AMD’s $5B Anthropic Investment, OpenAI’s $750B Compute Spend, Galaxy’s $3.5B Raise

Summary

Blockspace: AI & Bitcoin

Episode: AMD’s $5B Anthropic Investment, OpenAI’s $750B Compute Spend, Galaxy’s $3.5B Raise
Date: July 22, 2026
Hosts: Charlie Spears, Colin Harper

Main Theme

This episode dives into the convergence of AI and Bitcoin infrastructure, examining colossal investment deals and the rapidly evolving landscape of compute and data centers. The hosts break down AMD’s $5B investment in Anthropic, OpenAI’s jaw-dropping compute spend trajectory, Galaxy Digital’s massive capital raise for data center expansion, and the transformation of distressed mining assets to AI infrastructure—showcasing how former Bitcoin mining strongholds are pivoting to lead the AI revolution.

---

Key Discussion Points & Insights

1. AMD & Anthropic’s $5 Billion AI Infrastructure Deal

[02:03–09:45]

• Overview:
  • AMD, seen as the "little brother" in the GPU arms race, commits up to $5B to Anthropic for a 2 GW AI infrastructure buildout using AMD’s MI450 series GPUs.
  • The deal marks AMD’s increasing relevance, challenging Nvidia, and builds on prior deals—including with OpenAI and Meta.
• Industry Context:
  • AMD’s MI450 GPUs focus on higher memory per chip, an edge for massive-context AI models. Frontier model bottlenecks are shifting from raw compute to massive context windows.
  • AMD’s partnerships: Multi-gigawatt deals with OpenAI and Meta position MI450 as an emerging industry standard.
• Market Share Deep-Dive:
  • Nvidia dominates with ~87% of the data center market ($75.2B in Q1 2027), while AMD holds a modest but growing 5–8%.
  • Custom silicon (Google, Amazon, Microsoft, Meta) fills ~15–20% and is growing over twice as fast (44.6% YoY vs 16.1% for merchant GPUs).
• Circular Financing & Industry Interlocks:
  • OpenAI received warrants for AMD shares; now, AMD invests in Anthropic.

    "Is this, does this mean OpenAI actually by proxy kind of owns part of Anthropic?... Now does the equity move directly in a circle where the competitors now have proxy ownership to each other?"
    —D, [07:41]

• Deployment Timeline:
  • OpenAI is the anchor customer for new GPUs (receives first); Anthropic follows in 1H 2027—a first for running on AMD chips.
• Notable Quote:

  "The numbers stop making sense."
  —C, [09:45]

---

2. OpenAI’s $750 Billion Compute Commitment

[11:01–25:16]

• Headline:
  • OpenAI raises its projected compute spend to $750B through 2030 (from prior $600B estimate). Not to be confused with its $1.4T infrastructure goal.
• Investors & Partners:
  • Agreements with Microsoft, Oracle, AWS, CoreWeave; hardware affiliations with Nvidia, AMD, Broadcom.
• Financials:
  • Most recent numbers:
    • 2025 GAAP revenue: $13.1B
    • EOY projected run-rate: $20–21B/month ($240–252B annualized)
    • Not profitable—burning ~$27B/year now, projected $63B in 2027

  • "You're basically writing checks to be cashed in three to four to five years, which are multiples over your current gross revenue today."
    —D, [15:23]

  • Growth rate at present: 61% YoY, but even at that pace, revenue lags far behind spend projections.
• Data Center Strategy:
  • New $20B "Project Chelia" in Georgia (Effingham County), OpenAI’s first self-led build (not just leasing).
  • Contracted for 3.2 GW from Georgia Power (2028–2032); Brent Mayo (ex-XAI, built Musk’s Memphis Colossus facility) joins as architect.
  • Shift from nuclear generation to largely natural gas for fast deployment, reflecting reality of US energy mix expansion for data centers.
  • Georgia’s data center pipeline: rapid growth, projected to add 8,500 MW of load by 2030 (15B plan to lift capacity 50% over six years).

    "Georgia is definitely a state to watch as these build outs continue."
    —C, [24:23]

---

3. Galaxy Digital’s $3.5B Data Center Expansion

[25:47–31:18]

• Deal Details:
  • Galaxy Digital is raising $3.5B in debt for Helios data center expansion (Texas); proceeds fund Phase 2, build debt reserves.
  • Helios: two buildings, eight halls, 400MW utility capacity, 260MW critical IT.
  • CoreWeave is the anchor tenant (15-year lease, with options to extend); $10.4B in minimum lease payments expected.
• Backstory:
  • Helios was originally Argo Blockchain’s asset (acquired by Galaxy for $65M in Dec 2022 during crypto sector distress).
  • Massive ROI—pivoted from distressed mining to premium AI leasing, showing how AI is transforming former mining infrastructures.

    "To me, this is one of the, you know, best M and A deals in the data center space... The return on that investment is just insane."
    —C, [29:37]

• Naming Humor:
  • Industry penchant for mythological/astronomical names is leading to name clashes (e.g., "Helios" now both AMD’s GPU racks and the data center).

    "If we're going to circular finance everything, let's also put the Helios racks inside the Helios data center and then have ChatGPT bring a Helios model and run it on those."
    —D, [31:10]

---

4. Ionic Digital: From Bankruptcy to AI Hosting

[33:13–38:54]

• Background:
  • Ionic Digital was spun out of Celsius’s bankruptcy, managing a major mining fleet (12.2 exahashes).
  • Set to direct-list on the NASDAQ (July 28th), post $400M Series A at $2B valuation.
• AI Pivot:
  • Signed a 10.5-year, $1.95B contract with Inscale (Microsoft is end-client for AI workloads).
  • Revenue recognition: $15.5M/month from contract, with $44M in Q1 2026—vastly outpacing plummeting mining revenues.
  • After mining shutdown (Dec 2025), their 234MW Texas site now hosts compute for Inscale; fixed rent rises to $182.5M/yr after ramp, with expansion options.
• Industry Implications:
  • Example of mining-to-AI transitions for distressed assets.

    "Who is Inscale's customer? Inscale's customer is Microsoft, supposedly running Microsoft's own AI... Microsoft is a great, great tenant... the m in the Mag 7."
    —D, [37:21]

• Texas as Hotspot:
  • Ward County, Texas—prime, but congested—so additional expansion is promising but not guaranteed.

    "A lot of demand there for so that 700 megawatts might be harder to achieve than not."
    —C, [38:12]

---

Notable Quotes & Memorable Moments

• "All these companies are ultimately in each other's pockets one way or the other."
  —C, [08:37]
• "The numbers stop making sense."
  —C, [09:45]
• "You're basically writing checks to be cashed in three to four to five years, which are multiples over your current gross revenue today."
  —D, [15:23]
• "Georgia is definitely a state to watch as these build outs continue."
  —C, [24:23]
• "This is one of the best M&A deals in the data center space... The return on that investment is just insane."
  —C, [29:37]
• [On asset pivots and customer solidity:]
  "Microsoft is a great, great tenant... the m in the Mag 7."
  —D, [37:21]
• [On industry naming conventions:]
  "Maybe just... put the Helios racks inside the Helios data center and then have ChatGPT bring a Helios model... That's a joke."
  —D, [31:10]
• [Meta-humor on investment scale:]
  "I think we saw more than a trillion dollars worth of billions mentioned in deals, cumulative deals today.”
  —D, [01:17]

---

Timestamps for Major Segments

• [02:03] AMD x Anthropic $5B, 2GW deal reporting, background and context
• [04:24] Market share analysis for Nvidia, AMD, custom silicon
• [07:41] Circular finance and equity relationships discussion
• [09:45] Surreal scale of investment numbers
• [11:01] OpenAI’s $750B compute spend
• [14:59] OpenAI’s revenues and losses, financial model scrutiny
• [18:46] Details on OpenAI’s Project Chelia data center, Georgia
• [20:36] Georgia Power, nuclear and gas, energy grid expansion
• [25:47] Transition: Market focus Georgia → Texas
• [27:05] Galaxy Digital’s $3.5B debt raise, Helios site details
• [33:13] Ionic Digital, bankruptcy to NASDAQ, AI contract economics
• [37:21] Microsoft's indirect involvement via Inscale
• [38:12] Texas market challenges, expansion context

---

Podcast Tone and Style

• Analytical, irreverent, and conversational
• Hosts blend deep technical info with industry lore and playful banter
• Humor surfaces in discussions of market absurdities and naming conventions

---

Takeaways for Listeners

• The AI-crypto infrastructure space is awash in headline-grabbing, eye-watering capital raises and reinvestment cycles—numbers that truly beggar belief.
• Companies like AMD are rapidly closing the gap in AI hardware, forming critical supplier alliances and leveraging next-gen product strategies.
• Power grid expansion and choice of energy source (with the South leading) is now a central battleground for AI compute growth.
• Former crypto mining assets are finding new, lucrative life by retooling for AI, as Ionic Digital and the Helios site prove.
• Industry actors are increasingly intertwined through both business and equity deals—at times, almost forming a corporate ouroboros.

---

For further in-depth articles, rapid daily recaps, and show summaries, visit Blockspace Media or subscribe to newsletter.blockspacemedia.com.
Next episode: Expect more on these megadeals as data centers, energy, and AI markets continue their relentless evolution.

Transcript

[A]

Y', all.

[B]

Welcome back to Blockspace live presented by CleanSpark.

[C]

Packed docket today, Charlie.

[B]

We're going to start with AMD and Anthropic teaming up.

[C]

A little curveball there from the little brother in the GPU arms race, but

[B]

AMD and Anthropic are teaming up for 2 gigawatts of capacity and AMD committing

[C]

a $5 billion investment to the Frontier Lab.

[B]

Following that we've got a story on OpenAI raising its compute spend guidance to 750 billion through 2030.

[C]

Also some notes on a proposed data center in Georgia by the Frontier AI Lab.

[B]

Following that we've got Galaxy Digital coming out seeking to raise 3.5 billion in senior secured notes. And at the end of the show we will cover Ionic Digital, which is set to debut on the NASDAQ on July 28th. And taking a look at its in scale AI deal from the former Bitcoin

[C]

miner turned AI infrastructure play.

[D]

That's right, Blockspace goes live every weekday at 1pm Eastern featuring quick hits on AI, data centers, emerging tech and markets. If you like what you hear, you'll love the full breadth and scope of everything Block Space. Go to Blockspace Media for our website to get articles, news, more podcasts and anything Block Space related. We're not just a daily livestream. You can get the daily livestream summary and other news directly to your inbox every single day. Go to newsletter.blockspacemedia.com for links to sign up. This show is brought to you by CleanSpark. Nasdaq listed ticker CLSK more on CleanSpark later on in the show. Colin the billions are pouring in. I cannot believe it. I think we saw more than a trillion dollars worth of billions mentioned in deals. Cumulative deals. Today. I'll let you kick off with the AMD anthropic story.

[C]

Yeah, so interesting story to come across the desk. We don't get much on amd. As I said, they are the little brother in the GPU arms race.

[B]

But punching above their weight perhaps AMD commits up to 5 billion to anthropic

[C]

and 2 gigawatt AI infrastructure deal. This is coming to y' all fresh from Block Space this morning. I'm gonna go ahead and give y' all a TLDR collapse. This article here, the TLDR for this one.

[B]

As the headline reads they are partnering

[C]

for a 2 gigawatt data center or 2 gigawatts of data center build out. Excuse me, of AMD's MI 44450 series GPUs and and and AMD is committing a 5 billion dollar in future investments tied to the deal. More on circular financing, kind of similar to what we covered on the show yesterday. Charlie, can you collapse this for me? The just take the go back to dual view here.

[B]

So a little bit of contacts, a

[C]

little more context with this. This rollout of the Mi 450 is AMD's next gen design, with the first gigawatt deployment slated to begin in the second half of 2026.

[B]

And this is coming off of a

[C]

number of deals that AMD has struck with other leading AI companies. Just to go through a few of them here. AMD has a 6 gigawatt multi generational deal with OpenAI that was inked in October 2025 and with Meta.

[B]

So AMD starting to move in a

[C]

little bit on Nvidia's territory.

[B]

But that being said, they've got a

[C]

long way to go before they close the gap. So I was curious, I was doing the research for this one, Charlie. It's like what is the actual market share?

[B]

Now? There's no tracking of this.

[C]

That's clean, right? Because there's no single source of data and the companies are not reporting this.

[B]

But looking at a few third party

[C]

estimates here, Nvidia's data center revenue of 75.2 billion in the fiscal year Q1 2027 that was the quarter that ended or ended at the end of March

[B]

represented 87.4% of the combined 86 billion

[C]

in data center revenues reported by Nvidia, AMD and Intel for that quarter. So this broadly puts their market share at somewhere between 80, 88%. And AMD is estimated at roughly 5 to 8% of the market. And AMD's Instinct GPU line generated an estimated 7 to 8 billion in 2025, capturing roughly 5 to 7% market share. One estimate puts AMD at between 6 to 8%, so squarely somewhere between 5 and 8% for 2026.

[B]

The rest is made up from custom silicon.

[C]

And this is the fastest growing tracked within this sector of AI. Compute hyperscaler, custom silicon such as Google's TPU, AWS's Trainium, Microsoft's Maya and Meta's MTIA collectively sits near 15 to 20% of the total chunk of compute for AI. And custom. ASIC shipments are growing at about 44.6% year over year versus 16.1% for merchant GPUs according to TrendForce data from May 2026.

[D]

Yeah, real video gamers and PC builders will probably not differentiate hugely between Nvidia Nvidia and AMD because they, especially in late years, AMD has had pretty competitive GPU models. I'm a fan of running certain AMD architecture, but that's for the consumer market. I kind of had to do a little digging on specifically this particular GPU class and the kind of pods they go in. The Mi 450 class is kind of interesting because it takes a memory forward approach. So it's got basically more memory per GPU and per cluster than Nvidia, which is really interesting because the bottleneck right now for frontier training and long context inference has shifted from RAW compute to kind of a context model. Look at the latest Kimi model with a trillion or a billion. However, a lot of parameters, the parameters on all these models is growing as context becomes a problem that the frontier models realize they can just. If you, if you can get RAW compute, the next problem is going to be like raw large context. So this is a strategic, this is a pretty like strategic and fourth and fourth thinking move by AMD and within, with, with AMD investing in Anthropic, this means that the Mi 450 class has three anchor tenants. OpenAI, Meta and now Anthropic. And then I'll also call back and, and mention that OpenAI invested. That OpenAI invested in AMD or more specifically AMD issued OpenAI warrant for up to 160 million shares of AMD common stock. So OpenAI got equity and now AMD gets anthropic equity. Is this, does this mean OpenAI actually by proxy kind of owns part of Anthropic? Because like the joke has been that it's one big circle. The money keeps moving to circle, but now does the equity move directly in a circle where the competitors now have proxy ownership to each other? It's just too, too weird to me. Maybe I'm missing something here, but there's, you know, there's some, there's some funny kind of insight on that. Yeah.

[C]

I mean, it just goes to show that all these companies ultimately are in each other's pockets one way or the other. Right.

[B]

And it makes sense because there are

[C]

so few of them that can actually move its size. Right. And have the, the financial firepower to deploy these things.

[D]

Yeah. And then I'll also note that of this class of GPU for OpenAI, the OpenAI gets these, these GPU clusters first and then Anthropic will get them sometime later. Second half of 2026 is the OpenAI timeline. Anthropic gets them in the first half of 2027 and this is also anthropic. They haven't run their models on AMD GPUs before, so that's kind of a novel thing. So that's about as much insight I have to this deal casually throwing around numbers such as $5 billion now for financing and equity investment. So

[C]

like we said before, the numbers stop making sense.

[D]

Yes. And we've got some more stories. The big one with the biggest numbers OpenAI we'll go to next after a word from our sponsor, CleanSpark.

[A]

We are CleanSpark, America's Bitcoin miner, a publicly traded company with the largest operating hash rate powered entirely by self operated infrastructure across four states. This is our proof of work and we are setting the standard for what's next. Learn more about the intersection of energy and [e-mail]

[E]

if Bitcoin's actually the best money and it's the thing that people should accumulate and it's the best risk adjusted asset, I lose zero sleep about whether or not that's gonna happen. I just ask the question of when

[B]

is literally matrix math that you're running on large pieces of data.

[D]

The Bitcoin miners can absorb that energy and in many ways this feels like

[C]

a second bite at the apple to build a new Internet.

[D]

Okay, this next story is OpenAI the number gets even bigger Bigger. So the headline that we put out this morning, OpenAI lifts planned compute spending to $750 billion through 2030. This is a scoop from the Wall Street Journal and the highlights are as follows. The Wall Street Journal reported that originally the estimate for OpenAI to spend on compute through 2030 was $600 billion earlier this year. That has now been raised officially to about $750 billion spent on compute by 2030. This covers the projected compute expenditure rather than OpenAI's valuation. This is not the same thing as the 1.4 trillion in infrastructure commitments previously that CEO Sam Altman had mentioned, and different than the $1.15 trillion estimated covering hardware and cloud agreements through 2035. This means OpenAI has assembled agreements including with pretty much everyone, Microsoft, Oracle, aws, Core Weave and its hardware partners now include Nvidia, AMD and Broadcom. This is also on the heels of earlier this year in February, OpenAI announcing that they had closed a round valuing them somewhere a little over 750 billion. Again, that's the valuation and the rounded close with 122 billion of committed capital. I'm sorry, correction. At an 852 billion dollar post money valuation. And we've got some numbers on quarterly and monthly revenues from OpenAI. But I have some numbers that I can dig in there if you want. I'll toss it to you, Colin, for some comments on this story.

[B]

Yeah, so I don't know if our

[C]

wires are getting crossed here if there were two $1.4 trillion calls from Sam Altman, but the Wall Street Journal does

[B]

actually say, quoting directly here, after Altman's

[C]

proclamation that OpenAI plan to spend 1.4 trillion on compute capacity set off concerns over the company's financial stability, the CFO walked back that claim quietly reassuring investors that OpenAI plan to spend 600 billion through 2030.

[B]

You just imagine, it seems like no

[C]

matter who the CEO is, the CFO is always having to pick up the bag. Do you imagine your CEO going on some sort of show and saying, we're going to spend 1.4 trillion through 2030? And the CFO is like, where did you get that number? Nothing in our sheet says that nothing in our model says that we can do that reliably. So they actually had to walk it back. But you're almost halfway there with this upsize to 750 million.

[B]

And I'm going to be honest, I

[C]

mean, it's a huge number. I mean, 750 billion, it's a huge number, don't get me wrong.

[B]

But when I was looking at through 2030, this is why you become so

[C]

desensitized to the numbers, I was like,

[B]

huh, only 750 million over the, over

[D]

the next four years. Really?

[B]

But it's a massive spin and it goes back into what we've been saying

[C]

about most of the compute capacity constraint angles. Right.

[B]

I think it's obvious enough right now

[C]

that compute is still incredibly constrained.

[D]

Yeah.

[B]

We keep seeing these posts about rental prices for Blackwells and for H2 hundreds and H1 hundreds and even A1 hundreds are up throughout the summer.

[C]

Right. Because they're just like, quite frankly, not enough to go around. A few other things to highlight on this.

[B]

I think you said you were going

[C]

to go into revenue numbers, right?

[D]

Yep.

[C]

Yeah. So maybe we will just go ahead and go into that because that was going to be my next follow up was just looking at exactly what they're pulling in currently according to their most recent valuation.

[D]

Yeah. So we have the numbers that OpenAI says they're going to spend and then we also have the numbers that OpenAI has booked in revenue. Now the revenue sounds pretty spectacular for like any company, tech companies, especially new tech companies in 2020. 5 Their GAAP. According to GAAP accounting standards, revenue was 13.1 billion, which would put but now they're saying that the run rate at the end of 2025 was about 20 to 21 billion, or around $2 billion per month. That sounds pretty incredible. However, how do you square the fact that you're only making 25 billion a year and in revenue and you have 80 to 100 billion in annual compute commitments coming to over the next five years? And on top of that, they're not profitable today. For every dollar they make, they lose about 1.$2. And so they're burning around 27 billion per year this year and roughly 63 billion in 2027. So while I totally understand the incredible crazy growth metrics, but the reality is you're basically writing checks to be cashed in three to four to five years, which are multiples over your current gross revenue today. So you're betting that you can cash those checks?

[C]

Colin well, I'm betting on a windfall from either equity, sales or financing once they go public. But if we look at the growth rate right now of the run rate of 21 billion holds up, that's 61% growth from 2025.

[B]

Now obviously that's going to compress as

[C]

the company matures, but if you assume that growth rate, then they will potentially net again, this is assuming the 61% growth rate, they'll net roughly 34 billion in 2027.

[B]

But to your point, Charlie, that's not

[C]

enough to close the gap between that's not nearly enough to cover the more than 150 million that they plan to spend per year, or roughly 150 million they plan to spend per year going forward through 2030 on compute alone, not to mention infrastructure if they decide to build their own sites. You know, that's the other part of this that I actually wanted to flag. The interesting bit about this is the Wall Street Journal article also outlined a data center that they are planning in Georgia. And specifically this is kind of they

[B]

almost framed it as they almost framed

[C]

it as a not pivot away from but OpenAI kind of taking the reins for themselves on data center development after there are questions about whether or not Stargate is actually going to get online. So that was the 5 gigawatt data center that was announced in January 2025. SoftBank was supposed to lead financing for it and OpenAI was going to be involved in it, along with a host of other names. I haven't seen any updates on that recently and I'm not Even totally sure what the plan for that is right now.

[B]

But OpenAI is, according to this Wall

[C]

Street Journal report, looking to build this data center in Georgia. And I am struggling to find my notes here.

[D]

I've got it here. A $20 billion data center called Project Chelia in Effingham County, Georgia. OpenAI's VP of compute strategy, Sakin Kati, said the the company has contracted with the utility Georgia power to receive 3.2 gigawatts of power between 2028 and 2032. So we can get into that. This is, this is, this is really, really big.

[B]

And I think probably the most interesting

[C]

part about this to me at least again going back to the idea of OpenAI trying to take the reins for development for the Wall Street Journal article, the project quoting directly from it.

[B]

The project represents the first site in

[C]

which OpenAI is the lead designer and developer. At its other sites, OpenAI rents chips from cloud providers such as Oracle and AWS.

[B]

And critically, OpenAI has hired Brent Mayo, one of the architects of Elon Musk's

[C]

data center build out the Colossus data center in Memphis, according to People with Knowledge with the matter quoting further from

[B]

Wall Street Journal, Mayo, who left Musk's

[C]

XAI earlier this year, played a key role in helping that company build his first Colossus supercomputer facility in Memphis.

[B]

Which they stood up and what Was it like 122 days? It was some absurd, I mean it was like breakneck speed.

[C]

It was almost within a quarter. And that to me is one of the more interesting bits about this which shows that they're very serious about getting this done timely and getting it done right. You know, like they really seem to be taking, taking, they really seem to be taking the asset heavy model seriously here for themselves in the sense they want to own this infrastructure and they want to have one of the premier architects for these data centers to lead it for them.

[D]

Yeah, a little insight on Georgia Power because I think if bitcoin miners are a little bit familiar because Georgia has been popular as bitcoin miners of the past five years have used the abundant nuclear power from that utility over the past years. It's a clean spark stronghold, over half a gigawatt I believe in the area. But Georgia Power is the utility with the, with the most recent new nuclear reactors at the plant. Vogtle is the name of that. And they brought on a couple nuclear reactors recently at 1.1 gigawatts each. First newly built US reactors in a generation. There's a whole Issue of them being expensive and over budget and delayed. But the interesting thing is, I was like trying to look into where is the power coming from. Is Georgia Power adding more nuclear reactors? The answer is no. The new power from Georgia Power is natural gas. This is the case with everybody. It's mainly natural gas. They do anticipate having some solar and batteries, but the lion's share of it would be natural gas. Equivalent to basically 10 new nuclear reactors if assuming a nuclear reactor is roughly 1.1 gigawatts. This is reported from by Data Center Dynamics this past winter. Kind of diving into the dynamics and plans for the Georgia power grid. Another interesting thing is the Georgia Data center pipeline, kind of like the ERCOT Q, has jumped from 1.6 gigawatts in 2021 to 19 gigawatts in 2025, projecting 8,500 megawatts of load growth and a 2,600 megawatt jump in peak demand by 2030. So regulators in the area have approved a $15 billion plan to raise capacity 50% over six years. I can see with Georgia being a pro energy state here, at least from what I can see why OpenAI would want to go there. Seems like you have a lot of regulatory blessing. And we've talked a lot about Texas being the great place for energy and building data centers. Georgia's right up there. It looks like they want to. It looks like they want to build out their corner of the power grid and generate a lot of natural gas, which is great for data centers.

[B]

Yeah.

[C]

And clearly a very progressive view on how to build out more generation. I mean, the Vogel, the Vogdal nuclear plants were the first nuclear plants to come online, as you said, Charlie, and I want to say at least a decade, maybe more. They were the only one being constructed at the time, and right now, the only one that is currently in the queue. Who knows if it'll actually get online because it's a small modular reactor, is in Oak Ridge, Tennessee. But then that gas angle makes a

[B]

lot of sense to me.

[C]

Like you said, nuclear is very expensive to build.

[B]

It also takes a long time because it is choked in red tape because

[C]

no one wants to have another Three Mile island or Chernobyl or Fukushima. Right.

[B]

So you can understand, I'm sure there.

[C]

There are a lot of nuclear heads that say we could pare down some of that regulation. I'm sure that's the case, like with anything else. But if you're talking about speed to

[B]

power and also not pissing off the environmentalists too much, then that gas is

[C]

your only answer and it's going to be the thing that is a stopgap for baseload for these data centers states over. And it's not. It is really no surprise that honestly

[B]

in a lot of regards the south

[C]

is dominating in this respect because this is very pro business friendly and it's they typically don't want too many regulations outside of the ones that will keep people safe and they want people to come build Midwest and Rust Belt. Definitely pulling a lot of weight here as well given the fact that that used to be the industrial heartland. There are a lot of opportunities for brownfield sites up there too. But not too shocking for me to see Georgia position itself as a premier data center market going forward. You basically have Texas in the southwest, Georgia in the southeast leading the charge. Virginia, it seems tapped out. It's already like 25% of their energy goes to data centers. I don't really know what they're doing to build any more capacity there, but Georgia definitely a state to watch as these build outs continue.

[D]

Yeah, we'll see. We have effectively infinite natural gas. It's all domestic and it's really just a matter of getting at it from point A to point B and in the right processed form. I've seen a lot of people say things about natural gas being squeezed, but I think those people are wrong. Natural gas, we have infinite natural gas. More. We have more natural gas than the then previous Fed chair has dollars to print, if you're familiar with that meme. We have.

[B]

We have more natural gas than your entire country.

[D]

Exactly. We'll move on. We're going to go talk about data centers. We're going to go from Georgia to Texas. Galaxy secures $3.5 billion in senior notes. Before that, a word from our sponsor, Luxor Foreign.

[B]

This episode of Blockspace Live is brought to you by Luxors Commander Bitcoin miner management software for enterprise operations. Luxors Commander gives you real time fleet monitoring, bulk remote commands across your fleet and Intelligent miner. 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. Ercot backtest showed 10% improved profitability with intelligent mining versus binary mining. Commander Pro is $100 per megawatt or a 25 basis point pool fee adder. But you can also try it for

[C]

free for 60 days.

[B]

So if you'd like to learn more,

[C]

go to Luxor Tech Forward slash Commander to get started.

[B]

All right Charlie, let's go ahead and hop on over to Galaxy Digital. Galaxy Digital plans to raise 3.5 billion, 3.507 billion debt offering for Helios data center expansion. So there's not too much about this

[C]

because the note itself has not been closed. So we don't have all of the details, but the TLDR here is that this will go towards their Helios facility. 3.5 billion in senior secured notes due 2031 in a private transaction. They will use the net proceeds to fund part of Helios Phase 2 in Dickens county and build debt service reserves.

[B]

The project consists of two buildings with

[C]

eight data center halls offering a total of 400 megawatts of utility capacity and 260 megawatts of critical IT capacity.

[B]

Now, Coreweave is the tenant for the

[C]

whole of this Helios site. They have a 15 year lease with

[B]

two 5 year extension options.

[C]

And Galaxy is targeting minimum rent commencement in the second quarter of 2027

[A]

for

[C]

the full second build out, while its presentation lists 10.4 billion in minimum contract leases payments over the initial 15 year term of the lease.

[B]

The thing that's most interesting about this

[C]

to me, Charlie, is old heads will

[B]

remember that Helios was never a Galaxy site. Originally it was Argo Blockchains, which a

[D]

name many have forgotten by now, cast

[C]

into the annals of history and the dust of memory. Yes.

[B]

Aria blockchain.

[C]

I don't think that they.

[B]

I don't. They're still public, apparently.

[D]

They still public. I think they got delisted. I mean, that shows you kind of where the things are. Yeah, market cap of 46 million.

[B]

They're on NASDAQ Capital market.

[C]

So I guess they were delisted from the major exchange.

[B]

But Galaxy purchased Helios. I need to fact check myself here. I believe it was 65 million. Yes, for 65 million. In December 2022, when Argo is distressed

[C]

and looking for a way to restructure the company, sell assets that it could no longer manage.

[B]

Galaxy was originally being hosted at the

[C]

site, then they just purchased it. 65 million. Now it is an AI site that is commanding a $10.4 billion 15 year deal with Core Weave.

[B]

This to me is one of the. Obviously it was not suited for AI when they bought it.

[C]

They've had to invest a lot of capital in order to get that thing revamped.

[B]

But to me, this is one of

[C]

the, you know, best M and A deals in the data center space. As we as in terms of. For the AI landscape as we understand it today. I mean, the return on that Investment is just insane when you look at the fact that they got it for pretty cheap to begin with and then they made the pivot at the right time and got a premier deal with a Neo, a deal with a premier Neo cloud like Core Weave.

[B]

So I just think the history of

[C]

the site's really interesting and you know,

[B]

in terms of the financing itself, we'll

[C]

cover it once we have more, once

[B]

we have more information on the terms

[C]

when they close it, and that'll probably be within the next week or so.

[D]

Yeah, we're, we're already starting to run into name clashing here because as we reported in our first story, AMD's Mi 4550X class GPUs, when you put them in a cluster and a big old tower, you call that a Helios tower. So I mean like at some point you're like, okay, everyone's using astronomical nomenclature for their AI stuff. You have Soltera Luna from GPT. Oh my gosh. Maybe just if we're going to circular finance everything, let's just also put the Helios racks inside the Helios data center and then have ChatGPT bring a Helios model and run it on those. That's a joke. This is me spitballing, but I'm just remarking on a lot of the name conflicting here.

[C]

Yeah, I think we need more Greco Roman mythology. Helios being one of the original titans and the personification of the sun. We need a cooling company that goes by Poseidon or Neptune or something.

[D]

Yeah. And then we could always. And then, and you could have Sam Altman have his own personal model called Narcissus.

[B]

What does that make Dario?

[C]

Like what, who's, who's the satyr in Hercules? Is it.

[B]

It's not Hermit, is it Hermes?

[D]

Hermes who trains him? Hermes, the winged messenger.

[B]

Right. He's the winged God.

[C]

But who's the little goat man played by Danny DeVito? And Hercules.

[D]

Oh my goodness. I don't know. We're gonna.

[C]

Maybe I could get in a little bit of trouble here.

[B]

Mario can, can be the guy who

[C]

trains Hercules and that would make Hercules the US government.

[B]

Who knows?

[D]

On that note, we're going to move on from Greek mythology to Ionic, but not before word from our sponsor Lygo.

[B]

Bitcoin is making a recovery right now, but that doesn't mean that companies that have been playing funny games with your bitcoin are out of the woods. There's a long history of bitcoin backed lenders misallocating and misappropriating user funds and that has left many people with catastrophic

[C]

losses as they've used these services.

[B]

Do yourself a favor. If you're looking at a Bitcoin backed lender, do not become the next Celsius or FTX victim. Check out Lygos Finance. They are our preferred Bitcoin backed lender

[C]

here at Block Space.

[B]

They use Bitcoin native smart contracts to make sure you are always in control of your keys and no one else touches them while you are using them

[C]

for your lending service.

[B]

That's right, with Lygos Finance you always hold your keys. No one else. There's no rehypothecation, no bridging, no wrapping, no funny business, just competitive rates as low as 8.5 to 10% APR. Go to Lygos Finance to learn more. All right Charlie, let's close on another another company that was the product of distressed assets and bankruptcies. Ionic Digital and specifically Ionic Digital forecasts up to 195 million in 2026 revenue ahead of its NASDAQ debut. So some context really quickly. Ionic Digital was born out of Celsius's bankruptcy. Actually it was the entity formed to manage Celsius's mining arm their mining fleet

[C]

because they had a quite a sizable bitcoin mining fleet. In fact ionic still has 12.2 exahashes

[B]

under management and now they are going to debut on the nasdaq. This has been a long time coming. Believe they tried to make a run

[C]

at public markets in either 2024 or

[B]

2025 but now they will be doing

[C]

a direct listing on the NASDAQ as of on July 28th.

[B]

Now what the direct listing here means is that this is not an ipo.

[C]

So there's not going to be a

[B]

raise associated with this. But they will start trading on the

[C]

Nasdaq as of July 28th.

[B]

And this comes after a June 2026 $400 million Series A raise at $53

[C]

per share for an implied market value of 2 billion. Now here's where it gets interesting.

[B]

They already have an AI tenant for their sites. They have inked a deal with Inscale 126 month contract valued at 1.95 billion. And they received their first prepayment for

[C]

this contract in November of 2025.

[B]

What's interesting about this is the way that they recognize the revenue for this is they won't actually start drawing any more revenue until the site goes live

[C]

and N scale is delivered. The compute that they are hosting or N scale is delivering the services for the Compute, they're hosting at the site.

[B]

So that's coming in August 2026. They won't draw any more revenue until then. That being said, the way the accounting structure works is they're basically recognizing the

[C]

full value of the contract every month. So roughly $15.5 million per month. They're logging that as revenue because they're taking the full contract value and then splitting it up across 126 months and just recognizing a piece every single month.

[B]

So they've already logged, as a result of that accounting, 44 million in digital infrastructure leasing revenue in Q1 2026, while mining revenue was just 7.4 million, down

[C]

82% year over year from 41.1 million.

[B]

This makes total sense though because the

[C]

facility that is hosting N Scale was decommissioned for mining in December of 2025. And that site is 234 megawatts of total capacity. And after the initial 18 month ramp up period for this N Scale lease, the annual fixed rent at that facility at full capacity is 182.5 million per year.

[B]

There's also an extension option for 89 megawatts that currently Ionic Digital doesn't have secured.

[C]

They have to go out and actually get that capacity.

[B]

But There is a $250.5 million per

[C]

year potential if that additional 89 megawatt option is exercised by N scale. There's also a 3% annual escalator starting after year five.

[B]

Ionic is also targeting a future expansion

[C]

of 700 megawatts at the site. There's no contractual obligation for that though, but N Scale has the right of first refusal on it. But again, that's going to have to be a matter of execution because they have to actually get that capacity from the grid.

[D]

Yeah. So I don't know if you mentioned it, but who is Inscale's customer? Inscale's customer is Microsoft, supposedly running Microsoft's own AI. So as you know, a lot of the story is where does the buck stop? Who's providing the credit? Who's like the most reliable for paying the bills? Microsoft is a great, great tenant. The thing is, they are a tenant. You know, if Microsoft does pull some or part of its AI deal within scale, then you have a problem. But while they don't Microsoft. That's great. They have the money, they can write the checks. They're you know, the m in the Mag 7. So yeah, love. Sounds like a pretty good deal to me.

[C]

I mean a great deal for Ionic. And one more housekeeping thing, this is in Ward County, Texas. So we'll see one of the better places to build, obviously, but a lot of demand there for so that 700 megawatts might be harder to achieve than not. And they plan to roll out like 100,000 GPUs at this site. And again, the initial deadline for commencing lease payments on this site is August 2026.

[B]

It wasn't clear to me based on

[C]

the SEC filing how much capacity that is. I would be shocked if it were the whole 234megawatts because that would the turnaround time for that would be crazy.

[D]

Yeah. All right. That's all our news for the day. If you like this show, you'll Love the newsletter. Newsletter.blockspacemedia.com and check out the other content available on our website. A lot of written content, news stories that we don't even touch on here. You can find those on our website at BlockSpace Media. Make sure to like and subscribe. Drop a review 5 out of 5 stars. I know you want to give us a 5 out of 5 star rating. This show is brought to you by Clean Spark. NASDAQ listed ticker clsk. I'm Charlie.

[C]

I'm Colin.

[D]

We'll see you tomorrow.