependent nature of geography and regulations and the market, a lot of that gas does not have abundant transportation to markets and refining facilities.

217
00:40:55.852 --> 00:41:11.772
So if you'd go to the middle c- the middle of Appalachian Basin, you can have infinite gas, basically, and it trades often at a negative, uh, price, uh, even to that of Henry Hub, which is the regional pricing metric of West Texas.

218
00:41:12.932 --> 00:41:15.592
And because of this, um, it's...

219
00:41:15.852 --> 00:41:34.432
If you, uh, because of actually, like, transport rules and, like, certain, like, cross-state taxes, I was looking into this, um, you could build, say, a 10-gigawatt plant in Ohio, and you have some, uh, actual structural, like, regulatory advantages regarding to, like, how you, where you can consume and transport the gas.

220
00:41:34.812 --> 00:41:48.492
How much gas we talking about, Colin? 1.5 billion cubic feet per day of natural gas, which is roughly 5% of current total US production.

221
00:41:48.852 --> 00:42:00.312
So that would be basically how much natural gas it takes to run through a 10-gigawatt, uh, you know, 10 gigawatts of, of turbines and generation to a data center. Now- Wait, wait, wait.

222
00:42:00.432 --> 00:42:09.972
So that, that's not how much gas is in the Appalachian Basin. No, no, that's how much it would, uh, that's how much- Takes to power a 10-gigawatt-... a 10-gigawatt site.

223
00:42:10.352 --> 00:42:20.252
That's how much a 10-gigawatt site running solely off natural gas at the standard turbine efficiency and standard, like, uh, you know- 5% of the current- In-... daily production capacity of the entire US.

224
00:42:20.312 --> 00:42:31.182
Of the entire United States. Now, the, the United States, I'll, I will emphasize this, really, really dynamic production. So if there's demand, we can get it, and we can transport it, and we can refine it. Uh,

225
00:42:32.712 --> 00:42:45.242
that demand can be met very quickly. Um, if you were to basically cut through all the red tape, we could, we could ramp up production really, really quickly. So I think that's not as, like, indicative of, like,

226
00:42:46.212 --> 00:42:56.268
uh, I don't think that's as, as indicative of what, uh, the real constraint is here. That's just giving you a reference point for how much natural gas this site would... consume.

227
00:42:57.078 --> 00:43:00.867
There's another dimension to this, and I've said this a few times already, which is turbines.

228
00:43:01.408 --> 00:43:22.488
And this is where I think maybe one of the other real moats is besides the financing, besides the money, besides the region, and that is SB Energy, the power, uh, the energy side of this equation, has already placed, a while ago, a $10 billion order for nearly 170 gas turbines from GE.

229
00:43:23.948 --> 00:43:33.968
And the first of those is expected within a year, and the rest anticipated to come online within a decade. Or sorry, by the end of the decade, so by 2030.

230
00:43:34.888 --> 00:43:45.668
Um, the large manufacturers of these turbines, GE, Siemens, and Mitsubishi, are sold out years in the future. So this actually gets into the behind the meter trend going on.

231
00:43:45.688 --> 00:43:54.918
Because in order to go behind the meter in any state, Oklahoma, my, my neck of the woods notwithstanding, uh, you're basically gonna have to acquire gas turbines.

232
00:43:55.108 --> 00:44:12.628
And they are sold out, and they're especially sold out right now. So a $10 billion 170 turbine, like, order, um, basically means that SB Energy has cornered a little chunk of this market here.

233
00:44:13.588 --> 00:44:22.578
Uh, so you could argue that this is actually more of a moat than, uh, well like more of a competitive moat than an N- NVIDIA, uh, financing deal.

234
00:44:22.668 --> 00:44:34.648
Again, I'm gonna be more biased because I think I understand gas a little bit more than dollars. But, uh, yeah, really interesting kind of dimension to this, which is where's all the gas coming from?

235
00:44:34.708 --> 00:44:41.968
Where's the energy coming from? How are they gonna turn it on? And the answer is, uh, American shale. American gas.

236
00:44:42.088 --> 00:44:54.508
So- A- and you could see this being maybe much smaller in, in reality than what's being put on paper now. 10 gigawatts would make it the largest data center in the world several times over, if they got it off the ground.

237
00:44:55.548 --> 00:45:03.348
Yeah. 800 megawatts itself is big enough, right? That's a huge undertaking, especially if you're having to bring your own energy.

238
00:45:04.128 --> 00:45:15.688
N- and I, I do think it's interesting and notable that the United States government is taking a stake in this in the sense that they are helping broker the investment from Japan get part of the nat gas-

239
00:45:16.768 --> 00:45:26.828
I mean, at the, at the scale, this has- Get part of the nat gas offline or on the- At-... uh, you know, the plant online. Yeah, at this scale, like this is a nation state project.

240
00:45:27.308 --> 00:45:39.518
Even though the money is private, like it's so, it becomes so critical [laughs] to like the overall energy production that it may as well, it- it, you know, probably should have some state, uh, involvement in this.

241
00:45:39.588 --> 00:45:48.568
Right. It's like the Manhattan Project, except private companies are leading it, and the government's kinda taking a backseat and doing a little backseat driving, saying- Yeah, we're-... "Hey, actually, don't do that.

242
00:45:48.648 --> 00:46:03.288
Go here." Yeah, we're building like several Hoover Dams worth of stuff. It's, it's like these type, that scale of like civilizational, civilizationally, c- civilizational scale build. Um, so yeah, wild story.

243
00:46:03.508 --> 00:46:19.108
Um, we are gonna keep on rolling. We are gonna go to Chinese memory technologies, uh, ChangXin Memory Technologies, and their record-breaking Chinese IPO after a word from our sponsor, Ligos.

244
00:46:20.308 --> 00:46:29.608
[static sound] Bitcoin is charting a shaky recovery, my friends.

245
00:46:30.708 --> 00:46:44.088
But if you're using a Bitcoin-backed lender that's playing funny games with your coin, it- coins, it doesn't really matter what Bitcoin's price is at, because you may always be at the risk of becoming the next Celsius or BlockFi victim.

246
00:46:44.668 --> 00:46:52.768
So do yourself a favor. If you're looking into Bitcoin-backed lending or you are using a Bitcoin-backed lender, look no further than Ligos Finance.

247
00:46:53.168 --> 00:47:02.928
They are our preferred lender here at Blockspace because they use Bitcoin native smart contracts to make sure that you always are in control of your keys. That's right.

248
00:47:03.388 --> 00:47:18.228
With Ligos, you always have custody of your Bitcoin and nobody else. With Ligos Finance, there's no rehypothecation, no funny business, or no wrapping, no bridging, just competitive rates as low as 8.5 to 10% APR.

249
00:47:18.628 --> 00:47:33.648
Go to ligos.finance to learn more. All right. So Chinese memory fac- me- memory manufacturer CXMT IPO'd in Shanghai this morning, China time.

250
00:47:34.408 --> 00:47:46.368
Uh, company's named ChangXin Memory Technologies. Apologies if I technically mispronounced that. And overnight, it became the largest publicly traded company in mainland China.

251
00:47:46.508 --> 00:47:54.588
Now, this, this does not nearly approach the scale of publicly traded companies in the United States, the pr- best place for capital markets.

252
00:47:55.068 --> 00:48:11.908
But in China, this quickly, overnight, jumped ahead of China's largest bank to be roughly half a trillion dollar market cap, specifically 460 billion. The company, CXMT, IPO'd,

253
00:48:12.888 --> 00:48:28.488
uh, after raising, uh, $8.5 billion, the largest ship IPO in Chinese history, and then went up 466%. Wait, what? What? Yeah.

254
00:48:28.608 --> 00:48:43.948
So talk about, yeah, it... Remember when SpaceX went up, you know, 50- Could you imagine being the guy who was pre-IPO and sold at the bell? Yeah. Yeah. So wild stuff here. And, and why? You know, what it...

255
00:48:43.988 --> 00:48:53.448
You know, you probably, even if you started getting into memory stocks recently, you might not have heard of CXMT [laughs] because this stuff happens so quickly.

256
00:48:53.458 --> 00:49:06.140
They are a Chinese, uh, memory manufacturer, uh, that now plays an important role alongside Samsung, SK Hynix, et cetera. Um, they make DRAM.

257
00:49:06.540 --> 00:49:19.520
So a little bit different than the high bandwidth memory, HBM, that is in such demand for AI compute. Um, DRAM, DRAM is basically lower on the stack.

258
00:49:19.700 --> 00:49:33.850
It's a less, uh, high throughput type of memory, and it typically goes in, like, chips for phones, computers, like gaming computers, DDR4, DDR5, if you're a gaming computer manufacturer.

259
00:49:33.960 --> 00:49:47.620
And CXMT is the world's fourth-largest DRAM manufacturer after Samsung, SK Hynix, and Micron. They had roughly 7.7 of the global DRAM market in 2025 per the IPO prospectus.

260
00:49:49.640 --> 00:50:00.730
One interesting thing here, which is CXMT does not make high bandwidth memory. They make the, uh, basic- the component to high bandwidth memory, which is DRAM. Uh,

261
00:50:01.820 --> 00:50:12.940
HBM or high bandwidth memory basically takes DRAM and stacks it all on top each other, interlaces it, and so that is what makes, that's, that's what makes it high bandwidth. Um,

262
00:50:13.920 --> 00:50:41.280
and, uh, one of the reasons here is because HBM is so, uh, constrained right now, the market and the attention and, like, the, the bottlenecks are now moving lower down the stack, not to the end product that you might almost consider, like, more refined memory, to the lower, like, base unit commodity of memory, which is [laughs] DRAM, which goes into high bandwidth memory, memory.

263
00:50:41.340 --> 00:50:52.600
And that's why this Chinese company kind of out of nowhere is so important. Why can't they make high bandwidth memory? Largely because they're locked out because of Chinese export controls.

264
00:50:52.680 --> 00:51:00.880
You need those fancy machines from, you know, the Netherlands. You need those, uh...

265
00:51:00.940 --> 00:51:09.300
I, I forget what they're called, lithography machines, uh, that you can't get, and, and there's also certain, like, IP and, uh, Chinese export controls.

266
00:51:09.360 --> 00:51:16.240
But I'm gonna toss it to you, Colin, for any comments you have before I get into a few different takes and insights.

267
00:51:17.080 --> 00:51:34.440
I don't really have too much to say on this other than it just seems like with each week, China chips away a little bit more at the West's and the non-Chinese East Asian countries' competitive advantage with regards to AI and adjacent industries.

268
00:51:34.960 --> 00:51:47.360
You had DeepSeek, you know, a year and a half ago, Kimbie K-3 two weeks ago. Now you have this, and it seems like there's almost this deliberate rollout of all of the

269
00:51:48.980 --> 00:51:59.160
key components for building an AI industry stack that are, like, slowly creeping out of China and coming out of the shadows. Yeah.

270
00:51:59.680 --> 00:52:07.820
So, uh, an interesting thing here is that CXMT is apparently adding more wafers than everyone else combined.

271
00:52:07.940 --> 00:52:21.480
That is, CXMT will add 85,000 wafer starts per month this year versus 60K at SK Hynix, 30K at Micron, and 15K at Samsung as per semi-analysis.

272
00:52:21.920 --> 00:52:33.760
So that means that the Chinese upstart is adding more DRAM capacity than the three incumbents combined, and that does not, uh, ameliorate the memory crunch.

273
00:52:34.360 --> 00:52:49.230
Additionally, as I mentioned before, because they have to manufacture the lowly, lowly, uh, DRAM com- and, and can't produce high bandwidth memory, um, the reason that they're in this position is be- is because, uh...

274
00:52:49.460 --> 00:52:59.880
Rather, I'm saying they're actually benefiting from this, this scenario of being locked out because the, all the other memory manufacturers basically rotated their en- their entire operations

275
00:53:01.020 --> 00:53:12.160
to focus on, like, building this high bandwidth memory, and because of that, did not expand their lower level, um, memory manufacturing, whereas CXMT has only been ramping that up.

276
00:53:12.240 --> 00:53:22.380
So it's in, like, in a way, they have a precursor to the more final product which is ultimately in demand. So pretty interesting here.

277
00:53:22.520 --> 00:53:51.720
Uh, I think if you are a normal person who doesn't, uh, who just wants s- consumer, end consumer products, um, uh, the new phone, the new smartphone, the, a new gaming computer, probably any piece of technology, uh, you are just going to get caught up in this because DRAM is now, uh, DRAM, which would go into your other normal non-AI HPC tech, is now getting crunched even though supply is ramping up like crazy.

278
00:53:51.780 --> 00:54:08.800
So we have a structural shortage of memory for a while now, and it's being reflected in the markets. It's reflected in this, uh, supposedly no-name Chinese company becoming the most valuable public company in China.

279
00:54:08.940 --> 00:54:23.660
Uh, wild story. Uh, yeah, that's all I got for this one. I think we can wrap it up there. We're gonna wrap it up here. Thank you so much for listening and tuning in.

280
00:54:23.780 --> 00:54:35.880
Block Space goes live Monday, Tuesday, Wednesday, Thursday, Friday at 1:00 PM Eastern, featuring quick hits in AI data centers and other compute-related markets.

281
00:54:36.000 --> 00:54:50.240
If you like what you hear, you can see everything we produce at our website, blockspace.media. This show is brought to you by CleanSpark, NASDAQ-listed ticker CLSK. Thanks for tuning in. I'm Charlie. I'm Colin.

282
00:54:50.720 --> 00:54:58.180
And we'll see you tomorrow. [outro music]