[BidClub_]
Sohn Conference Foundation · · 11 min

Jacob Rubin pitches Galaxy Digital at Sohn 2026

Jacob Rubin

YouTube
TL;DR
  • Jacob Rubin's Sohn pitch is Galaxy Digital—nominally crypto, actually a data-center story built around its Helios site in Afton, Texas. His sum-of-the-parts takes the crypto business conservatively, values balance-sheet cash and crypto, and runs a DCF only on what's contracted with CoreWeave—"that's today's stock price. Everything else is upside" on a "$29-to-$30" stock.
  • He visited Helios about six days before the pitch—flying from San Francisco through Denver to Lubbock, then driving 66 miles northeast to Afton—and calls it "a special, one-of-a-kind data center project." It has 1,500 contiguous acres with options to expand to a few thousand, an aquifer, a privately built substation under a 345 kV line, 70+ GW of renewable power, and 1,300 workers on site: "It is a beehive."
  • Power is the moat: 1.6 GW front-of-meter ERCOT-approved, half contracted to CoreWeave for 15 years at "a billion a year to us." Another 830 MW was approved before ERCOT switched to a batch system, and management says a customer will be announced by year-end; 1.8 GW more toward the 3.4 GW plan has studies done and has been submitted. In Load Zone West, thanks to the CREZ renewables boom, "we get paid for the power some of the time."
  • The mispricing exists because crypto is divisive, Galaxy only recently became US-listed after the GENIUS Act changed the rules, and cash flows are still negative on a ramping build. It missed the Russell determination date by about two weeks last year—the new determination will be announced in about 10 days. On EV per approved megawatt, ignoring the outlier, it's "75-plus percent upside just to the peers"; the gap formed "when Bitcoin rolled over last year."
  • The comp math is explicit: Hut 8's 500 gross MW deal last week added $3.2B of market cap. Applied to Galaxy's 830 MW, if that customer announcement hits in the second half, "that's $13 a share" on a "$29-to-$30" stock.
  • Even skeptics get the digital-assets business free: Mike Novogratz's franchise did over $600M of EBITDA in Q3 of last year, and market-making a volatile underlying "is valuable torque." "If you're not [a believer], this is a free call option. Don't worry about it. Hear me out on Helios."
  • Party favor: "We think Nokia is a monster winner" on optical and IP networking for data-center interconnect. New orders were 3x sales in that segment in each of the last two quarters; the Infinera deal vertically integrated two fabs and brought a strong pipeline, while Nokia provides the heft and scale to win hyperscalers. Nvidia invested $1B recently, and blending Ericsson plus optical-peer multiples gives "60% more to go."
Digest · the substance, structured for research

1. How a value investor ends up pitching data centers

  • Rubin's filter for Sohn: liquid, high-conviction, not boring, thematically relevant. The other ideas he mentioned were looksmaxing and peptides, plus a peace-and-rebuilding trade via uncovered dry-bulk shipping. Separately, he cited Kyivstar, a pure-play Ukrainian telecom at 4x EBITDA that had recently listed in the United States.
  • The breadcrumb trail: distressed unsecured Talen Energy bonds (hat tip Dave Rosen, Sohn 2023) went through bankruptcy and then rose 10x as the thesis shifted to an AWS co-located data center next to a nuclear plant. That taught the team about power and grid inadequacy, leading to ProPetro's behind-the-meter networks in Midland—and from Texas, to Galaxy's Helios. "The work never stops."
  • The macro he concedes upfront: $750B of hyperscaler capex this year, US data centers doubling, and power into them tripling—the question is how a cash-flow investor plays it.

2. Why the mispricing exists

  • His "too good to be true" checklist: crypto is emotionally divisive; crypto companies could not list in the US until the GENIUS Act changed the rules, so Galaxy had been in Canada; it missed the Russell determination date by about two weeks—the new determination is in about 10 days; cash flows are negative on a ramping asset "turning on this quarter"; and it's a multiyear build requiring patience. "When Bitcoin rolled over last year, a gap formed. That gap's the opportunity."
  • The story arc: crypto business today, data-center-dominated tomorrow, "and possibly split, but that's years away."
  • For skeptics, Novogratz's digital-assets franchise—staking, lending, trading and broker-dealer, with billions under management—is still a real business; it generated over $600M of EBITDA in Q3 last year. Rubin calls it a free call option and says the volatile underlying gives market makers "valuable torque."

3. Helios first-hand—six days before the pitch

  • Rubin flew to Lubbock and drove to Afton to kick the tires: 1,500 contiguous acres expandable to a few thousand, an aquifer needing little cleaning, and a private substation directly under a 345 kV line—"the biggest they make in Texas"—with 70+ GW of renewable power. On execution: the first data hall was delivered to CoreWeave on time and on budget; with 1,300 workers, a full parking lot and security clearance, "it is a beehive."
  • The power kicker: in Load Zone West, the CREZ drove such a renewable-power boom that the lines cannot pull it all the way away—"we get paid for the power some of the time... Go look around the country, you're not getting paid to take power."

4. The numbers and the next six months

  • In hand: 1.6 GW of front-of-meter ERCOT-approved power, half contracted to CoreWeave for 15 years at about $1B per year. Another 830 MW was approved before ERCOT changed to a batch system; the company says it will announce a customer by year-end. The remaining Phase 1 delivery to CoreWeave, the customer announcement and other near-term steps are "basically the next 6 months." Separately, 1.8 GW more is in motion toward the 3.4 GW plan, with studies done and submitted; more sites could create a platform.
  • On EV per approved MW, ignoring the outlier, the site offers 75%+ upside just to peers, though "it should trade at a premium."
  • The valuation math: conservative crypto, balance-sheet attributes, and a DCF on the CoreWeave contract equal today's price; everything else is upside. Rubin's financing flywheel is loan-to-cost at underwriting becoming loan-to-appraised-value at stabilization, freeing equity for growth capex—"look at Hut 8's 2042 6% bonds trading at par." Hut 8's 500 MW deal added $3.2B of market cap; applying that math to 830 MW, if the customer announcement arrives in the second half, is $13/share on a $29-to-$30 stock. His caveat, kept as said: "Is it risky? Absolutely... but so far, so good."

5. The eBay brick: Nokia as the value play on optical

  • Having missed optical-networking stocks that went up 6–7x, Rubin's value pick is Nokia—"this is the real brick. I got this on eBay. It doesn't even sell phones anymore." Optical and IP networking is "not only growing but accelerating." Connecting disparate data centers is expensive and creates heat and speed problems; new orders were 3x sales in that segment for the last two quarters.
  • The Infinera deal was game-changing for vertical integration of two fabs and its pipeline; Nokia supplies the "heft and scale" to win hyperscaler business. Scale-across—connecting different data centers or buildings—is underway; scale-out, connecting buildings or racks within a campus, could be next.
  • Optionality includes AI RAN and 6G, which Nvidia invested $1B recently to push; "the 2028 Olympics will be trialing 6G, I think"; European defense, for which Rubin says his team just stood up an internal effort; and operating leverage from cost takeouts as top lines inflect. Blending Ericsson for the rest of the business with optical peers for that segment gives "60% more to go. Boom."
Jacob Rubin

All right, thank you to Soane and to everybody here for supporting this great cause. I've got a lot to cover and not much time, so I am going to dive right in. A few legal disclaimers: do your own work.

My team and I were idea junkies. I could have come up here and pitched any one of 100 different ideas happily, but this is a special event, and for this event we thought really hard about what is most suitable for this occasion.

4 things came to mind. Number 1, the idea should be liquid and actionable for real funds. Number 2, high conviction—I really don't want to look stupid next week. Number 3, the flip side of conviction: if I'm so safe, it'll be boring, and I don't want anyone to fall asleep. Number 4, thematic relevance, which is increasingly vital in today's markets.

So, we came up with 3 finalists after applying that filter. In 3rd place, we thought about going into looksmaxing. That's a real term. Look it up if you haven't heard about it. It's absolutely crazy. The word of the day is peptides. Very exciting, but not quite for today.

Number 2, peace. Peace is what follows war. We've got a lot of war right now. I don't know when, but when it comes, peace means rebuilding, and rebuilding will benefit dry-bulk shipping. It is not covered, and it's absolutely amazing the kind of torque those companies will see when we have rebuilding.

1. Data Centers Become The Pick

We've also found a pure-play Ukrainian telecom called Kyivstar at 4 times EBITDA, and they just recently listed in the United States. But today I want to go with something you've never heard about. You did not hear about it just now. He never mentioned the word. I want to talk about data centers.

There's a twist. Data centers, when you're a die-hard value investor who cares about cash flow and valuation, are not easy. You see, we value investors have a tough lot. We don't have our heads stuck in the sand. We see the tidal wave.

We see $750 billion of hyperscaler CapEx this year, and that's growing. Data centers in this country are doubling. Power into the data centers is tripling. We see it, but how do we play it? Here's how we did it.

About 3 or 4 years ago, we got into distressed unsecured bonds of a power company called Talen Energy. Hat tip to Dave Rosen—he pitched it here in 2023. That's how a value person gets into something. We went through bankruptcy, and then when it reemerged, it went up 10 times.

What happened? The thesis dynamically changed to a co-located data center deal with AWS next to their nuclear plant. We started learning about power and the grid. There's not enough of it, and it's inadequate, respectively. That led us to off-grid, behind-the-meter, distributed power networks, and ProPetro out in Midland, Texas, was at the forefront of that. It's been a great one for us.

When we were in Texas, we learned of Helios data center. It is owned by Galaxy Digital. So, today I'm going to talk to you about Galaxy, really focusing on Helios.

2. Galaxy Builds Beyond Crypto

What is Galaxy? Galaxy has some balance-sheet assets—cash and crypto. Spot those. It has 2 main business lines: digital assets and a data center business. That's what we're going to go through today.

If you're a value hunter, you always have to ask, is it too good to be true? How does this exist, especially when it's the biggest economic boom in human history? It exists because when I said the word crypto, you may have felt some emotional reaction. It's very divisive. Either you're the laser-beam-eye adherent and you say the word fiat in your sleep, or you're a skeptic and think it's all nonsense.

Number 2: technical. Until somewhat recently, last year, before the GENIUS Act, crypto companies couldn't list here. So, this was in Canada. That has changed. We're now in the United States. We missed the Russell determination date by about 2 weeks last year. I would footnote that because the new determination will be announced in 10 days.

Cash flows: negative cash flows. This is a ramping data center. We don't see it yet. It's turning on this quarter. And finally, this takes patience. This is a multiyear build.

So, past, present, future. This is a story in change. What was a crypto business is now a mix of crypto and data center. In the future, this will be dominated by the data center side, and possibly split, but that's years away.

Let's spend a minute here. In the digital asset arena, what Mike Novogratz has built here is an absolute player: staking, lending, trading, broker-dealer—tentacles everywhere, billions under management. It's a real business. Even if you're a skeptic, in Q3 of last year, they did over $600 million in EBITDA in 1 quarter.

The key is that the underlying security here is volatile. And if you're a market maker in a volatile underlying, that is valuable torque, even if you're a skeptic. So, let's say you go into the 2 camps. If you've got laser-beam eyes, I don't even have to explain it; you're already buying the stock. That's great. If you're not, this is a free call option. Don't worry about it. Hear me out on Helios.

3. Helios Has A Rare Setup

This is the pitch. In fact, it's so important that a teammate and I, last week, flew from San Francisco to Denver, Denver to Lubbock. We stayed at the Cotton Court Hotel. We drove 66 miles in a northeast direction to Afton, Texas. We spent the day at the site. So, what I'm about to tell you is a firsthand account from about 6 days ago.

This is a special, one-of-a-kind data center project, and I'm going to tell you a few reasons why. First of all, 1,500 contiguous acres, with options to expand to a few thousand; an aquifer with water underground that doesn't take too much cleaning; and a key node under a 345-kilovolt line, the biggest they make in Texas, with 70-plus gigawatts of renewable power. And it's right underneath the line—you see it—with its own privately built substation.

Then you've got execution. They just delivered the first data hall to CoreWeave on time and on budget. When I visited, I wanted to see whether anybody was there. It is a beehive: 1,300 workers, general contractors, a full parking lot, security clearance, everything.

Then you've got power. They have 1.6 gigawatts of front-of-meter, ERCOT-approved power. Half of that is contracted to CoreWeave for 15 years—$1 billion a year to us. Another 830 megawatts just got approved before ERCOT changed to a batch system. The company said they will announce a customer for that 830 megawatts by the end of the year.

Finally, Load Zone West. This is key. In Load Zone West, there's something called the CREZ. Go look it up. It led to a megaboom of renewable power in the region, so much so that the lines can't pull it all the way. We get paid for the power some of the time. It's that cheap. That is special. Go look around the country; you're not getting paid to take power.

4. Helios Expands Its Lead

Like I said, they've come a long way with Helios. What's next? More approvals. The plan for this is for 3.4 gigawatts. We have 1.6 approved. The other 1.8 are in motion: studies are done, and they've been submitted.

More sites can be added, creating a platform. We will announce that customer for the 830 megawatts, and they're going to deliver the rest of Phase 1 to CoreWeave. That's all in basically the next 6 months.

So, compare it to peers, and you'll find that this site really stands out. It should trade at a premium. Instead, it trades at a major discount if you look at enterprise value over approved megawatts. Ignoring the outlier, if you go to the peers, there's 75-plus percent upside just to the peers.

So, why? When Bitcoin rolled over last year, a gap formed. That gap's the opportunity.

5. Galaxy's Valuation Leaves Upside

This is how we value it: sum of the parts. We take the crypto business conservatively, value the balance-sheet attributes, and then do a DCF on what's contracted with CoreWeave. That's today's stock price. Everything else is upside.

The 1 thing I want to mention about the math is that we did a lot of math. When you underwrite the debt on a project, you do loan-to-cost. When you deliver it and it's stabilized cash flows, it's loan-to-appraised value. If and when you do that, you pull equity out, and it funds the next phase of growth CapEx. And if you don't believe me, look at Hut 8's 2042 6% bonds trading at par. The market is really good for this.

So, is it risky? Absolutely. It's a huge project they've got to execute, but so far, so good. To recap, what they've got in hand covers the stock, and there's a ton of ways to win, with an index inclusion about 10 days away.

That was last week on site. Hut 8 did a deal for 500 gross megawatts last week. They added $3.2 billion in market cap. If you apply that math to our 830 megawatts when that customer announcement hits in the second half, that's $13 a share on a $29-to-$30 stock price.

That would be it. I've got 43 seconds left, and I asked him if I could have 30 extra. I have a party favor. I have 1 more.

6. Nokia Powers The Network Boom

You see, when I said Talon went to ProPetro, went to Galaxy, the work never stops. Connecting disparate data centers is extremely expensive. You have to solve for heat problems and speed problems. That has led to huge growth in optical networking.

But those stocks went up 6 or 7 times. I'm the value guy. I missed it. Not so fast. We think Nokia is a monster winner.

This is the real brick. I got this on eBay. It doesn't even sell phones anymore. They sold that 10 years ago. Nokia is optical and IP networking. That's going gangbusters—not only growing but accelerating.

New orders were 3 times sales in that segment for the last 2 quarters. They did this deal for Infinera last year. That deal was game-changing for the vertical integration of 2 fabs and the great pipeline. What Infinera needed was heft and scale, which Nokia provides to win hyperscaler business.

So, scale-across is what they're doing now. But there are other ways to win. Scale-out could be next. That's connecting different buildings on a campus or, within a campus, different racks. These buildings are really big.

You've got AI RAN and 6G. Nvidia invested $1 billion into this company recently to push that. The 2028 Olympics will be trialing 6G, I think. There is also European defense.

That’s cutting edge. We just stood up a team internally to go after European defense. And finally, operating leverage. They’re doing cost takeouts at the same time top lines are inflecting. If you use Ericsson for the rest of the business and the optical peers for that segment, blend it together—60% more to go. Boom. Thank you.

Jacob Rubin pitches Galaxy Digital at Sohn 2026 | BidClub