Aaron Cowen

1. Key Themes

Power Infrastructure as the Defining Constraint of the AI Era

Dan Dreyfus argued that the AI buildout is not primarily a compute story — it is a power story. He framed data centers as refineries where electricity is the crude oil input, and argued the supply chain of critical minerals, construction timelines, and regulatory approvals make a multi-decade shortage nearly inevitable even without AI demand.

"We do not need AI demand to keep the power markets incredibly tight for the next 20 years. AI demand just turbocharges. That's all it does. And it creates shortages." 00:14:31

"Jensen was just recently quoted that we need a thousand times more power than we currently have. Now, if that's remotely true, we need every single source of power that you can imagine." 00:18:09

Buying Hard Assets Below Replacement Cost in a Capacity-Constrained World

Dreyfus invoked Sam Zell's framework — buy hard assets below replacement cost when new capacity will be needed, hold, then sell at a premium — as the core thesis for Talon Energy.

"If you can buy an asset, a hard asset at below replacement cost for an asset that's going to be needed in the future, where we're going to need to build new capacity of that asset, then you buy that asset at the discount to replacement cost, you hold it and you sell it at a big premium to replacement cost when the market wakes up." 00:15:29

"Today in the stock market, as a good speculation, you could purchase this company at a 25 billion dollar enterprise value. The replacement cost is 45 billion and because they've got debt, it means that the equity value just to get to replacement cost is more than a double from where it's trading today." 00:15:57

Hidden Option Value in Geographically Mispriced Casino Assets

Aaron Cowen's MGM thesis rested not on Vegas but on a Japan casino license opening in 2030 and a Dubai property with embedded gambling optionality — both almost completely ignored by the market. His insight: the market historically begins to price in a new casino opening about three years before it opens, and that window is now.

"What they have now is a license to open a casino in Osaka, Japan... If you go to the company slide presentations, they sort of mentioned this, but they're not really talking about it." 00:05:29

"When you take the Vegas assets, which we think are worth about 60, when you take Japan, which we think is worth about 50 bucks, if Dubai happens, that's worth another 40 or $50. So we think the stock is a triple." 00:08:36

Decentralized Crypto Networks as Infrastructure Bootstrapping Mechanisms

Kyle Simani argued that GeoNet's crypto-incentive model — paying hobbyists in tokens to deploy base stations on their rooftops — allowed it to build a 22,000-node global RTK network in four years, surpassing incumbents like Trimble, Hexagon, and Topcon that took 20–30 years to build ~12,000 combined stations.

"Those base stations are being deployed by any random guy or hobbyist or professional or small business owner who wants to make some extra money. You can go on GeoNet's website today, you can buy one of these base stations, they're a few hundred bucks, you put it on your roof of your house or your small business, it broadcasts radio waves, you make money." 00:42:40

"The GeoNet network is taking of that 11 million in revenue, roughly 80% of it is being used to make open market purchases of Geo tokens... That means $8.8 million right now per year is going into buying GeoNet tokens on the open market." 00:46:24

Radiopharmaceuticals as a Structurally Moated Oncology Platform

Oleg Nodelman made the case that radiopharmaceuticals are uniquely defensible in biotech — generics don't traffic in them, and because they require specific radioisotopes unavailable in China due to different historical nuclear programs, the China biosimilar risk that has depressed much of biotech does not apply here.

"Unlike most of biotech, there's a real moat. And now... generics generally don't traffic in radiopharma, and because the class involves radioisotopes, it's off limits to China." 00:33:37

Casino Loyalty Databases as the Real Competitive Moat in Gaming Expansion

Cowen shared a personal loss from the Rio investment — when Caesars separated the asset from its parent, the loyalty program went with it, and the standalone property lost its competitive engine. This reframes loyalty data as the hidden infrastructure of casino value.

"I did an investment with a couple of friends and we were buying the thing at $200 per square foot. The thing we forgot was when you separated from Caesars, you lost the loyalty program. And that ended up pretty bad investment." 00:10:42

Physical AI Tailwinds Are Creating a New Natural Monopoly in Precision Location

Simani's argument was that GeoNet is building something structurally similar to a telecom — a natural monopoly with network effects, lowest cost structure due to decentralization, and a physical AI wave (robotics, AVs, drones, precision ag) that creates sustained demand.

"What I love about GeoNet is it's a very obvious network effects business. This thing looks like a natural telecom. You have base stations all over the world, you've got to cover the whole planet. Telecoms naturally form monopolies. Historically, I think the same is likely to be true here." 00:45:33

GLP-1 Drugs as the Most Underappreciated Longevity Intervention Available Today

In a brief but significant aside, Nodelman argued that GLP-1 obesity drugs are already one of the best longevity drugs available — because caloric restriction is the only thing empirically shown to extend life, and these drugs mechanically replicate that effect.

"One of the best longevity drugs out there... folks don't even realize it in the GLP-1s and the obesity drugs. So one of the only things that's ever been shown in actual actual data to extend life is caloric restriction. And that's literally what all the obesity drugs do." 00:37:52

2. Contrarian Perspectives

The Osaka Casino Is Worth More Than MGM's Entire Current Market Cap — and Nobody Is Talking About It

Cowen argued that Japan's gambling market at $40 billion (vs. Vegas at $10 billion and Macau at $30 billion) means the Osaka casino could generate ~$2 billion EBITDA at full operation, worth ~$50/share against a ~$48 bid price. The market is ignoring this entirely because it doesn't open until 2030 — but historically, casino markets price in openings three years before launch, which means now is the entry point.

"Japan actually has a reasonably large gambling market. They have pachinko parlors, and they have horses. That's about a $40 billion market. If you look at the market in Macau, that's 30 billion. And if you look at Vegas, it's only $10 billion. So this could be a massive opportunity." 00:05:57

"When did they start caring about Macau? They started caring about it about three years before it opened... we're almost in that timeframe, which is why we think it's opportunistically the right period of time." 00:12:00

We Don't Need AI to Justify the Power Build — AI Just Creates the Crisis Point

Most investors frame the power buildout as an AI story. Dreyfus's contrarian take is that the power market was already going to be extremely tight for two decades due to prior under-investment, deindustrialization, and the prior efficiency cycle. AI just accelerates the timeline to crisis.

"We do not need AI demand to keep the power markets incredibly tight for the next 20 years. AI demand just turbocharges. That's all it does." 00:14:31

"PJM alone needs 106 gigawatts of new power in 10 years... That's the size of what Japan consumes today for one little part of the US." 00:19:04

The China Biosimilar Risk That Has Suppressed Biotech Broadly Does Not Apply to Radiopharmaceuticals

Most biotech investors have been burned by or are pricing in Chinese replication risk following the Amgen-Sanofi Supreme Court ruling. Nodelman argued that for radiopharma, this risk is structurally absent — because the specific radioisotopes required come from US nuclear waste streams with no Chinese equivalent.

"Actis' radioisotope payload is Actinium. And Actinium is manufactured from radium-233, which was used in our own nuclear programs in the US in the 50s and 60s. So it's a waste product from there. So Actinium is not even available in other countries like China because they had a completely different program with enriched uranium and plutonium." 00:34:46

A Revenue-Share Token at 150M Fully Diluted Returning $8.8M/Year Growing 3x Is Radically Undervalued Because of Crypto Bear Sentiment

Simani's implicit contrarian claim: the asset is being priced as a speculative crypto token, but structurally it is a revenue-share instrument in a natural monopoly infrastructure business. The discount exists purely because it lives on a blockchain and crypto is out of favor — not because of business fundamentals.

"They're returning capital to shareholders. This thing is returning $8.8 million to shareholders, it's trading at $150 million valuation, and it's going to grow 3x this year. It's an unbelievably cheap asset. It's just people aren't paying attention because it's crypto bear market right now." 00:50:15

3. Companies Identified

MGM Resorts International

Description: Major casino and hospitality company, operator of 13 Las Vegas properties and international assets. Why mentioned: Aaron Cowen's primary pitch — cited for hidden value in Japan Osaka casino license (opening 2030), Dubai optionality, aggressive share buybacks (half the float in 6 years), and Barry Diller's accumulation of 26% ownership with a $48/share bid. Won the bestie vote.

"When you take the Vegas assets, which we think are worth about 60, when you take Japan, which we think is worth about 50 bucks, if Dubai happens, that's worth another 40 or $50. So we think the stock is a triple." 00:08:36

Talon Energy

Description: US power producer with 2 gigawatts of nuclear and 6 gigawatts of natural gas baseload power, operating primarily in the PJM region. Why mentioned: Dan Dreyfus's primary pitch — trading at $25B enterprise value vs. $45B replacement cost, generating ~$50/share free cash flow on a high-$300s stock price, with Amazon data center contracts and potential to expand to $70-$100+/share FCF as power demand explodes.

"The stock today is sort of in the high 300s. If they just do absolutely nothing... these guys will be generating $50 a share of free cash flow per year. Again, the stock is in the high 300s. So it's about seven times free cash flow, good infrastructure assets in the US traded about 15 times." 00:21:44

Actis Oncology (ticker: AKTS)

Description: Clinical-stage radiopharmaceutical company with a mini-protein platform targeting Nectin-4 and B7H3 in cancer treatment. Why mentioned: Oleg Nodelman's primary pitch — $1B market cap, $500M enterprise value, 3+ years of cash runway, 18x oversubscribed IPO with $100M Eli Lilly anchor order, China-proof moat via Actinium isotope sourcing, clinical data expected Q1 2027 (Nectin-4) with potential $10B/$200/share outcome.

"We think Actis could be worth $10 billion or $200 per share if even one of their programs makes it to market." 00:34:32

GeoNet (token ticker: GEOD)

Description: Decentralized RTK (real-time kinematics) precision geolocation network built on Solana blockchain, with 22,000+ base stations in 150 countries. Why mentioned: Kyle Simani's primary pitch — world's largest and fastest-growing RTK network, $11M annualized revenue growing 3x YoY, 80% of revenue used to buy back tokens ($8.8M/year), customers include John Deere, DJI, TomTom, USDA-subsidized precision ag programs.

"Today, GeoNet is the world's largest RTK network in the world. And it's also the fastest growing. The three companies you see... Trimble, Hexagon, and Topcon... all of them combined have roughly 12,000 base stations deployed around the world. GeoNet was founded in 2021... and today they are roughly twice the size of the next three guys combined." 00:41:45

Eli Lilly

Description: Major pharmaceutical company best known for GLP-1 weight loss drugs. Why mentioned: Backstopped Actis Oncology's IPO with a $100M anchor order, signaling strong pharma validation of the radiopharmaceutical platform. Also referenced for GLP-1 drugs as potential longevity intervention.

"Recently went public with a 300 million-dollar IPO that was 18 times oversubscribed and backstopped with a 100 million-dollar order by Eli Lilly, the folks who bring you all the weight loss drugs." 00:31:24

Constellation Energy

Description: Nuclear power operator that owns the Three Mile Island reactor. Why mentioned: Cited as the operator that Microsoft convinced to restart Three Mile Island at $100/MWh for 20 years, as a landmark example of hyperscaler-driven power procurement desperation.

"It was Microsoft that told them they needed to start it up. And in order to incentivize... they said, power prices stay at $50 a megawatt hour. We'll pay you a hundred a year for 20 years minimum price for you guys to start this up." 00:20:50

Solana

Description: High-speed Layer 1 blockchain on which GeoNet's token trades. Why mentioned: GeoNet is built on Solana. Simani noted all GeoNet token purchases are verifiable on the Solana blockchain in real time.

"The token does trade on the Solana blockchain. If you want to buy it trades 24-7. The ticker is GEOD." 00:48:15

Multicoin Capital

Description: Crypto-native venture fund co-founded by Kyle Simani. Why mentioned: Led investment rounds in GeoNet and was an early investor in Helium. David Sacks seeded Multicoin early.

"I found that a firm, Multicoin Capital, about eight and a half years ago... I was probably most well known for leading all three rounds of investment in Solana prior to Solana's network launch in 2020." 00:40:19

Helium

Description: Decentralized wireless network (5G signal), structurally similar to GeoNet. Why mentioned: Simani led Multicoin's investment in Helium 6-7 years prior and remains a strong believer. Called out as GeoNet's closest structural analog.

"Yes, I actually led Multicoin's investment in Helium six or seven years ago, and I continue to be a very big long-term believer. They actually had big news go out this morning." 00:49:53

Wynn Resorts

Description: Luxury casino and resort operator. Why mentioned: Opening a casino in Ras Al Khaimah (UAE) in ~2 years, which Cowen argued would validate and potentially accelerate Dubai casino legalization adjacent to MGM's Dubai property.

"Two years from now, Wynn is going to open a casino in a place called Ras Al Khaimah, which is 45 minutes away from Dubai." 00:08:11

John Deere

Description: Agricultural equipment manufacturer. Why mentioned: Confirmed GeoNet customer via its Global Unmanned Spraying Systems (GUS) autonomous spraying vehicles powered by GeoNet's RTK network.

"John Deere... they have a new service that they rolled out recently called Global Unmanned Spraying Systems or GUS. These things drive around. They literally spray plants with pesticides and other things like that." 00:44:05

DJI

Description: World's largest drone manufacturer. Why mentioned: GeoNet customer across many of their drone models, sending significant traffic over the GeoNet network.

"The world's largest drone manufacturer, DJI, is a GeoNet customer. It's not in all of their models, but it is in a lot of their models." 00:45:06

TomTom

Description: Mapping and navigation technology company. Why mentioned: GeoNet customer using GeoNet data to update maps for autonomous vehicle programs; described as a supplier to "basically every AV program in the world."

"TomTom is one of GeoNet's customers. TomTom is a supplier to basically every AV program in the world, excluding maybe a couple." 00:44:33

RevMed

Description: Biotechnology company developing targeted therapies including a new drug for pancreatic cancer. Why mentioned: Briefly cited by Nodelman as one of the most promising recent examples of progress in hard-to-treat solid tumors.

"What's most promising that I think a lot of folks have probably heard of is a new drug for pancreatic cancer from a company called RevMed with just another targeted therapy." 00:39:33

Trimble, Hexagon, Topcon

Description: Legacy RTK network operators that have been building precision geolocation infrastructure for 20-30 years. Why mentioned: Cited as incumbent competition that GeoNet has already surpassed in total base station count despite being only 4 years old.

"All of them combined have roughly 12,000 base stations deployed around the world. GeoNet... today they are roughly twice the size of the next three guys combined." 00:41:45

4. People Identified

Description: Founder and portfolio manager of Suretta Capital, a $4B generalist hedge fund in New York; previously ran equities at Soros and was CIO at SAC/Steve Cohen. Why mentioned: Pitched MGM as a triple with hidden Japan and Dubai optionality; won the bestie vote. 29 years of hedge fund experience.

"I run a $4 billion firm in New York called Suretta Capital. Before founding my firm, I ran the equities business for George Soros. I was then CIO for Steve Cohen. And I've been doing hedge funds now for 29 years." 00:02:22

Barry Diller

Description: Media and internet entrepreneur (ABC, IAC, Expedia); now 26% shareholder in MGM with an active $48/share bid. Why mentioned: His accumulation of MGM shares to 80% of his NAV and formal bid for the company was cited as the key catalyst confirming the Japan/Dubai thesis and providing a floor on the stock.

"You have Barry Diller, who's a legend, aggressively buying the stock, and it's also now 80% of his NAV... What he's really doing is now trying to pick off the company to get the Japanese opportunity, which we think is worth will more than double the stock." 00:04:38

Dan Dreyfus

Description: Commodities and energy investor; pitched Talon Energy. Why mentioned: Made a compelling structural case for a decades-long US power shortage; won the audience vote with 50% of ballots.

"There's been a structural and permanent perception shift where both sides of the aisle are going to be leaning into nuclear in a big way. I'm massively optimistic." 00:02:08

Sam Zell

Description: Legendary real estate and distressed asset investor, known for Equity Office Properties. Why mentioned: Dreyfus invoked Zell's core framework — buy hard assets below replacement cost — as the conceptual foundation of his Talon pitch.

"If you can buy an asset, a hard asset at below replacement cost for an asset that's going to be needed in the future... you buy that asset at the discount to replacement cost, you hold it and you sell it at a big premium to replacement cost when the market wakes up. That's exactly what we did with equity office properties." 00:15:29

Oleg Nodelman

Description: Founder and Managing Director of EcoR1 Capital, a $2.5B San Francisco-based value-oriented biotech fund, 25 years in biotech investing. Why mentioned: Built a 10x fund since 2013 at 20% annualized returns. Pitched Actis Oncology's radiopharmaceutical platform as structurally moated biotech with China-proof defenses.

"Since inception, we've 10xed to our investors and annualized at 20%. And today we have about 2.5 billion under management." 00:28:41

Kyle Simani

Description: Co-founder of Multicoin Capital; led all three pre-launch investment rounds in Solana. Why mentioned: Pitched GeoNet as a natural monopoly in RTK geolocation infrastructure. Credited David Sacks as an early seed investor in Multicoin.

"I was probably most well known for leading all three rounds of investment in Solana prior to Solana's network launch in 2020." 00:40:19

Mike Horton

Description: CEO of GeoNet. Why mentioned: Described by Simani as a strong operator with a background in IoT and smart device infrastructure.

"The CEO's name is Mike Horton. Really, really good guy. He's been building in this kind of IoT smart device space for a while." 00:49:02

Jensen Huang

Description: CEO of Nvidia. Why mentioned: Quoted as saying the world needs 1,000x more power than currently exists — used to frame the extreme scale of the AI power demand thesis.

"Jensen was just recently quoted that we need a thousand times more power than we currently have." 00:18:09

David Einhorn

Description: Founder of Greenlight Capital, prominent hedge fund manager. Why mentioned: Chamath cited him as having publicly rejected the Amazon trillion-dollar prediction in 2015, framing it as a cautionary example of consensus being wrong on big tech theses.

"David Einhorn, who's a friend of mine, but who is totally wrong, said, I know trillion dollar companies. This is not a trillion dollar company. Wrong." 00:00:28

Gavin Baker

Description: Investor; ran a biopharmaceutical fund in the past; appears to be one of the All-In hosts/participants at this event. Why mentioned: Provided commentary on biotech's disappointing post-genome-sequencing track record; bought $200K across all four pitches in real time; invited to pitch silicon and memory supercycle next year.

"I ran that fund right after the human genome had been sequenced. And there was an expectation that the sequencing of the genome was going to lead to this explosion in therapies, personalized medicines... And I don't think broadly speaking, we've made as much progress over the last 25 years as maybe people thought in the early 2000s." 00:36:43

5. Operating Insights

The "Catalyst Path" Framework for Timing an Entry

Cowen described a discipline from his SAC days: before entering a position, explicitly map out the catalyst path — what events will force the market to recognize the value you see. He noted that Barry Diller's bid short-circuited his planned catalyst path (an investor day), but validated the thesis faster than expected.

"One of the things we focus on is catalyst path. So what was the catalyst path? The catalyst path was they would have an investor day, blah, blah, blah. Barry just showed his cards." 00:11:30

Skin-in-the-Game as a Filter for Idea Quality

Chamath articulated the Stan Druckenmiller principle as an explicit filter: if a presenter doesn't have personal capital at risk, the idea is worth less regardless of its analytical quality.

"I apply the Stan Druckenmiller school of invest and investigate. I really believe in it. Yes. If you don't have any skin in the game, you don't care." 00:54:45

Size the Position to the Liquidity, Not Just the Conviction

Chamath made the point that liquidity constraints — not conviction — should determine position sizing in small-cap and crypto assets. He estimated he could not get even $1M into GeoNet without moving the market.

"I love Kyle's idea. The problem is I could not get enough working for me where I don't even think I could get a million dollars in today. It would move the market. So I would have to probably be like 10, 20, 30,000 and then maybe start to buy into it." 00:55:13

Terminal Multiple Depends on Revenue Mix: Contracted vs. Merchant

Dreyfus gave a precise rule of thumb for infrastructure investors: the more contracted (bond-like) the cash flows, the higher the multiple (20x+); the more merchant/spot-exposed, the lower. Blend at 15x as a heuristic.

"The more contracts, the higher the multiple, the less, the lower the multiple. Use 15 times as a good rule of thumb, and you'll probably get to the right answer." 00:26:46

Customer Loyalty Programs Are Infrastructure, Not Marketing — Losing Them Destroys Asset Value

Cowen's Rio investment failure illustrates that when a casino spins off from its parent, it loses the loyalty database — which is the actual driver of customer return visits. This applies to any subscription or loyalty-dependent business being separated from its parent network.

"The thing we forgot was when you separated from Caesars, you lost the loyalty program. And that ended up pretty bad investment." 00:10:42

6. Overlooked Insights

Entertainment Yield Per Venue Is Radically Under-Monetized at MGM — and Barry Diller Knows It

David Sacks dropped a single statistic in passing that reframes the entire MGM investment thesis: a friend is making $1 million per day in incremental EBITDA on nights when a show runs at the Sphere at the Venetian. This suggests the entertainment-to-gambling revenue multiplier is enormous and currently uncaptured at MGM properties. Sacks explicitly tied this to Barry Diller's activity on MGM's board — that Diller's real thesis may not primarily be Japan/Dubai but rather unlocking this entertainment multiplier at scale across 13 Las Vegas properties.

"They're making a million bucks a day in incremental EBITDA every day that they have a show at the Sphere at the Venetian hotel, which is an unbelievable statistic, which tells you that when you have the entertainment draw, the gambling revenue just flies. And so Barry Diller, I have heard separately has been spending a lot of time on trying to reinvent the entertainment at these properties and thinks he has an idea on how to do it." [00:01:01:00]

This was never picked up or challenged by the other participants or Aaron Cowen (who admitted he didn't know the answer), yet it could represent a third, entirely separate value unlock — incremental EBITDA from a new entertainment strategy — that would further compress the timeline and magnitude of the return, entirely independently of Japan or Dubai.

GeoNet's Customer Revenue Expansion Curve Is Unusually Steep — This Is the Real Compounding Engine

Simani mentioned in passing a specific expansion metric that the other participants did not probe: GeoNet customers go from spending $60,000/year in year one to $170,000/year in year two — nearly a 3x expansion in a single year from the same customer. Combined with 5x net new customer growth last year and 80% revenue buyback, the compounding math is non-linear and materially different from how a typical SaaS or infrastructure business scales.

"The average GeoNet customer is growing their revenue with GeoNet about 3x in that second year. Obviously, then we look they're just their customers. They've signed up in the last two years. You can see they 5x their customer base last year. Those are net new customers." 00:47:23

This means GeoNet's growth is not only driven by new customer adds — existing customers are themselves compounding at ~3x annually, which at scale creates a revenue base that is far more durable and predictable than the headline "3x growth" framing suggests. No one in the room followed up on this point.