Soluna Q2 Earnings and CFO Interview, NBIS Hits a Snag in NJ, What it Takes to Build Compute Futures

WEBVTT
1
00:00:00.100 --> 00:00:11.430
On y'all. Welcome back to Blockspace Live. We are coming at y'all fresh and early for me, but not for the rest of the world who are on civilized time zones, on this Friday.
2
00:00:11.860 --> 00:00:23.200
We have got our final earnings week update today, with a note on Soluna's Q2, plus an interview with Soluna's CFO, Mike Pitchie. Those will be our first two segments today.
3
00:00:23.240 --> 00:00:38.180
Following that, we have Luxer Hansen, CEO [laughs] Luxer Hansen, Nick Hansen, CEO of Luxer, on to discuss what else but AI and compute. And then we are going to be hopping into Hunter Brook's latest hunt.
4
00:00:38.810 --> 00:00:42.100
They are hell-bent on anything Bloom Energy.
5
00:00:42.140 --> 00:00:57.060
They released a report today about how Bloom Energy at one of its sites in New Jersey, in Vineland, its data center operator, DataOne, has been issued a stop order by the local government that for the Bloom Energy cells they talked about on their Q2 earnings call.
6
00:00:57.070 --> 00:01:08.080
Don't wanna miss that one. A lot of juicy details there. And then following that we have Brett Harrison, the CEO of Architect Financial Technologies, and that's the show.
7
00:01:08.700 --> 00:01:21.400
Big packed lineup for today, and, uh, a lot to dig into, Charlie. Yeah, banger episode. Three guests today. Blockspace goes live every weekday, 1:00 PM Eastern.
8
00:01:21.580 --> 00:01:35.000
We are compute's daily live show, featuring quick hits on AI, data centers, emerging technology, and markets. If you miss the live show, you can always catch us when it hits the podcasts.
9
00:01:35.460 --> 00:01:48.740
Later that day anywhere podcasts are streamed. And you can go to our website for more content, including a lot of written content. Uh, website is blockspace.media. That is blockspace.media.
10
00:01:48.820 --> 00:02:06.260
This show is brought to you by CleanSpark, NASDAQ listed ticker CLSK. More on CleanSpark later on in the show. But we have to do our Luxur hash rate index update.
11
00:02:06.300 --> 00:02:18.280
We're doing it a little bit differently this time, 'cause this time we are talking about compute and not hash rate so much. Colin, what do we got?
12
00:02:18.700 --> 00:02:25.780
Yeah, we were getting a little bit fatigued with saying that Bitcoin mining metrics were not doing anything week in and week out, so we're switching this up.
13
00:02:25.820 --> 00:02:30.080
Going forward, we're gonna start highlighting different data sets from hash rate index and also blog posts.
14
00:02:30.100 --> 00:02:37.060
They have a really good repository of resources, and they're starting to do more AI and com- AI and general compute content as well.
15
00:02:37.120 --> 00:02:46.980
So for today we are highlighting an article by Michael San Miguel, friend of the show, been on a few times now, about Vera Rubin, what I'm telling operators before they buy.
16
00:02:47.470 --> 00:02:57.820
And Michael's got some really good insights here, and the TLDR of this blog post, he's trying to answer the question of like, when does it make sense to actually deploy one of these systems?
17
00:02:58.340 --> 00:03:12.040
You know, this is the, uh, the diamond in the, in N- Nvidia's portfolio now. It's a rack design with a CPU and GPU together, and the TLDR from Michael is the following.
18
00:03:12.049 --> 00:03:19.920
"Nvidia won't sell you a Vera Rubin," quoting directly from the article, "rack until it approves the site you run it in. The facility is the gate, not the chip.
19
00:03:19.980 --> 00:03:31.400
Whatever you buy will run at close to 100% utilization, so the expensive mistake isn't overbuying, it's waiting. For a lot of inference, the RTX Pro 6000 does the job. You don't need a Vera Rubin rack for most of it."
20
00:03:31.460 --> 00:03:42.920
I wanna highlight that last point, and it's kind of central to Mike's thesis here. [sniffs] Um, he, he gets into what he would actually buy depending on a data center's needs.
21
00:03:42.930 --> 00:03:50.520
Seen here, quote, "You've seen the headl- " Uh, uh, quoting directly here, "You've seen the headlines, up to 10% the inference throughput per watt versus Blackwell.
22
00:03:50.530 --> 00:04:01.790
That's real, but it concentrates in specific regimes, large mixture of experts, long context, decode-heavy serving, and leans on the Groq integration, not the Rubin GPU alone. It's not universal cut to your bill.
23
00:04:02.400 --> 00:04:10.140
So here's how I think about it. Vera Rubin earns its place when you're running latency-sensitive, decode-heavy inference at large scale, and you can clear the facility bar.
24
00:04:10.580 --> 00:04:21.420
But a lot of inference isn't that, and it doesn't need a facility gated rack at all. A current generation RTX Pro 6000 Blackwell gives you basically what you need depending on your cost profile."
25
00:04:21.930 --> 00:04:34.520
Another interesting thing I thought about this, Charlie, and I'll toss it to you for second takes, is this idea that we've gotten to the point that for these higher end Nvidia systems, Nvidia wants to see that you actually have the ability to run it on site.
26
00:04:34.530 --> 00:04:49.360
They're going to need to see receipts for your infrastructure to make sure that you're not just placing an order for something that ultimately won't be beneficial at all. Uh, you're muted, Charlie. Ah.
27
00:04:49.380 --> 00:04:56.020
Well, see, my comment is, um I'm gonna throw your screen back up. Scroll down- scroll down to, "The expensive mistake isn't overbuying."
28
00:04:56.640 --> 00:05:12.690
So this is an interesting, uh, take because, uh, if we go back and take the lessons that we all learned from Bitcoin mining, everybody was eager to get their hands during the peak of the 2021 run on, like, nine-month, 12-month delivery, uh, new machines.
29
00:05:13.060 --> 00:05:27.000
But this is the opposite take with, um, of that lesson that the industry learned about GPUs. Specifically, and this comes into, like, the, the, like, one of the themes of this earnings week, is that,
30
00:05:27.980 --> 00:05:42.720
uh, while token prices are dropping, uh, a useful lifespan of these rigs is much longer than GPU bears and depreciators have anticipated.
31
00:05:42.780 --> 00:05:58.520
So, like, the three to five-year depreciation timeline, uh, is, uh, vastly under, uh, uh, is not i- i- as we've seen, now that we're six years into the A100, uh, not totally accurate.
32
00:05:58.600 --> 00:06:15.220
So basically, Michael says- Buy your GPUs now. Uh, that's a bigger risk than maybe buying the not perfect optimal one for your specific rig. So, uh, we could,
33
00:06:16.540 --> 00:06:26.620
uh... Yeah, so we're, we're having Luxur CEO, Nick Hanson, on here in a- about 20, 25 minutes, uh, to...
34
00:06:26.660 --> 00:06:33.199
And we'll talk about this, but mainly about, um, the landscape of, uh, Southwest Power Pool and the new Luxur energy product.
35
00:06:33.520 --> 00:06:47.780
Before that, though, we have in the wings Mike, CFO of Soluna, who we'll bring on up here to riff on their latest earnings right after a word from our sponsor, CleanSpark. [gentle music] We are CleanSpark,
36
00:06:48.700 --> 00:07:01.160
America's Bitcoin miner, a publicly traded company with the largest operating hash rate. Powered entirely by self-operated infrastructure across four states.
37
00:07:03.020 --> 00:07:20.620
This is our proof of work, and we are setting the standard for what's next. Learn more about the intersection of energy and Bitcoin at cleanspark.com. All right, we've got Soluna CFO, Mike Pichi, in the wings.
38
00:07:20.720 --> 00:07:28.600
We're kind of accelerating the timeline here. We were gonna cover some of the earnings before. We'll do that after Mike gets off 'cause he is a early bird gets the worm.
39
00:07:28.640 --> 00:07:34.480
Yeah, he showed up- We appreciate a guest showing up on time... showed up promptly. We got it. Yeah, let's bring him on up here. Hopefully. We haven't done a mic check, so we're just gonna bring him on. Let's do this.
40
00:07:35.140 --> 00:07:43.280
Mike, hello. I think we're okay. Good. How are you? Great. And is it Mike Pitchy or Mi- or how do I say your last name? Pichi. Pichi. [laughs] And you gotta raise your hand.
41
00:07:44.000 --> 00:07:52.780
Yeah, Pichi, and he's got the peach-colored polo, too, this man. [laughs] I'm on brand. So Mike, thank you. Appreciate you joining. And for...
42
00:07:53.500 --> 00:08:00.540
To, to kick off our conversation, I wanna start with some of the consolidations that we saw from Soluna, um, over the first half of this year.
43
00:08:00.620 --> 00:08:12.540
So we've got, um, Dorothy One fully consolidated in terms of equity interest, as I understand it. You also, um, purchased the Briscoe Wind Farm, 150 megawatts in, uh, in April.
44
00:08:13.460 --> 00:08:25.780
And this all leads towards a, uh, you know, Soluna furthering its stake in these sites that formerly it was either, you know, renting the power, um, on the Briscoe site. It's also where y- or some of the operations sit.
45
00:08:26.340 --> 00:08:37.400
And also consolidating that equity interest. Can you walk us through what actually changes oper- operationally at these sites, and what can you do now that you couldn't do six months ago with this consolidation?
46
00:08:38.040 --> 00:08:50.080
Yeah, sure. Thank you for having me, uh, today. Operationally, nothing changes. We were already operating the sites as if we were the majority owner. Uh, in one case, we did have more than 50%.
47
00:08:50.600 --> 00:09:03.250
But now that we have 100% control at bo- both Dorothy One A and One B, um, that 50 megawatts of Bitcoin hosting capacity, um, can continue to work.
48
00:09:03.320 --> 00:09:15.060
But it helps us in the future in terms of thinking about, uh, an AI campus just across the street from Dorothy One and Dorothy Two. So our new AI campus is called Dorothy Three,
49
00:09:15.980 --> 00:09:25.860
and all three campuses are essentially built in the shadow of the Briscoe Wind Farm that we bought, and that's a 150-megawatt nameplate, uh, wind farm.
50
00:09:26.060 --> 00:09:32.760
And so now we can have more flexibility that we own 100% in terms of how we allocate power at the site.
51
00:09:33.250 --> 00:09:50.700
And this has been really important with Governor Abbott's audit requirements because already energized power, uh, essentially doesn't need to go through the audit process, uh, and has a lot more value, um, than power that's lined up in a queue to get approved down the road.
52
00:09:50.760 --> 00:10:03.880
So, uh, we're pretty proud of, uh, being ahead of the curve and buying the generation asset that, as you mentioned, for years we had been using as a customer, and now we are our own customer.
53
00:10:04.120 --> 00:10:11.420
Uh, we own the generation, uh, and the, uh, uh, compute, so it's wonderful. So I'll jump in here.
54
00:10:11.580 --> 00:10:30.060
So, um, on the back of that, looking out towards AI sites, uh, is a more consolidated model something that you want for those sites, or is, uh, this, uh, this consolidation of the mining, uh, stuff, is that unique to just that package?
55
00:10:30.520 --> 00:10:36.420
And, and I'll kind of throw in a add on there, too, in terms of the JV that y'all have for your most recent AI deal.
56
00:10:36.560 --> 00:10:43.260
So I'm just cur- curious as well about the structure, if, like, JV is something y'all privilege or if you want full ownership for some of these going forward.
57
00:10:44.120 --> 00:10:57.220
Yeah, I think, um, uh, on the first part, on Charlie's, uh, point, we would certainly be opportunistic if there were other wind farms available, uh, and we could pair them with building a data center campus in the shadow of those wind farms.
58
00:10:57.740 --> 00:11:11.120
Um, it's a great business model, um, in terms of being vertically integrated. So yes, I'd say m- moving forward, we'd be opportunistic, uh, at looking at that. With regard to, um, Coddy Two, uh, our
59
00:11:12.240 --> 00:11:20.470
project that's furthest along, we did announce in the, in the second quarter that we had completed our joint venture formation, and we partnered with a, a firm called MetroBlocks.
60
00:11:20.520 --> 00:11:30.240
They've built a number of AI-scale data centers around the country. Um, and they also had, um, expertise in terms of customer acquisition, design work, and the like.
61
00:11:30.300 --> 00:11:41.150
And so we thought for this first big, uh, AI campus and initial phase one would be, uh, roughly 100 megawatts, it made sense to partner, uh, with MetroBlocks.
62
00:11:41.200 --> 00:11:51.740
And, and we'll share economics down the road, but we shared some economics... Uh, we shared some details on those economics in our, um, AK filing when we completed the joint venture.
63
00:11:52.230 --> 00:12:03.220
And Soluna, w- we're the fun- uh, founding and funding partner. Um, but we'll get a return on our equity-invested capital. We'll get a return of that capital.
64
00:12:03.230 --> 00:12:13.812
We'll get a return on joint venture costs that we're funding to get the project launched. Get a 14% IRR on those costs, and then we also get credit for the land that we're contributing into the project.
65
00:12:13.872 --> 00:12:28.832
So, um, it, it, it was a win-win. We, we, we de-risked the launch of that project. Um, and as you see in, uh, in our, uh, most recent, uh, filings, we are under a letter of intent for a customer at that site.
66
00:12:30.752 --> 00:12:35.392
Mike, I wanna turn our attention really quickly to financing for some of these builds.
67
00:12:35.612 --> 00:12:50.532
Uh, 'cause the funding sources for s- the f- uh, first six months of the year, um, were mostly from at the market sales, I believe 113.5 million from the ATM, um, 74.2 million shares issued in Q2.
68
00:12:50.592 --> 00:13:02.362
Should we expect more equity financing for AI builds as y'all move further into this sector? Or will you look for project-level financing, other forms of debt? What's your strategy with regards- Yeah...
69
00:13:02.372 --> 00:13:14.092
to financing these builds? Definitely, uh, more project-level financing. And so that $113 million that we've raised in the ATM in Q2, and we've used it to further these projects along.
70
00:13:14.472 --> 00:13:30.002
But once it flips into a signed lease, now you've got to go raise 1.2 to $1.3 billion related to the build-out. And that bigger nut, um, we will fund w- at the project level.
71
00:13:30.142 --> 00:13:39.472
We estimate probably 80% debt, mostly likely in the firm of high yield notes. In fact, we've already hired an investment bank to do that, uh, to start working on us with that.
72
00:13:39.962 --> 00:13:49.832
And that's a repeatable transaction that's been done across the sector by many, many companies, whether it be Applied Digital, uh, TerraWolfe, Cipher.
73
00:13:50.432 --> 00:14:03.872
Um, and so there's a, a great market for those bonds, uh, to fund 80% of the build-out, and then the remaining 20%, um, will be funded by raising additional Soluna equity.
74
00:14:03.992 --> 00:14:16.772
Um, but yeah, there's no-- it's not at all contemplated that we're gonna use equity, equity to fund 100% of the roughly 1.2, $1.3 bill- billion dollar build-out. Yeah. So it's a lot of money to throw around.
75
00:14:16.872 --> 00:14:30.512
I think that when investors look at, um, some of the equity issuance of, of any company, and they're asking about that for the AI forward stuff, um, if you actually look at the numbers, it doesn't make any sense 'cause you can't really reliably do that with equity, right?
76
00:14:30.572 --> 00:14:41.552
So, like, dilution risk oftentimes overstated, I think, from some of the more jittery investors with regards to that kind of stuff. Uh, I, I would love now to talk about Briscoe specifically.
77
00:14:41.692 --> 00:14:51.672
I, I'm, I'm, I'm just trying to make sense of, like, what exactly the offtake on this is and how much of that can be localized to Dorothy.
78
00:14:51.772 --> 00:14:58.672
Um, specifically, like, note two in the Q says Briscoe sells power to Golden Spread Electric Cooperative under a PPA.
79
00:14:58.692 --> 00:15:04.412
But then it also says all PPA revenue is eliminated in consolidation because all sales are with Dorothy entities.
80
00:15:05.052 --> 00:15:19.232
So is Golden Spread Electric Cooperative still a third party offtake or post-acquisition, or is that offtake now serving Dorothy load? Uh, a little of both. So Dorothy one in aggregate is fifty megawatts.
81
00:15:19.352 --> 00:15:28.532
Dorothy two is forty-eight megawatts, so just under a hundred megawatts deployed there, and the wind farm's 150 megawatts nameplate capacity.
82
00:15:28.552 --> 00:15:42.632
So at capacity, the balance would be sold, um, essentially in merchant in the open market. But you could tell from the elimination, um, on a gross basis in Q2, the wind farm had $2.5 million of revenue.
83
00:15:43.272 --> 00:15:57.442
Call it that third-party merchant revenue was three hundred and sixty-six thousand of the 2.5 million. So we were our own biggest customer, uh, by, uh, you know what, eighty-five percent roughly in the quarter.
84
00:15:58.172 --> 00:16:09.012
Now, that'll change as you move through quarters. There'll be more merchant activity, we would think, in the later part of the year. But, um, that's where it was in that first quarter of us owning the Briscoe Wind Farm.
85
00:16:10.352 --> 00:16:26.532
So j- just to be clear, um, like are-- is there any contractual obligation to deliver a specific amount of megawatts to the, uh, to the company, or is-- does, does Soluna have flexibility with how much they wanna sell?
86
00:16:26.592 --> 00:16:32.292
We have flexibility with how much we wanna sell. Okay. Um, and- Yeah... and, okay, that, that, that makes a lot of sense.
87
00:16:32.332 --> 00:16:49.672
And this might be a redundant question, um, but y- just, uh, kind of drawing this out, uh, Briscoe reported, I think you just said this, roughly three hundred and sixty-six thousand of revenue against 2.25 million of cash cost and one million in depreciation and accretion in its first quarter of ownership.
88
00:16:50.492 --> 00:16:57.932
Um, and that's against a projection of twenty to twenty-four point four million of annualized revenue and six to eleven million of year one adjusted EBITDA.
89
00:16:58.532 --> 00:17:08.552
Um, if the offsetting benefit is a landing and cost of data hosting revenue rather than in the wind line, how much did the intercompany elimination reduce that cost line in Q2?
90
00:17:08.632 --> 00:17:19.312
And, like, does Briscoe still hit that year one range, do you think? Yeah. So in terms of, uh, uh, thinking about the elimination and the cost, the energy costs are passed through to the end customer.
91
00:17:19.461 --> 00:17:27.452
So, uh, there is a, there is a markup there. And with regard to, uh, th- the range, you know, we're, we're early days in this.
92
00:17:27.552 --> 00:17:34.812
Uh, that original projection certainly was pr- priced for perfection in terms of wind conditions and uptime.
93
00:17:34.992 --> 00:17:54.732
You know, when we, uh, actually inherited the wind farm, ten out of the eighty-one turbines were down, so roughly 12% of the capacity was down, and that's part of the $1.5 million of repair costs we incurred in the quarter, uh, as part of that $2.2 million of operating expense at the wind farm level.
94
00:17:54.832 --> 00:18:07.912
Um, we certainly would expect that to come down. I think in terms of, uh, the summer is the, you know, Q2 is the shoulder season. The wind doesn't blow as hard. We'd like to see a couple more quarters how this plays out.
95
00:18:08.012 --> 00:18:18.628
Um, we've- Uh, did the repairs, got those 10 turbines back online here by, uh, early third quarter, and the wind will pick up here in, uh, in the fall and winter.
96
00:18:18.688 --> 00:18:31.418
So, um, think we'll, we'll just wait and see how the next couple of quarters play out. Um, but the real value in that asset isn't in terms of being a supplier of power for the Bitcoin hosting operations.
97
00:18:31.928 --> 00:18:47.408
It's the value of that asset when it's supplying an AI campus in late 2027, early 2028. Uh, Mike, I've got one last question, and, um, if Charlie has anything as a closer, I'll throw it to him afterwards.
98
00:18:47.808 --> 00:19:02.208
But it, it's specifically related to, um, a 19.3 million, I, I believe services liability on, um, Soluna's books, and it's related to this GPU purchase that dates back to June 2024.
99
00:19:02.238 --> 00:19:15.388
Hewlett Packard terminated its agreement with Soluna in March 2025, and y'all's subsidiary, Cloud Co, hasn't made any payments on this in 17 months. I'm just curious what the status on that 19.3 million is.
100
00:19:15.468 --> 00:19:22.828
Is it disputed, in negotiation, or litigation? I'm just curious why it hasn't been addressed yet. Yeah. Essentially, I'd say it's dormant.
101
00:19:23.008 --> 00:19:36.808
Um, both parties agreed to kind of just, uh, go their own ways for now, um, and, and see how the businesses, uh, business develops. So there's been no activity on that, um, in almost a year, I think, so.
102
00:19:38.868 --> 00:19:50.248
I'll, uh, I'll, I'll wrap it up. First of all, you have a new, uh, cheerleader in your court. You got, uh, Mike with Alpine Fox disclosing that, uh, they, let's see, trimmed Cipher and added Soluna.
103
00:19:50.408 --> 00:19:59.348
So, uh, new cheerleader in your court. He's a fun cheerleader to have. Um, I guess zooming out, I know, I know you're more f- finance focused, but,
104
00:20:00.248 --> 00:20:13.748
uh, look, you know, looking at what the Governor Abbott in Texas, uh, has said, hold off on the interconnection queue. Can you speak to this broadly? What do you think's going on here? How do you think the next, um...
105
00:20:13.788 --> 00:20:24.068
How do you guys think about the next six months as the interconnection queue and people are, and, uh, and people are vying for power and, like, desperate to get these deals unlocked?
106
00:20:24.108 --> 00:20:29.748
Can you speak broadly just to the Texas market here? Yeah. So, um, we're thrilled to be in the Texas market.
107
00:20:29.788 --> 00:20:43.148
We've got multiple project sites across the state, um, and, uh, it's a great opportunity for us with, as a company focused on renewables, primarily, uh, wind power, uh, building data centers in the shadow of wind farms.
108
00:20:43.168 --> 00:20:53.898
We're in the right, right state. Um, Governor Abbott's, uh, mandate, uh, we think helps the industry, helps Soluna, one, because we already have energized power.
109
00:20:54.058 --> 00:21:05.018
So that's really valuable to customers that are looking at, "How am I gonna be powered up by the end of '27 or early 2028, uh, if I wanna go build a, a new data center?"
110
00:21:05.068 --> 00:21:15.168
Well, at least Soluna already has power at both of those AI campuses, um, available and, and energized today and not subject to, to the review period.
111
00:21:15.868 --> 00:21:26.688
Um, and then the other thing, you know, we, as John Belizaire, our, our co-founder and CEO said in his fourth point on this, uh, that we welcome the review. We do.
112
00:21:26.788 --> 00:21:34.228
We, we think it'll call out, um, a number of maybe not so serious, um, projects that are in the, that are in that pipeline.
113
00:21:34.268 --> 00:21:46.468
I think it was over 450, uh, thousand megawatts, uh, that the, in, in the state of Texas, and so this will cull out some of that, uh, as well. Um, but net-net, we think, uh,
114
00:21:47.468 --> 00:22:00.268
Soluna's, one, already aligned with what the mandates want in terms of, of behind the meter, uh, paying for our own energy infrastructure, not impacting rate payers with what we're doing on the data center side.
115
00:22:00.908 --> 00:22:14.468
Um, and two, that we already have that energized power at our first locations. Mike- Pici... Pici. [laughs] Thank you for joining, man, and, um- Thank you... be- best of luck for the rest of the year.
116
00:22:14.478 --> 00:22:25.848
We've got, um, uh, we've got, uh, the CEO, John, coming on, uh, next week, so we'll have more- Okay... up on Soluna for listeners who are interested. But we'll be checking in after Q3 for earnings then too.
117
00:22:25.988 --> 00:22:32.888
Love to have you and/or John back on to talk about the results. Thank you for having me. I look forward to coming back. Thank you, Colin and Charlie. Thank you, Mike. Thanks, Mike.
118
00:22:35.568 --> 00:22:44.388
All right, Charlie, I'm gonna call an audible here. Call an audible. I think we should, I think we should do the Hunter Brook Nebia story next before- Yeah... we have Luxor on. I'm down.
119
00:22:44.428 --> 00:22:56.428
And, uh, then we'll, we'll keep rolling. Yeah. And this is a spicy one because Nebia's had their earnings, uh, blends together, was it yesterday, Thursday? Yeah, and this came out- Yeah... specifically this, uh, the
120
00:22:57.388 --> 00:23:08.088
approvals for nat gas use on the site, so. Okay. So, uh, we'll do that, but first... A word from our sponsor A word from our sponsor, Luxor. Luxor.
121
00:23:08.488 --> 00:23:19.288
[static] This episode is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations.
122
00:23:19.688 --> 00:23:25.368
Luxor's Commander gives you real-time fleet monitoring, bulk remote commands across your fleet, and Intelligent Miner.
123
00:23:25.428 --> 00:23:32.358
That's an automated profitability engine that runs every five minutes and tests your fleet's power settings against live energy and hash rate markets.
124
00:23:32.868 --> 00:23:38.708
In fact, ERCOT back tests showed 10% improved profitability with intelligent mining versus old-fashioned binary mining.
125
00:23:39.328 --> 00:23:47.588
Commander Pro is roughly half the cost of competition, $100 per megawatt, or a 25 basis point pool fee adder, but you can also try it free for 60 days.
126
00:23:47.648 --> 00:23:58.448
So if you'd like to learn more, go to luxor.tech/commander to get started. A quick note on the docket today. We are gonna go ahead and skip over to this story from Hunter Brook.
127
00:23:58.988 --> 00:24:06.168
We will be covering Soluna's earnings after the Luxor interview. Got a lot of qualitative and some quantitative stuff to put you through there.
128
00:24:06.228 --> 00:24:16.178
Just don't have time for it in the s- uh, short amount of time we have between now and when Nick Hansen, and I believe Jamie from Luxor joins us as well. So- I, I, I actually think it's, uh, just Nick.
129
00:24:16.308 --> 00:24:28.764
Called another audible. We got a few audibles in the air. Lot of audibles today. But this is a new- Report from Hunter Brook. Disclosure out front, they are short Bloom Energy, and they're involved in this story.
130
00:24:28.804 --> 00:24:37.844
And we covered their short outlook on Bloom Energy on the pod, I believe a month or two ago. But this is coming from their media team.
131
00:24:37.884 --> 00:24:52.084
Tweet here, headline reads, "Vine City officials issued a stop construction order on August 10th at the Neebus Data One data center in Vineland, New Jersey, citing installation of Bloom Energy fuel cells with, quote, 'no prior approvals, plans, or permits,' end quote."
132
00:24:52.704 --> 00:25:05.284
They have a link to the article here on Hunter Brook's X account, and they break down the TLDR within the tweet thread. So a few things to note here with regards to what's going on.
133
00:25:05.324 --> 00:25:13.304
Data One is the data center operator, Neebus is the tenant, and the offtake for that compute is none other than Microsoft.
134
00:25:15.224 --> 00:25:21.944
And this is interesting to us with regards to what it says about the approval process for this site.
135
00:25:22.884 --> 00:25:34.604
Neebus fielded questions about not having approval for nat gas generation on the site during their Q2 earnings call, and they said that they are using Bloom Energy as a kind of stopgap for this.
136
00:25:34.624 --> 00:25:43.164
Their fuel cells will then take that nat gas and convert it to energy without having to have generators on set. The core facts of the case are as such.
137
00:25:43.244 --> 00:25:53.064
The two stop construction orders in, have come in one week from the city of Vineland, both issued to the site owner, Data One Vineland LLC. At the data center, Data One is building for Neebus.
138
00:25:53.164 --> 00:26:02.044
The first one, August 6, halts installation of an en- a liquefied natural gas tank, citing no appro- prior approvals or permits.
139
00:26:02.084 --> 00:26:09.444
Then an August 10 order halts work on the Bloom Energy fuel cell units that they are deploying, citing no prior approvals, plans, or permits for those either.
140
00:26:10.204 --> 00:26:16.494
The Bloom units are the power source for the roughly 300-megawatt site. The capacity, again, is for Microsoft. They're contracted to use it.
141
00:26:17.384 --> 00:26:29.384
Um, Hunter Brook s- uh, sourcing links directly to the SEC filed exhibit for that, so the Microsoft offtake is primary sourced. It's not speculation. Now, why this permitting sequencing matters.
142
00:26:29.424 --> 00:26:48.444
The amended site plan, which will add nine Bloom power areas, a 8,660 square foot LNG tank, that's huge, and an LNG filling station, went before a roughly six-hour special planning board hearing that adjourned without a vote amid resident opposition.
143
00:26:49.124 --> 00:26:55.933
Chalk one up again to the rabid anti-data center hordes for stopping a site's permitting in its tracks.
144
00:26:56.564 --> 00:27:08.064
A continuation was scheduled for the following Monday, with public comment again on the agenda and no guarantee of a vote. So the equipment was being put in ahead of the approval for the plan that authorized it.
145
00:27:08.144 --> 00:27:20.124
Hunter Brook first flagged the installation before approval on August 5th, and points to a staffing agency holding a, a July 31st hiring event in Vineland for Bloom assembly technicians with August start dates.
146
00:27:20.204 --> 00:27:26.564
Uh, circumstantial evidence, but it's consistent with what we're looking at here. Now, there is a disclosure story here.
147
00:27:26.684 --> 00:27:32.704
Neebus held its Q2 call this Wednesday, we covered it here on the show, two days after the second order hit.
148
00:27:33.044 --> 00:27:45.904
When they were asked about Vineland, the chief communications officer, Tom Blackwell, said, quote, "Overall, we are on track with our delivery," end quote, and that the company was confident the layout complies with all applicable local, state, and federal laws.
149
00:27:46.374 --> 00:27:54.894
The chief infrastructure and product officer said Bloom fuel cells deploy- deployment, quote, "Should be fast," end quote, with no significant impact expected on the project timeline.
150
00:27:55.794 --> 00:28:02.324
Now, neither executive mentioned the stop orders. The questions they were answering, though, was about last week's planning board meeting.
151
00:28:02.424 --> 00:28:16.484
So they didn't duck the topic, uh, but they didn't address what's being talked about here. Now, here's the thing. They're technically correct. The site is on time, and it is, has been built, and they can...
152
00:28:17.884 --> 00:28:26.064
Under New Jersey's uniform construction code, um, work can't continue in violation of a stop order except to abate a dangerous condition or by a court order.
153
00:28:26.404 --> 00:28:35.244
But each day of non-compliance is a separate offense at up to $2,000 per violation and no, um, certificate of occupancy issues until penalties are paid.
154
00:28:35.754 --> 00:28:38.744
Vineland could seek an injunction if they decide to eat that penalty.
155
00:28:38.784 --> 00:28:48.384
But basically, as far as I can understand, they, Data One and maybe Neebus, I assume Data One would probably have the liability here, maybe split it with Neebus later.
156
00:28:48.424 --> 00:29:00.464
They can eat that $2,000 a day fine and continue to do what they're doing if they wanna be brazen with it. Um, but the per day penalty, um, it, it, it's pretty trivial for a 300-megawatt project, honestly.
157
00:29:00.504 --> 00:29:04.654
That's not really that much money given the scope of the CapEx going into this.
158
00:29:04.704 --> 00:29:14.804
Um, but the certificate of occupancy is not, and that's kind of the gating item that will be a, uh, that will, that is the item that will be gating delivery capacity for Microsoft.
159
00:29:15.924 --> 00:29:18.104
Last few things here, Charlie, then I'll toss it to you.
160
00:29:18.944 --> 00:29:29.154
A Data One spokesperson told Hunter Brook the notices were, quote, "limited to equipment placement work under the approval site plan," end quote, and that they did not expect the pause to affect construction and delivery timeline.
161
00:29:29.524 --> 00:29:37.724
Again, that's important if you're looking at the Q2 call. Neebus is not wrong to say that they, that nothing within the construction is actually being delayed here for the data center itself.
162
00:29:38.484 --> 00:29:47.704
Um, and, and that's kind of meaningful pushback here. It asserts that the work fell within an already approved plan. Vineland's building code office said it was instructed not to discuss the issue.
163
00:29:48.064 --> 00:30:00.284
Neebus and Bloom also did not respond. Again, important caveat here. Hunter Brook is short Bloom Energy, so they are, they have an interest here in making Bloom look as bad as possible.
164
00:30:00.784 --> 00:30:08.284
That being said, from all I can tell, those stop orders are legitimate, and DA Davidson, potentially as a result of this, cut its nee...
165
00:30:08.364 --> 00:30:18.804
or definitely as a result of this, cut its Neebus price target 30% on the delays at Vineland, New Jersey, threatening 2026 guidance. So I don't know how this gets resolved.
166
00:30:18.924 --> 00:30:25.456
It seems like they're gonna have that, uh, follow-up on the permitting. They're gonna have another v- uh- They're gonna have another meeting on that.
167
00:30:26.136 --> 00:30:34.106
If it is adjourned without a vote, then this can continue to stretch on. If it, the, it's rejected, that throws a huge wrench into this plan, but there's still a chance that this gets approved.
168
00:30:34.656 --> 00:30:39.206
Maybe they just eat these fines until that vote happens. I'm not sure. Uh, they wouldn't be the first.
169
00:30:39.236 --> 00:31:02.216
I mean, uh, if I'm not mistaken, Elon Musk with their gas turbines at, uh, in Memphis out at Colossus just roll those things onto the property, hooked 'em up, and is gonna ask questions, ask permission later, p- uh, and probably eat some fines because it's like if you know you can just pay a little bit of extra money and get it online sooner,
170
00:31:03.456 --> 00:31:13.096
who am- who, you know, what's the difference? Uh, there are still bottlenecks for permitting, and actually this gets into, I brought up Elon again, um,
171
00:31:14.196 --> 00:31:23.156
you know, just a moment ago Elon said, uh, by 2029, Earth will hit bottlenecks in both power availability and land permitting.
172
00:31:23.256 --> 00:31:28.986
You know, when people talk about bottlenecks, they talk about, you know, memory, generation, energy, transmission.
173
00:31:29.676 --> 00:31:40.876
Very few people are talking about permitting, but it seems to me that everyone is gonna be talking about permitting. Um, unclear how that progresses. So, um, enough about Nebius.
174
00:31:41.336 --> 00:31:58.056
We have in the wings Luxor CEO Nick Hashbender Hansen, and we're just gonna bring him on up here. Let's do this. Let's talk a little bit about compute. Nick, welcome to the show. Nick Hashbender Hansen. Let's go. Yeah.
175
00:31:58.156 --> 00:32:10.096
In the, in the intro I said, uh, Luxor Hansen. [laughs] So maybe you can start going by that. Luxor Hansen? Luxor Hansen. We u- we used to call, uh, Dan Lawrence, the CEO of- Yeah... Foreman. Yeah. Dan Foreman.
176
00:32:10.706 --> 00:32:21.156
Um- Dan Foreman. Uh, f- yeah, uh, deep Luxor lore there. Uh, came from a, a hiccup from a marketing hire at one point- Yep... who thought his last name was Foreman. Easy mistake to make.
177
00:32:21.656 --> 00:32:30.756
He said, "Congrats to Dan Foreman," or something like that. Dan Foreman for winning a miner. Yep. Yeah. Yep. Um, a- as ever- Yeah... a reminder to do your homework, but anyway. All right.
178
00:32:30.796 --> 00:32:40.696
Okay, so, um, we've been talking- What do you got for me today, guys? All right. Yeah, well, well, okay, so, uh, unfortunately we don't have the smart guy Jamie on, so we'll have to lob some easy, uh, softballs to you.
179
00:32:40.716 --> 00:32:47.746
But- Give me some meatballs. Yeah, I know. Well, uh, we- [laughs] You, uh, you guys are coming out [laughs] you guys are coming out with the Southwest Power Pool.
180
00:32:47.916 --> 00:32:55.476
Luxor Energy coming to my home state of Oklahoma, uh, and adj- and adjacent states to this RTO.
181
00:32:55.556 --> 00:33:05.456
Um, can you give me, like, a sec- a 60-second, like, high level of what Luxor Energy is and kinda what's going on coming to the SPP? Yeah.
182
00:33:05.616 --> 00:33:16.396
Luxor Energy is an energy provider by Bitcoin miners, for Bitcoin miners, and now hardware, uh, AI, AI deployments as well, AI data centers as well. So, um, you know, it's, we started in ERCOT.
183
00:33:16.536 --> 00:33:28.476
ERCOT's the easiest to get started in. Really what... You can kinda think of it as a, uh, uh, you know, uh, an energy provider that, uh, is denominated in Bitcoin. So we, we accept your Bitcoin hash rate.
184
00:33:28.876 --> 00:33:38.396
Uh, so we have this product called auto-settlement. Basically your pool, your pool rewards then just net out against your, uh, your power, uh, your power bill, um, and at the end of the month you get a power bill for $0.
185
00:33:38.936 --> 00:33:47.956
Um, which is, uh, which is great. That's been probably the biggest uptake, uh, of all. Uh, of course, we also, we also have, uh, Bitcoin collateral, uh, for those energy deposits, you know.
186
00:33:48.236 --> 00:33:59.496
Um, so basically we are a Bitcoin-denominated energy provider. Uh, and we're bringing that to SPP. So basically all that same, uh, you know, the same, uh, features and functionalities we're bringing to SPP, so.
187
00:34:00.076 --> 00:34:08.436
Y- you said that ERCOT is the easiest one to get started in, and that's kind of a foregone conclusion for anyone who knows m- anything about Bitcoin mining and energy specifically.
188
00:34:08.996 --> 00:34:16.246
What, what were some of the, uh, differences between getting involved in SPP? Like, what, was the hurdle higher? Yeah, so SPP just...
189
00:34:16.246 --> 00:34:27.116
Yeah, just for example, like you had to place $500,000 in credit with SPP just to start trading. Um, which is, um, you know, it's not, it's not, uh, it's not a, uh, it's not $0.
190
00:34:27.396 --> 00:34:35.536
It's not the most money you've ever heard, especially when, you know, Bitcoin miners are raising $9 or $10 billion every other day. Um, but it is, you know, it is a barrier.
191
00:34:35.836 --> 00:34:42.816
Uh, and also the, really the thing is for, for us is like the enrollment period is, is very, you know, it's quite long. It takes a long time to get enrolled.
192
00:34:43.176 --> 00:34:51.816
Um, you have to go through a lot of different hoops and hurdles to get registered. Um, and ERCOT just makes it quite a lot easier for us. So honestly, um, yeah.
193
00:34:52.216 --> 00:35:03.976
You know, we had, we've had a g- we've had a great run in ERCOT. Um, and I mean, I guess, you know, everybody had been building in ERCOT just assuming that, you know, Texas would just always say yes, uh, forever.
194
00:35:04.616 --> 00:35:15.216
Um, and then this summer they said, "Hold on. Uh, we're gonna slow that down for a little bit." Uh, which immediately kind of, you know, shook, I, I think shook some people and, and, and made SPP even more attractive.
195
00:35:15.296 --> 00:35:24.246
You know, you just have to go a few miles north, um, maybe more than a few, but, uh, go head north, uh, and you'll find, you'll find, uh, you'll find power.
196
00:35:25.176 --> 00:35:34.176
And, uh, we'll come back to, you know, comments on ERCOT and four CP, moving to 12 CP, but I wanna talk about SPP right now.
197
00:35:34.216 --> 00:35:46.836
And maybe you're not, like, primed and ready to talk specifically about SPP as much as Jamie would be, but, um, very heavy on wind. A lot of people would be surprised- Yeah... to know- The most wind in the country. Yeah.
198
00:35:46.916 --> 00:35:56.396
I mean, talk about that. What does that mean in, in context of demand response, Luxor Energy? Absolutely. Yeah. Yeah. Um, I would like to hope that I'm, uh, well prepared. [laughs] But, uh,
199
00:35:57.416 --> 00:36:03.536
the, the, uh, the, you know, it, it is, it is the most wind heavy, and wind, uh, causes lots of amplitude, uh, in, in the demand curve.
200
00:36:03.936 --> 00:36:15.456
Uh, what I mean by that is the peak demand and, um, trough demand, uh, basically you want your p- you want your generation curve to match that. And unfortunately, um, we don't control the wind.
201
00:36:15.736 --> 00:36:20.216
Um, they say there are people that can, but we can't. Um, but eventually, um,
202
00:36:21.436 --> 00:36:35.704
but really the problem is trying to get that, those curves to match, uh- And, and what, what, what ends up happening is you, you know, a lot of the load, a lot of the demand curve i- is, is not flexible, meaning I don't really care what's going on outside of my house.
203
00:36:35.764 --> 00:36:43.244
If it's 110 degrees outside, I'm keeping the AC on. The same thing is, is, is applies to, like, economic, uh, uh, loads as well.
204
00:36:43.284 --> 00:36:51.704
So, like, uh, somebody that's running a factory that's building cars, um, most likely the value of the cars is much higher than the value of the energy, regardless of what the cost of energy is.
205
00:36:51.744 --> 00:37:04.084
So they're going to keep running the factory. Um, Bitcoin is really the first very flexible load that exists, um, meaning it is purely economically driven. And so if the price is, is high, they will turn off.
206
00:37:04.364 --> 00:37:14.074
A- also, they, they, even if the price isn't high or there's some other incentive mechanism for them to turn off, uh, they, they, they will do so. Um, now, bring that, all of that all the way back to wind.
207
00:37:14.504 --> 00:37:26.224
Uh, when the wind is blowing, energy prices are low. Um, miners can spin up. You can use aftermarket firmware like LuxOS or whatever to, uh, to actually crank them even higher, consume more power.
208
00:37:27.114 --> 00:37:35.164
Um, you know, in Texas from time to time, power prices are even negative, or I should say ERCOT are even negative, meaning the more power you use, you're actually getting paid by the grid to do that.
209
00:37:35.224 --> 00:37:49.304
So, um, and then ener- and, you know, and Luxur Energy provides signals to miners to either ap- you know, accept, you know, or, sorry, to, uh, to ramp up, ramp down, to basically ride that curve based on what we're seeing in the market.
210
00:37:49.444 --> 00:37:58.164
Um, and bring it, again, bringing it back to wind, wind makes that very, very dynamic. So that curve changes really abruptly depending on what, when the wind is blowing.
211
00:38:00.024 --> 00:38:09.284
Given that and the intermittency of the wind and wind alley there, I mean, do you think there's any wiggle room for data centers in that area?
212
00:38:09.424 --> 00:38:15.374
I mean, is there anything else you've seen with an SPP that would indicate that they're adding enough base load to, uh, furnish this demand?
213
00:38:15.984 --> 00:38:26.184
Or are you mostly thinking that this is going to be a new mecca for Bitcoin miners, so to speak, in the US for those who can't get power in Texas or elsewhere? Oh, geez, mecca. Mecca. [laughs] Charlie- Hey...
214
00:38:26.204 --> 00:38:36.904
do you think it's gonna be a mecca? I was a big preacher of Oklahoma, but the problem is it's just why would you build a Bitcoin mine when you could build an AI mine? You know, I, but I, I think it, Oklahoma's great.
215
00:38:36.944 --> 00:38:53.904
Grid pricing is still incredibly cheap, so- Yeah... add, add demand response, yeah, people should build here. So I call- SPP feels like an underfished pond. It's like a place where people don't go there right now.
216
00:38:54.004 --> 00:39:03.304
Well, they, up until I guess, uh, the ERCOT, um, um, batch zero debacle, um, people weren't going to SPP as much.
217
00:39:03.404 --> 00:39:09.804
I mean, p- opportunistic folks, people that had maybe an angle or they already had some land or something like that were taking advantage of it.
218
00:39:09.884 --> 00:39:22.004
Um, but now it feels like a destination as opposed to an opportunity area or opportunity zone. Um, and so people are actually going and taking a, a hard, a harder look. So, um, I don't know if...
219
00:39:22.413 --> 00:39:26.164
I, I, it's, it's, it's harder to say exactly how much mining is on SPP grid.
220
00:39:26.304 --> 00:39:39.524
Actually, Charlie, have you, have you guys done any analysis of what it is or- Yeah, well, I was so, yeah, so I mean, uh, on the board of the Oklahoma Bitcoin Association, which is now entering year two of kind of an indefinite hiatus because Bitcoin mining is just very difficult.
221
00:39:39.564 --> 00:39:44.744
But we, uh, we had, um, over half a gigawatt of mining at our peak.
222
00:39:45.244 --> 00:39:58.824
Um, what it looks like now kind of depends upon what the largest one or two, uh, players in Oklahoma, uh, are currently doing, which I don't have a lot of visibility. So yeah. Got it. Yeah.
223
00:39:58.904 --> 00:40:14.253
Um, but I guess throwing it back to you, uh, Nick. Um, uh, like, uh, the SPP is kind of slow behind ERCOT, and it's not like the, it's not like the SP- SPP wasn't attractive before.
224
00:40:14.284 --> 00:40:19.154
It's just that ERCOT seemed to catch all the attention of the miners. Um,
225
00:40:20.404 --> 00:40:39.044
and as economics are at the tightest they've ever been for mining and demand response, like, do you think that, um, with AI kind of taking over the Texas grid, like, we could see mining diversify outside of Texas and Georgia where, where it's, like, been concentrated for a while?
226
00:40:42.864 --> 00:40:53.424
I think you have to, right? With, with, um, you know, what is, what is the interconnect queue? Something like 470 gigawatts, uh, of interconnect. H- how much of that is actually real?
227
00:40:53.524 --> 00:41:03.464
Like, if they approved it, how many, how many, um, how many of those gigawatts are actually gonna get t- turned on? Uh, it's hard to say. But if even half of it is, is real, um, it has to go somewhere.
228
00:41:04.104 --> 00:41:11.364
So I mean, it feels like SPP is probably the n- best next natural. Um, it is a little, it's just different.
229
00:41:11.484 --> 00:41:18.644
And, and maybe, maybe the market has just gotten, become so used to ERCOT, like we always just talk about ERCOT, like ERCOT, when you think of mining, you just think of ERCOT.
230
00:41:18.704 --> 00:41:27.334
But like the vertically, like it's, it's much more vertically integrated, where there's not like the, you don't have like the retail market.
231
00:41:27.644 --> 00:41:34.984
Like, you don't go out and pitch customers directly, um, like you do in, in, uh, in, in Texas. So that's totally different.
232
00:41:35.084 --> 00:41:44.074
Um, so when I said, you know, the under- underfished pond, like how do you, how do you start fishing there? Um, uncertain. Like a- a lot of it is, is just uncertainty.
233
00:41:44.144 --> 00:41:51.444
Like there isn't a playbook for SPP the way there is for ERCOT. Um, and so that, that gives a lot of people a lot of opportunity. Margins are a lot higher.
234
00:41:51.544 --> 00:42:01.844
Um, you know, for example, on, on, you know, ancillary services and, and demand response programs, the, the margins are much higher in SPP just because there isn't as much competition because it's harder to, to, to compete there.
235
00:42:01.964 --> 00:42:18.104
Um, ultimately, that's worse for the, the consumer, that's worse for the buyer. Um, but that is the way of the world, I guess. Um, thoughts on like other grids, PJM, MISO, uh, yeah, I'm kinda- P- P- Yeah.
236
00:42:18.684 --> 00:42:34.136
PJM especially, given what's been happening in the news recently with, uh, it's like multiple gigawatt shortfall or something like that. The, what, the, the, what's the que- the, about the regulatory environment or the-
237
00:42:35.136 --> 00:42:39.746
I mean- Or just like the- Yeah... community response? Uh, kind of all of it. I mean- Okay...
238
00:42:39.756 --> 00:42:53.336
PJM seems like it was one of the more promising zones to build in, and now it's seeming like it's plagued with a host of issues, like regulatory, communities are not very favorable there. I see, yeah.
239
00:42:53.436 --> 00:42:59.056
They're raising their bills. Yeah. That has become a pretty common, pretty, pretty common problem.
240
00:42:59.136 --> 00:43:14.456
I mean, you talked, you even talked to some of the financiers that are financing these big AI data centers, and they're just running into local communities not letting you build or, for whatever reason, they just aren't gonna allow that, that, that permit to get pulled for, for your data center.
241
00:43:14.556 --> 00:43:25.496
Um, and I think, you know, you guys were just talking about Elon's tweet earlier. He's, I don't know if this is true or not, but it sounds like he's having permitting issues. Um, and so, a- and he's not the only one.
242
00:43:25.896 --> 00:43:38.966
Um, you try to build a data center, um, in a blue state, let's just say it, um, then you end up, you know, uh, you end up running into people that don't want your data center there. And, uh- This is a safe space-...
243
00:43:38.966 --> 00:43:48.376
no matter what you do... Nick, you can say these things. I mean, you say that, but, I mean, I'm s- I, here in northeast Oklahoma, people hate data centers. It's out of, it feels out of nowhere. It's the reddest- Yeah...
244
00:43:48.486 --> 00:43:58.466
state in the union. Um, we talked about this like with, uh, Brandon Bailey, uh, you know, looking at DI metrics, kind of reviewing county by county what's common.
245
00:43:58.656 --> 00:44:12.496
And while it, it does like have like at a, at the public national level a like political component, um, grassroots, it almost seems pretty agnostic to politics and it's just, um- I think that's true.
246
00:44:12.636 --> 00:44:23.156
I think the one thing though is with the, the left and with people in blue states, they're much better at collective action. Like we've seen this in terms of being able to organize behind things.
247
00:44:23.756 --> 00:44:32.956
They're, they're much more likely to protest, they're much more likely to get together in these groups, and I think you're missing that a little bit on the right, and push back in rural areas.
248
00:44:32.996 --> 00:44:37.576
It's, to me it's pr- it's, to me it's pretty simple. They both hate the same thing for different reasons.
249
00:44:37.616 --> 00:44:45.996
The left hates it because of the environmental reasons, the right hates it because of like the AI s- like, uh, like takeover, you know, takeover, taking our jobs, et cetera.
250
00:44:46.536 --> 00:44:52.536
And the, when I say normie, I just mean like people that are not in Bitcoin and, and, and AI and data centers.
251
00:44:53.016 --> 00:45:00.736
Their interactions with, with AI are very, very different than the ones that you and I, and probably most of the people listening to this have.
252
00:45:00.816 --> 00:45:13.556
Uh, their interactions with AI are ChatGPT Pro, uh, sorry, uh, free, free version, or some suit, like some AI generated content on Instagram, et cetera. And people are like, they just hate it. They absolutely hate it.
253
00:45:13.816 --> 00:45:21.716
Um, my mom hates it. Uh, people that are like, you know, that don't really use the, the tools like the way that we use them, they hate it.
254
00:45:22.236 --> 00:45:39.696
Um, and that then goes to, "If I hate this thing, like I don't like any of this stuff, why would I care to let them build their stupid data center that's gonna let them face swap AOC and Donald Trump, like, and, and make, uh, a picture, you know, make a corgi dance in the woods?
255
00:45:39.816 --> 00:45:43.916
Like I don't really care about any of that, and I don't wanna build a data center next to my house that allows it."
256
00:45:44.836 --> 00:45:51.516
To be fair, it's pretty funny to face swap AOC and Donald Trump, but yeah, maybe not, maybe not if you have a looming data center.
257
00:45:51.556 --> 00:45:58.196
Okay, I kinda wanna switch some ge- I wanna switch gears, and, uh, we, we've been leaning really into AI lately.
258
00:45:58.326 --> 00:46:09.736
But I wanna go back to the source, Nick, and throw a question or two out to wrap this up about the latest drama in Bitcoin. I was gonna do it. [laughs] Don't do it. Don't do it.
259
00:46:11.176 --> 00:46:22.176
Can the, can the plebs mine Luke's fork of Bitcoin? What's going on here? Okay. Origi- Okay. Uh, where do I even start with this?
260
00:46:22.436 --> 00:46:28.736
So I've been banned from a lot of nonce discussions, uh, unfortunately, which is, which is sad, honestly. It's sad.
261
00:46:28.856 --> 00:46:38.676
But, um, before I got banned, uh, they were talking about the, the proof of work fork, and I was like, "Oh yeah, if you use Blake2b," um, well, it was one of the options.
262
00:46:38.716 --> 00:46:52.216
There was like EagleSong, and, uh, SHA-3, and I don't remember, a bunch of other, um, alternative proof of works for, that most of them have been vetted by alternative coins, altcoins.
263
00:46:52.336 --> 00:47:03.056
Um, and so anyway, they ended up deciding on Blake2b, and they're like, "Oh yeah, we can use the old Siacoin hardware." So Siacoin is Blake2b proof of work. Um, we can use the Siacoin hardware.
264
00:47:03.516 --> 00:47:12.776
Well, the problem is that the nonce space, the nonce space, which is what your miner is actually running through, it's basically just incrementing a number, hashing, incrementing a number, hashing.
265
00:47:13.336 --> 00:47:18.856
You can see here on the screen, the nonce space in Bitcoin is at the end, and it's eight bytes, or f- sorry, four bytes.
266
00:47:19.356 --> 00:47:28.496
Um, the nonce space on a Sia header is somewhere in the middle, uh, and it's, it's, uh, it's eight bytes, not four. Uh, so it's different size, uh, and it's in a different location.
267
00:47:28.556 --> 00:47:41.796
The problem is that the chip is wired to flip the last eight bits of the header, and so it can't... Or sorry, the last 32 bits, uh, four bytes of the header. Um, and you can't change that.
268
00:47:42.256 --> 00:47:54.376
Turns out they are going to change the header format, uh, to support that. So if they do that, then you will be able to mine, uh, BIP 100 or BIP 110 chain with, uh, with Siacoin, uh, miners.
269
00:47:55.396 --> 00:48:03.516
Well, good for them because otherwise all of these plebs just bought ASICs that are gonna be very, very expensive paperweights. Yeah, I was worried they were all gonna be bricked.
270
00:48:03.696 --> 00:48:15.636
I guess the last one, just last final comment is like, you know, Bitcoin has a really pretty mature ASIC, uh, and hardware chain now. It's really dynamic. Yeah. It's pretty commoditized.
271
00:48:15.696 --> 00:48:26.736
You guys have been instrumental in making that happen. Um, but Sia, Bla- you know, Blake2b, ASIC aren't, and I know I'm probably losing a lot of people 'cause they don't care about this, but like, you know,
272
00:48:27.696 --> 00:48:36.356
d- what is structurally different about- All right... this particular hardware chain? So Blake2, Blake2b, um, is ac- was specifically designed to be ASIC friendly.
273
00:48:36.416 --> 00:48:46.661
Like it's, uh, it's, it's easy to design for, it's made to be efficient on, on, um-... uh, on, uh, on physical chips. Like it's, it's made to be efficient.
274
00:48:46.952 --> 00:48:53.692
Um, there's a lot of smarter people that could probably describe it a little bit m- better why it actually is more efficient, but it is.
275
00:48:53.792 --> 00:49:06.512
Um, I mean, if we, if, if Bitcoin had been, um, had been Blake2b instead of SHA-256... Actually, I'm pretty sure Satoshi would've pl- picked Blake2b had it existed back when he originally designed, uh, Bitcoin.
276
00:49:06.632 --> 00:49:23.372
Of course, we're not gonna change to Blake2b now, but, um, anyway, uh, the, if, if, if he had picked Blake2b in the ear- in the early days, the, uh, the hash rate would be something like, it would probably be like 40 or 50% higher, um, if not even more, uh, because it is so much easier to make chips for it.
277
00:49:23.682 --> 00:49:38.142
So there's a good chance, and, and this may be a really good marketing scheme for them, that they actually are able to grow hash rate, hash rate, uh, to be much higher than Bitcoin's, like pretty easily, just because of how easy it is to improve the machines, uh, and get better efficiency.
278
00:49:38.472 --> 00:49:40.842
Um, basically more hashes for the same amount of power.
279
00:49:41.492 --> 00:49:51.292
Um, though it would be like comparing like, you know, back when Ethash was, um, you know, I think, I think Saylor famously compared the amount of hash rate that Bitcoin had to the amount of hash rate that Ethash had.
280
00:49:51.812 --> 00:49:57.492
Um, but, uh, yeah, it would be kinda like comparing those two, 'cause they are completely different algorithms.
281
00:49:57.652 --> 00:50:08.722
Um, and they use different amounts of energy and, and have different numbers of computations and all that stuff. And Saylor didn't understand that? That's, I'm really surprised. No, he like tripled down on it.
282
00:50:08.722 --> 00:50:16.672
It's a legendary thing he said. I'm really surprised about that, though. Maybe he can put it in one of those nice little graphics of a rocket ship with the Stretch logo on it.
283
00:50:16.712 --> 00:50:25.272
You know, with the, with the chick I know that'll help people make sense of it. Yeah. The chick that looks always like, kinda like Natalie Burnell. [laughs] Yeah. Just, yeah, I've never made that connection before.
284
00:50:25.392 --> 00:50:34.012
I wasn't meant to live an uncomfortable life. Um, Nick, HashBender Hansen, thank you so much for your time today. Uh, shout out, everybody go check out Luxor.
285
00:50:34.042 --> 00:50:45.492
Pay, uh, Nick, where p- can people go to, uh, check out Luxor Energy? Just @Luxor on, on Twitter, man. We secured, we secured the grail at Luxor. Hey, N- Nick, I'm gonna ask you a, a quick uncomfortable question.
286
00:50:45.612 --> 00:50:56.732
When GPU- [laughs]... compute futures? Oh, geez. Okay. Wow, that's, uh... Who have you talked to? You just talked to p- am I... Brett Harrison's coming on too? Well, he's on after you.
287
00:50:56.792 --> 00:51:07.502
I mean, we've, we've talked to everybody, man. We've, yeah, so we've talked to the gold players. That's the guy, that's the guy to talk to. Um- Like, is that something that y'all wanna do? Oh, absolutely- Yeah...
288
00:51:07.502 --> 00:51:18.792
we wanna do it. I mean, we list Orange on, uh, we list Orange index on Hashrate Index now just because it's like so interesting. But here's the question, is everything a commodity? I think the answer is no.
289
00:51:18.892 --> 00:51:30.312
I think actually very few things are a commodity, and I'm starting to come around to the idea that maybe compute is not a commodity. I don't know. We might have to dig into that more later, but I'm not sure.
290
00:51:30.952 --> 00:51:43.252
I, I mean- If I rented you my bike, I am not, I'm not finding the bike rental index and f- settling our trade to that index. I'm just saying, "Hey, it's 100 bucks for the month for you to rent my bike."
291
00:51:44.172 --> 00:51:52.452
That's not a commodity. That's just a swap. I like this. We should, we should, uh, we should kick this can down the road sometime.
292
00:51:52.552 --> 00:52:00.912
Do like a compute commodity skeptic, uh, you know, uh, pro, con skeptic everything discussion. So- Yeah...
293
00:52:00.972 --> 00:52:09.692
I'm, I'm also curious- I'm very keen to hear Brett, uh, because we're, we're, we're, I'm, I'm very excited about Architect and very excited to work with Brett. Um, so I wanna, I wanna see what he thinks about it.
294
00:52:10.132 --> 00:52:16.172
Um, the, the, the challenge... Okay, so getting back to your q- q- the challenge we have right now is what I call heterogeneity.
295
00:52:16.432 --> 00:52:22.712
It's basically how do you come up with an index for compute because all of the computes, the computers are very, very different.
296
00:52:23.232 --> 00:52:34.312
A B100 and a B300 and a H100, and an A100, which CoreWeave just recently announced to having still r- in, in production, um, they all can do very different things.
297
00:52:34.332 --> 00:52:47.432
And they all have different GPU, uh, hour, uh, dollars per GPU hour rates. And even once an identical A100 to another identical A100 has a different rate because of how it's being run.
298
00:52:47.492 --> 00:52:59.072
Like, what if your uptime is different or you have a different SLA? Like, the way that you produce that dollar per hour for a c- for a GPU is, is, is different per site, I think. So, um, so it's causing...
299
00:52:59.222 --> 00:53:07.272
I, I think the, the heterogeneity of indices is the, is the challenge. Uh, a lot of people will point to Brent crude and say, "Oh, yeah, there's a lot of different crudes."
300
00:53:07.352 --> 00:53:16.172
They say corn, there's a lot of different corns. Um- Yeah, that's what Kusp Aviary of Orange said on our show a few times. Yeah. Like, eventually it'll collapse into a kind of blended rate like that. Yep.
301
00:53:16.182 --> 00:53:28.192
And so maybe that happens. Um, it's just there's so much evolution. And we're miners talking about, about AI on a mining show. We have been part of the hardware hamster wheel for so long that we're like, I don't know...
302
00:53:28.212 --> 00:53:36.512
They're, they're only, like, these AI guys, they're only on year like four or five of this, and we've been doing it for 14 years. Um, you know, three times as long.
303
00:53:37.092 --> 00:53:45.152
The, the hardware w- flywheel does not, or the hardware hamster wheel does not slow down, but eventually somebody gets kicked out of the ham- the, the, the hamster wheel.
304
00:53:45.192 --> 00:53:58.332
The, you know, the S9 or whatever has obviously been kicked out. S19 is close to being kicked out. Um, how do you index those? Those are w- if, if, if those didn't produce... Anyway.
305
00:53:59.152 --> 00:54:05.212
I know you guys got to get to- No, I, I, I have one last question on this actually- Okay... Nick, 'cause this feeds well into our conversation we're about to have with Brett.
306
00:54:05.272 --> 00:54:16.692
Uh, Luxor's head of derivatives, Matt Williams, has at least, I don't know if he's publicly said this, he said this to me, that he feels like the, like, Orange compute futures are actually solving for the wrong rate.
307
00:54:17.372 --> 00:54:29.692
Um, or sorry, the, um, the, uh, the index is, is, is showing the wrong rate in terms of the rental hours, when really what you wanna be looking at is the, um, end output of that compute for what's being used on the other side, as I understand it.
308
00:54:29.752 --> 00:54:37.692
To me, the analogy here is almost kinda like hash price doesn't reference what it costs to rent out a Bitcoin miner on a cloud contract.
309
00:54:37.752 --> 00:54:45.832
It actually references what you can juice from that hash rate when you're using the machine. Does that make sense to you? Am I totally garbling what Matt was saying with that?
310
00:54:46.848 --> 00:55:00.868
I think, I think, yeah, um, you- no, you're not, not. Uh, that makes sense because, yeah, really what we're pricing is the, the value of the output, and in this case, the output of these AI machines typically is tokens.
311
00:55:01.488 --> 00:55:10.298
But even in tokens there's a lot of heterogeneity, where in, in Bitcoin, the output of hashes is blocks, um, or some expected value of the block.
312
00:55:10.368 --> 00:55:14.788
But, um, that's what, you know, hash price is, is the expected value of a peta hash per day.
313
00:55:14.948 --> 00:55:27.248
Um, the, the problem with, um, [clicks tongue] with c- like, doing that on a compute basis is that the tokens all have different values. So, like, a DeepSeek token and a Claude Table5 token are different.
314
00:55:27.378 --> 00:55:36.148
They have different values. Like, you can go look at, uh, OpenRouter and see, like, how much do you pay per million tokens. Um, we're getting to, like, a rate, though. That's nice.
315
00:55:36.368 --> 00:55:41.348
Um, but the problem is, now how do you... You know, my... Is my token factory gonna be swapping models?
316
00:55:41.628 --> 00:55:56.428
I can't, I can, I can swap between K3 and DeepSeek in GLM, but I can't swap to Fable, I can't swap to Opus, uh, Codex. Those are private, you know, frontier closed source models. Um, it's very difficult. Very difficult.
317
00:55:56.788 --> 00:56:07.128
But Brett is coming up next, and he's a much smarter guy on this than me, and he has got all of the answers, I guarantee it. Um, and I'm looking forward to hearing what he has to say.
318
00:56:07.188 --> 00:56:15.988
Hopefully, I didn't stick my foot in my mouth too bad. Yeah, you should- I-... stick around in the wings. [laughs] Stick around in the wings? What is in the wings? Yeah, like when we take you off stage.
319
00:56:16.228 --> 00:56:23.118
Have you never been in a theater? The wings on each side- Yeah, you can, you can throw que- you can throw questions in the, uh, in, there's, like, a studio side chat- Oh, all right. I'll-...
320
00:56:23.128 --> 00:56:32.528
and if you're lucky, or if you could just do it in the, uh, the comment section on the YouTube stream if you want. Um- YouTube or Twitter? Uh- Which is better? Uh, YouTube is probably preferable.
321
00:56:32.588 --> 00:56:40.567
You make more money there? [laughs] No, no, no. It's just easier to, [laughs] it's easier to see the comments. Nick. Okay. Great. Thanks for hanging out. Get out of here.
322
00:56:40.608 --> 00:56:50.608
We're gonna take you down and we're gonna do a quick ad read, uh, before we bring Brett up. So, cheers, Nick. All right. Thanks, Nick. Thank you guys. See you. All right. We have Brett in the wings.
323
00:56:50.708 --> 00:56:58.168
We're gonna talk more about compute. Future's got it teed up nice and well from Nick. But before that, a word from our sponsor, Ligos.
324
00:56:58.708 --> 00:57:07.528
[Instrumental music] Have you ever used a Bitcoin-backed lender and wondered, "Where do your coins go?"
325
00:57:08.368 --> 00:57:16.508
Well, instead of throwing them into the void and potentially becoming the next Celsius or BlockFi victim, you should do yourself a favor and check out Ligos Finance.
326
00:57:16.628 --> 00:57:25.468
They are our preferred lender here at Blockspace because they use Bitcoin-native smart contracts to make sure that your stack never leaves your hands. That's right.
327
00:57:25.488 --> 00:57:31.888
With Ligos Finance, when you are collateralizing a Bitcoin-backed loan, you control the keys to your kingdom, not Ligos.
328
00:57:31.928 --> 00:57:44.988
With Ligos, there's no rehypothecation, no wrapping, no bridging, just competitive rates as low as 8.5 to 10% APR. Go to ligos.finance to learn more. All right. We're gonna bring on Brett.
329
00:57:45.068 --> 00:57:53.568
We haven't done a mic check, so let's see if this works. Brett, welcome to the show. Hey, guys. Good to see you. Great. Good to see you too. Thanks for coming on. Thank you for joining.
330
00:57:54.168 --> 00:58:06.708
And yeah, Nick, Nick, uh, warmed this up for us pretty well. Um, but before we get into Compute Futures, give me a little TLDR of Architect and what you're building. I'm glad I caught the last five minutes to hear Nick.
331
00:58:06.808 --> 00:58:13.948
Uh, he's, uh, too funny and too kind. So yeah- [laughs]... looking forward to talking about all that with you. Um, so Architect, uh, we're a multi-exchange group.
332
00:58:14.048 --> 00:58:25.428
We have two different regulated exchanges, one in Bermuda, which is regulated for perpetual futures on traditional asset classes like metals, foreign currencies, interest rates, single stocks.
333
00:58:26.088 --> 00:58:48.348
And then in the US, we have a designated contract market license that we are launching soon as the American Innovation Exchange, a US-regulated futures and options exchange dedicated to compute as a commodity, as well as other commodities, currencies, and indexes in the supply chain of AI, everything from metals to electricity and g- oil and gas, uh, to different foreign currencies.
334
00:58:48.388 --> 00:59:04.488
Again, all in the supply chain of AI. And, uh, so we're talking Compute Futures. There's a mon- number of people who are trying to build this. It's a hot topic. Everybody wants them. Um, but what is, like, is what...
335
00:59:04.548 --> 00:59:11.968
Is there anything structurally different about them from other futures, like crude oil or soybeans? How are they different? How are they the same? Sure.
336
00:59:12.648 --> 00:59:23.648
Uh, they're the same in the sense that they are a derivative contract that will expire at particular expirations, let's say every month, every quarter, to a particular underlying index.
337
00:59:24.228 --> 00:59:37.588
Um, as the world of physical delivered compute is heterogeneous, which I'm sure is a topic that we'll discuss a lot, uh, most of the futures exchanges are starting off with cash-settled futures, not physical delivered futures.
338
00:59:37.628 --> 00:59:48.588
So at the end of that expiration, you have a cash amount equal to the expiring value of an underlying index. What is different about this is that it's kind of a commodity unlike any one we've seen before.
339
00:59:49.108 --> 01:00:02.408
Um, it, it's not... What exactly it defines, you know, an H100 hour, a B300 hour, is going to be different. There needs to be quite a lot of interesting index creation for this to be a valuable product in the market.
340
01:00:02.438 --> 01:00:19.028
There, there have been other interesting products that have been tried in the past along these lines, but this is, um, the combination of the current excitement about this commodity for the modern e- economy that it, it, it becomes so crucial for, and its heterogeneity is part of the, the novelness of the, of the futures product.
341
01:00:19.128 --> 01:00:30.438
Um, I, I would like to piggyback off of the physical versus cash-settled. How would you even go about physically settled futures in, you know, if you're a company taking, you know...
342
01:00:30.448 --> 01:00:39.698
If you're going long on this, this compute, unless you're an actual operator, I guess that's the only profile that makes sense to do physically settled. Like, how would that actually work in practice?
343
01:00:40.928 --> 01:00:47.828
So it would work on both the long and the short side. Uh, think of it on the long side, you're an inference provider. Uh, you know you're gonna...
344
01:00:47.888 --> 01:00:52.778
You're training some, you're working on b- some big model that you're not ready to train yet, but you're gonna train it starting in six months.
345
01:00:53.078 --> 01:00:57.768
So you might buy a future today hoping to take delivery in six months of compute.
346
01:00:58.372 --> 01:01:11.552
On the sell side, you know, you have excess capacity, let's say, that's going to come up in a year when some year contract is rolling off, and so you might wanna sell forward capacity one year from now to lock that in.
347
01:01:11.892 --> 01:01:23.932
So the buyer and the seller side both make sense. Um, but to your point, it- far be it from just one futures exchange to set the delivery standard for the entire market. That's gonna be quite difficult.
348
01:01:23.992 --> 01:01:30.292
I mean, first of all, it's moving so fast. I mean, e-e-every time we talk to someone, the chip they want is different, the place they want it delivered is different.
349
01:01:30.392 --> 01:01:39.932
Um, and so we're trying to keep pace with that with our futures products that we're launching soon. Uh, but, but second is that people have all different kinds of needs, and no c- no one compute is the same.
350
01:01:39.992 --> 01:02:12.892
So what we're trying to do actually is do a kinda best of both worlds approach by launching cash-settled futures, but al- also launching an over-the-counter exchange for physical market, an EFP market, which is something that's very common in commodities and even in e- in equity index worlds, where people can take their long future or their short futures position, and then engage in a, you know, bilaterally negotiated basis trade from that future to whatever the actual compute is that someone has, maybe pay a slight premium or a slight discount, and get the actual compute.
351
01:02:13.472 --> 01:02:22.682
I know you mentioned corn before. Obviously, corn is not a super analogous commodity to, uh, [chuckles] to compute, but n- a lot of people don't actually know how corn futures work.
352
01:02:23.152 --> 01:02:36.612
So if you buy a corn future, what you get delivered is number two yellow corn. But if a supplier actually has ya- number one yellow corn, they can deliver that instead, and the buyer pays a little bit of a premium.
353
01:02:37.092 --> 01:02:41.112
Or they can deliver number three yellow corn, and the buyer pays a bit of a discount.
354
01:02:41.172 --> 01:02:54.152
So there is already some precedent for the idea that this heterogeneous commodity world can have a deliverable commodity with different grades, with different offsets, depending on what is actually being delivered, and I think that is a standard that will eventually make its way into the compute world.
355
01:02:55.572 --> 01:03:07.972
Um, so I guess let's double tap on heterogeneity 'cause you kinda got there. Um, that's heterogeneity in, like, the chips. What about the, the, the tokens themselves, like running DeepSeek on,
356
01:03:09.012 --> 01:03:19.952
uh, versus Claude versus, you know, Kimi on, on the same hardware. Those tokens are... Are they, are they the same? Do- are they heterogeneous in some capacity? H-how do you solve for this?
357
01:03:19.992 --> 01:03:21.432
What do you-- how do you describe this challenge?
358
01:03:23.372 --> 01:03:40.132
So thinking about, again, analogous commodities markets, and obviously it's difficult to reason entirely by analogy here, um, th- there's something in, in commodity markets called a crack spread, which is where you take in crude oil and you output refined products like diesel or gasoline.
359
01:03:40.712 --> 01:03:52.682
So one is a raw input and one is a refined output. I think in the compute, uh, landscape, the compute hour, the GPU hour, is like the crude oil. It is the kind of raw input.
360
01:03:52.692 --> 01:04:05.152
It's the thing that you need to produce intelligence, whereas the refined product like diesel is the output. The commodities market largely doesn't care that much about what crude oil is being used for.
361
01:04:05.192 --> 01:04:10.292
There's lots of different needs and reasons for having crude oil. The actual refined product could be different.
362
01:04:10.372 --> 01:04:18.662
So, like, a Fortune 500 company is probably not consuming compute, I mean, maybe they might be, but for the most part they're not. They're consuming the output of compute, which are tokens.
363
01:04:19.162 --> 01:04:31.392
So I, I do think that, uh, maybe starting in the over-the-counter markets first and then working their way up to the sort of centrally cleared standardized product market, there will eventually be some kind of futures or options product related to tokens.
364
01:04:31.952 --> 01:04:38.372
In fact, the, the Shanghai Futures Exchange has announced that they're gonna start with that. They're actually not starting with the GPU hour as a futures product.
365
01:04:38.382 --> 01:04:43.322
They wanna start with tokens as their, like, main commodity they're trying to build a derivative based on.
366
01:04:43.332 --> 01:04:55.092
But to start, we're kind of focusing on what we think is the nexus of this commodity market, which is the GPU hour. Brett, I wanna shift our attention back to the concept of the GPU hour as the reference rate here.
367
01:04:55.232 --> 01:05:05.761
Do you think that that will be the reference rate that these futures and forwards are built on, uh, in, in the long term, or do you think that there's a different reference rate, like tokens or something else that we might be using?
368
01:05:06.282 --> 01:05:17.872
'Cause I do keep coming back to this idea of, and the analogy I use here is, is, uh, Bitcoin mining, which, like, the dollar per GPU hour is kind of like what a hosting or a cloud contract rate would be for that hash rate.
369
01:05:18.231 --> 01:05:27.672
But it doesn't actually give you the output of that hash rate in terms of what it's worth. Sure. Uh, it's just what the, the, the supplier is willing to sell it to the end buyer.
370
01:05:28.072 --> 01:05:34.332
So do you think that there's any flexibility for, like, multiple different structures here? Do you think we'll converge on something else? What are your thoughts on that?
371
01:05:36.022 --> 01:05:51.792
It, it's something that we're both following and trying to stay nimble and on top of as the market evolves because, yeah, as, as you say, like, maybe in six months we realized the H200 GPU hour was the wrong unit and that doesn't make a good future, and we'll get to that in a second.
372
01:05:51.802 --> 01:05:55.052
A-a-a-and I, I've been seeing other attempts at different kinds of indexes.
373
01:05:55.082 --> 01:06:11.512
So for example, uh, ICE, uh, Intercontinental Exchange, which, um, is the parent company of NYSE and they're launching their own compute futures, they announced they're gonna work with this company called Nativex that has something like a, a, a normalized GPU hour for power costs.
374
01:06:11.552 --> 01:06:16.452
It's like kind of like a power hedge GPU hour. Uh, maybe that's the right metric. I don't know.
375
01:06:16.992 --> 01:06:25.252
What, what, what I'm sort of seeing in the market is that if you see what people are texting back and forth to each other in the compute markets, they're asking for GPU hours.
376
01:06:25.812 --> 01:06:32.692
They're saying, "Hey, I need three years of 1024 B300 GPUs. H- what's the price per hour you can give me?"
377
01:06:33.232 --> 01:06:44.792
And because that is the thing everyone is asking for quotes for, almost like tautologically, that is the correct unit for now. You know, no one is saying, "Hey, what's my GPU per megawatt that I can get?"
378
01:06:44.892 --> 01:06:48.692
People are just saying, "What's my GPU hour?" So that's what we're building.
379
01:06:49.352 --> 01:07:04.692
Um, now, in terms of how you actually structure a futures product, uh, what we-- what the challenge we have as an exchange is figuring out how do we make something that is general enough so that we can aggregate liquidity in a few number of order books?
380
01:07:05.120 --> 01:07:12.550
But specific enough that it can provide a good hedge for people who have some kind of portfolio that is correlated with compute prices.
381
01:07:12.950 --> 01:07:24.740
So you can imagine we could do a all Nvidia GPU price index that takes A100s, H100s, H200s, B200s, B300s, and makes like an index almost like the S&P 500.
382
01:07:24.840 --> 01:07:35.540
Or we could go the other direction and say, "I want this particular SKU of H100 InfiniBand 32 gig in Virginia," and that's the future, and we'd have hundreds of those.
383
01:07:36.040 --> 01:07:45.950
And I think for now we've sort of, the market has kind of, uh, coalesced around this idea that the, the H200 aggregate US GPU hour is the right metric.
384
01:07:46.050 --> 01:07:49.560
But we might find that we should launch more generic or more specific ones in the future.
385
01:07:51.140 --> 01:08:12.880
And around this GPU hour, 'cause, uh, we spent, we're on, we're wrapping up two weeks of covering earnings of power shells, neo clouds in particular, and those earnings kind- they live and die by the financing and the deals they've announced, and this is downstream of GPU hour pricing, which has been very volatile since we've even been able to start tracking it.
386
01:08:12.920 --> 01:08:19.020
I mean, down something like 50% on some indices last fall or last, uh, winter.
387
01:08:19.760 --> 01:08:39.020
Um, maybe this is just the story of any kind of commoditization in any sector, but like as GPU hours anticipatedly become commoditized, what does that mean for the industry such as neo clouds and their ability to access financing and capital and like just change their business?
388
01:08:40.660 --> 01:08:41.180
Yeah, it's interesting.
389
01:08:41.300 --> 01:08:54.140
I, I, I, I put up a post this morning actually talking about the correlations between, uh, neo cloud stock prices and GPU hour prices, and right now there aren't really any correlations to speak of.
390
01:08:54.160 --> 01:08:56.530
And I think a large part of that has to do with the fact that
391
01:08:57.980 --> 01:09:09.700
in some ways a CoreWeave is, is already forward hedged for compute because they are selling out like five-year forward contracts on all of the capacity in their data center, and most of that is already booked up for years and years.
392
01:09:10.580 --> 01:09:29.380
So they're an example of a company which maybe today wouldn't be the, the perfect consumer of a, of a futures product on compute, but there is a long tail of other, you know, cloud providers or GPU-as-a-service providers or inference providers who either on the short or long side of that futures forward curve would, would have a lot of exposure.
393
01:09:29.460 --> 01:09:39.480
Um, what, what, what I think it means for the financing market is, is this, and this is what we've been hearing directly from customers in the space. Uh, so h- how do you actually build more capacity?
394
01:09:40.260 --> 01:09:53.380
These are obviously very expensive, very capital expenditure intensive prod- projects, and so you have to go out and maybe get a giant loan to go like build your data center, to get the electricity, all the, the hardware, the GPUs, the networking, the cooling.
395
01:09:55.040 --> 01:10:05.000
In order to get a loan at a reasonable rate, uh, y- you basically have to be a triple A rated, you know, corporate. That's the only way you're getting like a 6 to 8% loan in the market.
396
01:10:05.420 --> 01:10:21.130
If you're not, you're being offered something like 20 to 25% interest rate for a loan to be able to build out a data center, and that is completely like u- untenable for a lot of companies, especially if you're a startup in this space that's trying to kind of use, let's say, equity raised capital to build out a data center.
397
01:10:22.060 --> 01:10:42.070
So what we believe is that if, if there's actually some- something like an options product, a put option where you can guarantee to your lender that no matter what, I will be able to sell my GPU hours at at least $3 an hour or $3.50 per hour, even if the whole market for this tanks, I have completely hedged that risk and transferred to someone else.
398
01:10:42.080 --> 01:10:47.360
You can imagine the lender saying, "Well, because you're hedged, I can lend this out at 12% interest rate instead."
399
01:10:48.110 --> 01:10:55.380
And all of a sudden that unlocks all of this financing for a, a huge market that is, is completely supply constrained to grow faster and faster.
400
01:10:56.020 --> 01:11:10.940
And, and, and my take on this Nvidia announcement from a couple of days ago is that right now Nvidia itself is behaving as that like put option seller of last resort, which I- is fine for now, but that is not the way that the market is supposed to develop.
401
01:11:10.960 --> 01:11:26.059
Like these risks are supposed to be transferred to financial institutions like private credit firms and banks and hedge funds that can take on that risk and hedge it and manage it so that neo clouds don't also have to be their own like kind of like hedge fund of risk basically.
402
01:11:27.160 --> 01:11:41.690
This reminds me of something we heard a lot when I used to work at Luxer about when they launched their hash rate forwards and then eventually the futures, is that it does give them the ability to lock in better rates for their financing because they can give you that certainty on the revenue.
403
01:11:42.160 --> 01:11:54.560
Also kind of reminds me of the way that in some ways the way that prepayments are kind of working right now for some of these companies in terms of being able to fund construction costs upfront and maybe show the lenders that they actually have income coming in in the interim.
404
01:11:55.220 --> 01:12:01.220
O- One last question from me, Brett, and, uh, this is a closer for me, and then if Charlie has anything else, he'll chime in here at the end.
405
01:12:01.940 --> 01:12:18.820
As we look at the commoditization of compute and it continues, [clears throat] you know, the, the, the margins, if things go the way they usually do i- in these, uh, w- with the commodities, that they continue to compress, is it just a winner take all for the hyperscalers?
406
01:12:19.260 --> 01:12:23.860
Like do you see a world in which the neo clouds end up getting gobbled up by them?
407
01:12:24.440 --> 01:12:40.570
Like when margins are thin and you need, um, significant capital to continue to build out, do you see a world in which the current model, power shells, neo clouds, et cetera, stands or do you see a world in which the hyperscalers just gobble up all of the other parts of the value chain?
408
01:12:42.700 --> 01:13:01.980
I, I think that the industry will develop no differently than sort of the, the software startup industry, which is that of course there'll be a lot of consolidation at the top, and at times from the outside looking in, you'll look at it and say, "How could anyone possibly compete with like the giant machines that are these enormous hyperscalers?"
409
01:13:02.010 --> 01:13:09.780
And maybe at some point we'll, we'll call CoreWeave a hyperscaler. You know, like maybe they'll just consolidate and consolidate and get larger and larger, and that's what they will end up doing.
410
01:13:09.900 --> 01:13:23.772
Um, but the reality on the ground is there will always be space for a smaller, nimbler, more specific, more niche driven operator to be able to serve some need.... that a giant hyperscaler won't be able to.
411
01:13:23.881 --> 01:13:31.972
And that's why, you know, there's now hundreds of different neo clouds out there, and they're all serving different regions, different kinds of chips, different purposes. They might be
412
01:13:33.032 --> 01:13:43.502
focusing on open source models for inference, or they might be focusing on the very latest chip, or they're focusing on the very cheapest chips and providing really cheap inference for high-performance computing and scientific computing tasks.
413
01:13:44.262 --> 01:13:51.592
It, it-- there's gonna be such a wide ar- array of opportunities here that are gonna continue to grow that I don't, I don't see that pie shrinking up and consolidating too much anytime soon.
414
01:13:52.952 --> 01:13:58.532
I think that's a great place to end it. Brett, thank you so much for hopping on the show and talking compute with us.
415
01:13:58.672 --> 01:14:10.232
Uh, yeah, uh, where could be-- where sh- uh, what, uh, where could people go to follow more of your work, your blog, your insights, et cetera? Yeah, sure. So I'm just @brettharrison on Twitter. Uh, I write a lot there.
416
01:14:10.492 --> 01:14:22.972
Um, we publish a lot of those as blog posts on our website, arkitekt.co. And then people can, uh, pre-onboard to our exchange at ai.exchange, uh, for when we're getting ready to launch these compute futures. Cool.
417
01:14:23.332 --> 01:14:37.432
Thanks, Brett. Thanks a ton for your time. Thanks, guys. See ya. Have a good weekend, man. Thank you. You too. All right, Charlie, this is a marathon show, and we can't tap out yet. We can't tap out.
418
01:14:37.532 --> 01:14:46.432
We have one postponed, delayed earnings. We gotta do the SNLH, right? Yeah, we gotta get Soluna's Q2 up here.
419
01:14:46.552 --> 01:14:53.772
We were gonna do this f- at the front of the show, but as I learned from a lot of these earnings updates, these take longer than I usually think they should.
420
01:14:54.492 --> 01:15:07.672
So- Sometimes you gotta, sometimes you gotta, uh, rip it with the CFO first and then do the earnings. So, uh, pull that on up here. Yeah, so we're gonna get Soluna's earnings up here. Let me remove that. Here we go.
421
01:15:09.872 --> 01:15:19.852
This is coming to y'all from Blockspace. Headline reads, "Soluna Q2 revenue rises a hundred and forty-five percent as development pipeline reaches six point three gigawatts."
422
01:15:19.892 --> 01:15:31.672
I'm gonna go ahead and put a huge asterisk on that pipeline. Their actual... Y- you know, all, all of these Bitcoin miners turned AI companies kind of treat pipeline differently.
423
01:15:32.512 --> 01:15:36.412
I would call most of that six point three perspective, like things that they're looking at.
424
01:15:37.032 --> 01:15:48.072
But in terms of gigawatts that they're actively working on rolling out, from what I can tell from digesting their 10-Q, it's like roughly a gigawatt. Um, Bitdeer is kind of infamous for this.
425
01:15:48.192 --> 01:15:57.652
They would always post these crazy pipelines for gigawatts, and like most of it was stuff that they were basically just trying to do due diligence on to figure out if they could build there.
426
01:15:57.892 --> 01:16:09.092
[clears throat] But that's a lot different than having load studies and o- other things submitted to local grid authorities, right? So I just wanna throw that out there as a caveat to begin.
427
01:16:09.252 --> 01:16:22.732
And I'm gonna go ahead and cover some of the numbers from this, but mo- most of what we're gonna be focused on today is a lot of qualitative stuff with regards to their AI pivot and some, um, some of the things that we talked about with Mike Peachey on the interview.
428
01:16:22.852 --> 01:16:32.432
So, uh, data hosting revenue, which is Bitcoin mining hosting, I wanna also get that out front, none of this is AI-related revenue, was twelve point six five million Q2.
429
01:16:32.592 --> 01:16:40.552
Crypto mining, that's their self-owned mining, was one point seven two million. Wind energy generation came in at three hundred and seventy thousand dollars.
430
01:16:41.232 --> 01:16:51.692
Demand response came in at three hundred and twenty thousand dollars. HPC services came in at zero. Total revenue was fifteen point zero six million, up from six point sixteen million in Q2 2025.
431
01:16:53.092 --> 01:17:05.092
So there was a net loss for, um, for Soluna of twenty point seven one million. Most of that is related to general and administrative.
432
01:17:05.772 --> 01:17:13.012
Um, six-month stock-based comp was nineteen point seven million, so that's doing the majority of the lifting. That's against twenty-four point five million of six-month revenue.
433
01:17:13.432 --> 01:17:21.312
Also comes on top of a lot of the dilution that we talked about. They issued a hundred and thirteen million dollars worth of shares in the first half of the year, I believe.
434
01:17:22.052 --> 01:17:37.492
And so a, a decent amount of dilution coming from Soluna, hence our question with Peachey, who said that they were going to be approaching roughly 80/20 split between debt and equity for the AI builds that they have their eyes on, and they're going to try to lean on project-based financing.
435
01:17:37.852 --> 01:17:50.572
Think he cited like TerraWolf and Cipher, other companies that have-- basically looking at their structures and saying you should expect something similar. So yeah, ju- just covered the dilution, uh, piece there.
436
01:17:50.652 --> 01:18:03.552
Not really too much to add, but I wanna move on to the operations right now. So portfolio as of June 30th, they have five operating projects. Project Sophie, twenty-five megawatts in Kentucky.
437
01:18:03.592 --> 01:18:16.072
Dorothy, uh, which is split between 1A and 1B, that's 50 megawatts energized. And, um, up to, um, w- and they have an expansion at-- for Dorothy 2 at 48 megawatts.
438
01:18:17.212 --> 01:18:34.692
And they also have this W- uh, Willacy County, Texas, 83 megawatt site. The Briscoe Wind Farm is 150 megawatts. Um, as we discussed with Mike Peachey, the 10-Q is kind of-- Uh, the accounting on it was kinda confusing.
439
01:18:34.752 --> 01:18:38.452
I was like kinda wrapping my head against my keyboard this morning trying to make sense of it.
440
01:18:38.472 --> 01:18:52.212
But Mike basically made the point that the great thing about the Briscoe site, as I understand it, is they can sell to the power company that they have a PPA with when they need to, and otherwise they can consume all of that locally.
441
01:18:52.272 --> 01:19:00.632
Which is really good news when you consider that oftentimes when it's super windy in this corridor of Texas, power prices can be negative for that load.
442
01:19:01.112 --> 01:19:08.492
So the-- I- if you have to sell that back, that's a t- big problem for them. Obviously, I don't think they would have bought it if they had to do that.
443
01:19:09.152 --> 01:19:24.352
But they paid roughly fifty million dollars for the Briscoe site, fifty-one point four million in April of 2026. And again, that's 150 megawatts of wind generation and will be helping the Dorothy build-outs there.
444
01:19:24.832 --> 01:19:34.174
Charlie, I don't know if you have any commentary to add. No, this, this Briscoe site, is this, is this K-A-T-I, Kati? No, so this, the, the Coty, I'm glad that you brought that up. So the- Yeah...
445
01:19:34.174 --> 01:19:46.194
Coty site is, um, they have Coty one and Coty two, and Coty two right now is the crown jewel of their AI aspirations. They, on June 3rd, 2026, Soluna entered into a joint venture.
446
01:19:46.284 --> 01:19:49.244
This comes directly from their press release, I'm quoting here.
447
01:19:49.274 --> 01:20:00.024
"A joint venture agreement with Metro Blocks to develop Coty two, phase one, for a 100-megawatt critical IT data center development, with phase two adding a further 250 megawatts.
448
01:20:00.444 --> 01:20:11.024
Soluna holds Class A interest and serves as the manager of the joint revenue." [sniffs] That's really interesting, the, the joint revenue structure here. You know, Soluna has kind of--
449
01:20:12.884 --> 01:20:21.584
They've gone into a number of their sites in a kind of a joint venture structure, and we talked about this with Peachey.
450
01:20:21.624 --> 01:20:31.324
They actually purchased back all of the Class B equity for the Dorothy 1A and 1B site over the course of the quarter.
451
01:20:31.334 --> 01:20:38.564
We also talked to them, uh, talked to Mike about, like, whether we should expect that going forward with some of their other sites because the structures are a little bit split.
452
01:20:38.664 --> 01:20:45.704
I believe Soluna mostly owns all Class A for these sites, but for some of them, they do not own the full Class B share structure.
453
01:20:46.684 --> 01:21:14.764
[inhales] Uh, w-with regards to development outside of the Coty two, they've got roughly one point zero three gigawatts gross across seven ERCOT projects, Dorothy three, 300 megawatts projected for 2027, Rosa, a hundred and seven-- hundred and eighty-seven megawatts, uh, no, no timeline there, and a, a number of other sites, Gladys at 150, Hedi at 120, Ellen at 100, Faye at 100, and Annie at 74 megawatts.
454
01:21:15.484 --> 01:21:22.924
Um, all of these are greenfielded, uh, and all have allocation pending except for Dorothy three, which is in load study.
455
01:21:23.004 --> 01:21:36.184
[lip smack] What I wanted to highlight here, Charlie, is the AI business line, specifically the fact that the JV is really the first major milestone they've hit with regards to AI revenue.
456
01:21:37.144 --> 01:21:59.234
For listeners who were on the call with us when we had Mike Peachey, I asked him about this nineteen point three billion dollar liability that's on their books that was traced back to a Hewlett-Packard purchasing agreement they had from June twenty-four, twenty-- uh, June 24th, 2024.
457
01:22:00.424 --> 01:22:06.024
So back then, Soluna basically launched its cloud compute line.
458
01:22:06.044 --> 01:22:14.844
The headline from the SEC filing and the press release is, quote, "Soluna unleashes first batch of Nvidia GPUs targeting eighty million revenue over three years," end quote.
459
01:22:15.664 --> 01:22:27.564
Um, they took on a twelve point five million dollar credit facility at their new subsidiary, Soluna Cloud, the first tranche of a larger raise led by BitOODA Technologies and Imperial Capital.
460
01:22:27.644 --> 01:22:33.744
They ha- they bought Nvidia H100 GPUs under a three-year-- or they, they were, um, excuse me.
461
01:22:33.804 --> 01:22:43.604
They were launching Nvidia H100 GPUs under a three-year managed services agreement at a tier three hydropower data center run by an unnamed, quote, "strategic OEM partner," end quote.
462
01:22:44.044 --> 01:22:51.184
You gotta wonder if that was Hewlett-Packard themselves. Services were available starting as early as July 2024.
463
01:22:52.304 --> 01:23:03.734
Now, that twelve point five million dollar credit facility is listed in their accounting as a green cloud note, and, um, they, they launched into a few deals after this.
464
01:23:03.784 --> 01:23:14.064
So on December 3rd, 2024, they had the San Francisco Compute deal. This was a potent- This had a potential maximum revenue of two point four six six million over one year.
465
01:23:14.564 --> 01:23:23.624
The master services agreement listed an initial eighty H100 scaling to a possible 120 with variable pricing and expansion options.
466
01:23:23.633 --> 01:23:37.054
San Francisco Compute was described as the first customer out of the HPE Greenlake collaboration, and it was-- they were listed as kind of this anchor supplier, um, called, uh, the CEO, Ethan Anderson, called Soluna an anchor supplier.
467
01:23:37.754 --> 01:23:47.304
They also had this Atlas Cloud contract. Um, this was a one-year engagement for sixty-four H100s with an option to scale.
468
01:23:47.404 --> 01:23:59.104
[inhales] Um, Belzere called it the second engagement off of the HPE Greenlake deal, [lip smack] and Soluna said the deal took it to one-third of its initial HPE Greenlake cluster sold.
469
01:23:59.144 --> 01:24:12.804
Deployment began within a test phase, production to go live January 2025. Um, both had December 2024 announcement dates, and roll-up came in January 2025. Uh, here's the thing, though.
470
01:24:12.844 --> 01:24:21.604
None of those contracts are currently active. The Atlas agreement was explicitly one year from late January 2025 go live, so it would've expired this year.
471
01:24:22.114 --> 01:24:30.704
The San Francisco Compute MSA carried a maximum potential revenue of nearly two point five million over one year, and, um, that rolled over as well.
472
01:24:31.544 --> 01:24:47.524
Soluna's own framing at the time, the two customers took one-third of the initial cluster, but two-thirds seem to have never sold, and their total HPC services revenue recognized across the entire program was only twenty-eight thousand dollars all of the first half of 2025.
473
01:24:47.644 --> 01:24:56.704
So I-I just have a few questions with regards to, and I'm gonna field these with John Belzere. Like, flagship rollouts for a cloud
474
01:24:57.984 --> 01:25:11.544
platform, uh, only materialized in twenty-eight K in revenue for the first half of 2025, and it seems like maybe those, those counterparties, I mean, seeing that they didn't book any of the other revenue that they projected for this, they might have backed out of this.
475
01:25:12.244 --> 01:25:21.584
This also is important with regards to the financing for the HP, uh, the financing for the HPE deal or the purchase commitment.
476
01:25:21.594 --> 01:25:30.864
They made a total commitment for thirty-four million dollars over thirty-six months with Hewlett-Packard for these units. The initial prepayment was ten point three million.
477
01:25:30.904 --> 01:25:44.612
They still owed nineteen point three million as of March 2025, and they, uh, have Made no payments on that since. That was the $19.3 million liability I asked Peachy about.
478
01:25:44.652 --> 01:25:53.152
And I'm curious, like, he said that-- Oh, I can't remember the term that he used. He said it was, like, inert. I, I don't know why. They have $130 million in cash.
479
01:25:53.212 --> 01:26:03.392
I'm curious why they didn't s- just pay down that entire commitment. And I wonder if there is any-- There was nothing in the legal or the litigation parts of the 10-Q to suggest that there is litigation over that.
480
01:26:03.412 --> 01:26:14.152
But I am curious about that, that liability. I'm also curious what happened in terms of the earliest cloud contracts that they had inked and their aspirations.
481
01:26:14.192 --> 01:26:25.492
They had one-third of those HPE or the Hewlett Packard, uh, GPUs contracted. Didn't get the other two, two-thirds contracted, and those other contracts just fell to the wayside.
482
01:26:25.552 --> 01:26:34.812
I, I'm really curious about what, what happened with that. I just feel like I wanted to flag that 'cause that was something that was kinda buried in the 10-Q, and you kinda had to go digging looking for it.
483
01:26:35.772 --> 01:26:49.632
Hewlett Packard, not a name you hear [chuckles] that, that much anymore, I'll say. Uh, yeah, I don't have a whole lot of insight, uh, to this, uh, as, uh, earnings are definitely not my specialty.
484
01:26:49.692 --> 01:26:58.532
Yeah, one last note on it. Um, Hewlett Packard terminated the agreement with cause or for cause in March of 2025.
485
01:26:59.332 --> 01:27:06.272
So, uh, again, it's been since March 2025, and that the, the outstanding liability hasn't been extinguished.
486
01:27:07.052 --> 01:27:15.722
So all that being said, I wonder if what you'll see from Luna going forward, you know, they were kinda going, like, small cloud cluster for the original rollout.
487
01:27:16.232 --> 01:27:25.232
I would imagine with this JV, for example, they're gonna go more powered shell route, uh, which makes a lot of sense. Play to your strengths, manage energy, don't worry about the clusters.
488
01:27:25.272 --> 01:27:35.012
We've seen that from basically every single Bitcoin miner in our coverage, so. If you've got the pipeline, maybe the power shell- Yeah... model makes sense. Yeah. I think it does, yeah. I think it makes the most sense.
489
01:27:35.072 --> 01:27:43.262
Even with, like, Iron, you know, um, as far as I understand, Dell is actually managing the GPUs for the Microsoft contract.
490
01:27:43.822 --> 01:27:55.092
Like, I hear- Yeah, I mean, and, and if you look at Rittenhouse's critique, [chuckles] that's one of his big things, is they're not actually managing anything ev- So, um- Yeah, he's been on the- Yeah... warpath recently.
491
01:27:55.232 --> 01:28:05.172
I know. He has. Also against Nebius. Yeah. All right, I think that wraps it up for today, Charlie. Thank you for watching Blockspace Live coming at you every weekday, 1:00 PM Eastern.
492
01:28:05.392 --> 01:28:12.412
We are Compute's daily live show, especially today. Man, we were definitely Compute's live show, uh, this beautiful Friday the 14th.
493
01:28:13.372 --> 01:28:24.732
If you missed the live stream, you can catch the podcast on anywhere podcasts are found. And, uh, make sure to check out our newsletter at our website, blockspace.media.
494
01:28:24.852 --> 01:28:40.432
Newsletter direct sign up at newsletter.blockspacemedia.com. Thank you so much for listening. This show is brought to you by CleanSpark, NASDAQ-listed ticker CLSK. I'm Charlie. I'm Colin. We'll see you on Monday.
495
01:28:40.772 --> 01:28:51.352
[upbeat outro music]