Latest / Deep Values: Value Investing Deep Dives / Peter Thiel 13F: Trading Nvidia for the power grid
Transcript
- 0:00Right now, Silicon Valley is just pouring trillions of
- 0:03dollars into AI chips, server farms, and, you know, advanced
- 0:07code. Oh, absolutely.
- 0:09The skill is unprecedented. Yeah, we are basically in the
- 0:11middle of the most explosive artificial intelligence boom in
- 0:14history. But one of the most famous
- 0:17forward thinking tech investors in the world just looked at that
- 0:20boom, dumped an enormous stockpile of NVIDIA stock and
- 0:25bought, well, a bunch of dirt, natural gas and nuclear power
- 0:28plants. It is the ultimate contrarian
- 0:30pivot. I mean, you have a fund famous
- 0:33for defining the digital frontier, right?
- 0:35Exactly. For backing the most disruptive
- 0:37software on earth suddenly taking all its chips off the
- 0:39table. They're rotating massive amounts
- 0:42of capital into the most physical, heavily regulated old
- 0:45school industries imaginable. It feels like watching a Formula
- 0:48One driver suddenly trade in their race car for like a fleet
- 0:53of city buses. Yeah, that's actually a great
- 0:55way to put it. So our mission for this deep
- 0:57dive is to figure out why we want to understand how a high
- 1:01tech venture capital philosophy is currently merging with Warren
- 1:05Buffett style value investing. Because it turns out they are
- 1:08betting everything on the physical, real world constraints
- 1:12of artificial intelligence. But before we jump into the core
- 1:15analysis, I need to share this with you to access more Advanced
- 1:19Research companion for modern intelligent investors and
- 1:23leverage our latest Deep Values agents.
- 1:25Welcome to download Deep Values APP from App Store to track
- 1:30stock market temperature. Welcome to download iOS app
- 1:33Buffett Indicator from Apple App Store or Bubble Watch from
- 1:36Google Play. Disclaimer All analysis is based
- 1:40on publicly available sources and may be subject to revisions
- 1:44or differing interpretations. All contents from this channel
- 1:47is for research and entertainment purposes only and
- 1:50does not constitute any investment or financial advice
- 1:52or service. Conduct your own due diligence
- 1:55before making investment financial decisions.
- 1:57So to really make sense of what this fund actually bought and
- 2:00sold, you can't just scan down a list of ticker symbols.
- 2:03Right, because those are just the symptoms aren't?
- 2:05They exactly you have to understand the underlying
- 2:07economic philosophy that drives their entire decision making
- 2:09process. OK, let's unpack this, because
- 2:12we really can't make sense of this sudden massive rotation
- 2:16into power generation without first looking at the pursuit of
- 2:19what the source is called, creative monopolies over perfect
- 2:23competition. Right.
- 2:24And this goes back to a foundational doctrine
- 2:27articulated in Peter Thiel's book Zero to 1.
- 2:30There is this core belief here that capitalism and perfect
- 2:33competition are actually diametrically opposed.
- 2:38Which sounds completely backwards if you've taken
- 2:40Economics One O 1. It totally does.
- 2:42In school, you are taught that perfect competition is the ideal
- 2:46state of a market. You know, it's a suffent.
- 2:48Yeah, But from an investment perspective, perfect competition
- 2:52is an absolute nightmare. Because it just completely
- 2:54destroys your profit margins. I mean, if you and 10 other
- 2:57people are selling the exact same lemonade on the same street
- 2:59corner, the only way to get customers is to drop your price.
- 3:02And you just keep dropping it. Under perfect competition, you
- 3:06have multiple participants producing entirely
- 3:08undifferentiated commodities. It becomes an endless race to
- 3:11the bottom. Everyone lowers their prices to
- 3:14steal market share until the price gets driven all the way
- 3:17down to the marginal cost of production.
- 3:19Wow, so durable economic profits are just completely wiped out.
- 3:23Exactly. You might survive, but you
- 3:26certainly do not thrive. Real value creation and long
- 3:30term capital compounding requires escaping that trap
- 3:33entirely. You have to build a creative
- 3:36monopoly, yes. So perfect competition is
- 3:38basically like being on a really crowded treadmill.
- 3:41Everyone is just exhausting themselves running as fast as
- 3:43they can just to stay in the exact same place.
- 3:46That's a perfect analogy. But building a creative monopoly
- 3:49is like stepping off that treadmill entirely to build your
- 3:51own private elevator. You go up and nobody else even
- 3:54has the key to use it. And to build that private
- 3:56elevator, you need a few structural advantages.
- 3:59You need proprietary technology, massive economies of scale,
- 4:03strong network effects, or, you know, highly defensible brand
- 4:08ecosystems. Something that acts like a Moat.
- 4:09Right. You are looking for an
- 4:11enterprise that can structurally escape competitive friction.
- 4:16You want robust pricing power that lasts for decades.
- 4:19But finding that elevator requires differentiating between
- 4:22two types of progress, doesn't it?
- 4:24It does. There's horizontal progress
- 4:26which is going from one to north.
- 4:28That is basically just globalization, taking something
- 4:30that already works and copying it everywhere.
- 4:32Like opening 1000 new coffee. Shops.
- 4:34Exactly. Then there is vertical progress
- 4:37going from zero to one that is creating something entirely new.
- 4:40And to achieve vertical progress, you have to discover
- 4:43what this philosophy calls secrets.
- 4:45Yes, secrets. These are fundamental, non
- 4:48obvious truths about market inefficiencies or structural
- 4:51bottlenecks that the rest of the consensus driven market is just
- 4:55completely ignoring. So in public equities, this
- 4:57means finding the choke point in an industrial supply chain long
- 5:01before the Wall Street analysts update their quarterly earnings
- 5:04models. Precisely, and this ties
- 5:06directly into how they view the future.
- 5:08The sources contrast Wall Street standard indefinite optimism
- 5:12with this funds definite optimism.
- 5:14Which honestly sounds a little arrogant at first glance.
- 5:17I mean, it sounds arrogant until you look at the actual mechanics
- 5:20behind it. Indefinite optimism is basically
- 5:23the philosophy behind passive index funds.
- 5:25Right where you just buy everything.
- 5:27Yeah, it assumes the market will generally drift upward over
- 5:30time, but you have no concrete plan for how or why.
- 5:35So you just scatter your capital across 500 companies and hope
- 5:38the aggregate tide lifts your boat.
- 5:40Definite optimism is the exact opposite of that.
- 5:43It really is. Yeah.
- 5:44It is the belief that the future is predictable and can be
- 5:48deliberately engineered through focused high conviction capital
- 5:52allocation. You aren't guessing what the
- 5:54future will be, you are actively funding the specific
- 5:56architecture of it. Exactly.
- 5:58And that leads right into the power law of capital.
- 6:01Which is the mathematical reality that a tiny fraction of
- 6:04individual holdings is going to generate the overwhelming
- 6:07majority of your aggregate financial returns.
- 6:09That is why a fund operating on this definite optimism framework
- 6:13completely avoids broad diversification.
- 6:15They don't want the mediocre positions.
- 6:18They prefer highly concentrated portfolios built around massive
- 6:22high conviction macro theses. Absolutely.
- 6:25So what does this all mean for the AI boom?
- 6:28Did the fund just panic during a tech bubble, or are they seeing
- 6:31one of those structural bottlenecks the rest of the
- 6:34market is completely ignoring? Well, looking at the timeline of
- 6:37their thirteen F filings, they definitely saw a bottleneck.
- 6:40Because you don't just wake up 1 morning and decide to tear your
- 6:43entire portfolio down to the studs.
- 6:46But that is exactly what they did.
- 6:47It is a truly remarkable timeline.
- 6:49We can trace this massive portfolio rotation through late
- 6:532025 and into early 2026. The first major warning sign
- 6:57flashed in the third quarter of 2025, right?
- 7:00Yes, this was the AI Semiconductor divestment.
- 7:03The fund executed a complete 100% exit from its premier
- 7:07artificial intelligence holding, NVIDIA.
- 7:09Wait. A complete exit.
- 7:10Completely, they disposed of over half a million shares.
- 7:13That was roughly $100 million. And if I'm reading the sources
- 7:17right, that position used to make up 40% of their equity
- 7:19assets. It was a massive anchor and they
- 7:21just liquidated it. Totally liquidated and in that
- 7:25same window they slashed their Tesla holdings by 76%.
- 7:30And interestingly, they also fully divested an initial stake
- 7:34they had in Vista Core, which is a power company we will
- 7:37definitely be coming back to. So what was the rationale there
- 7:40it? Was a classic bubble warning.
- 7:42The funds management cautioned that hardware valuations had
- 7:45detached from near term economic productivity.
- 7:49So they were drawing explicit analogies to the late 1990s tech
- 7:53cycle. Exactly.
- 7:54Just like when companies were laying thousands of miles of
- 7:57fiber optic cable long before there was enough Internet
- 8:00traffic to actually justify the cost.
- 8:02You know, everyone likes to say NVIDIA is selling the picks and
- 8:04shovels for the AI gold rush, but I guess this fund realized
- 8:07that once everyone has a pickaxe, you stop selling
- 8:10pickaxes. That's a great point.
- 8:11So they take all that cash from NVIDIA and Tesla.
- 8:13They park it temporarily and mega caps like Apple and
- 8:16Microsoft. But then we hit the end of 2025,
- 8:19five in early 2026, and this is the part of the filing that
- 8:22genuinely dropped my job. Total equity neutralization.
- 8:25Yes, they took the aggressive step of liquidating everything
- 8:29else. The remaining Apple, Microsoft
- 8:32and Tesla positions were completely sold off for two
- 8:35consecutive quarters, the fund reported Literally $0.00 in 13 F
- 8:40equity assets. Going to 0 equities in the
- 8:43middle of the biggest tech bull run in history is either the
- 8:47most genius move of the decade or the kind of thing that gets a
- 8:50fund managers access badge permanently deactivated.
- 8:53It takes incredible conviction. They just sat on the sidelines
- 8:56in cash and short duration equivalents.
- 8:58They kept their powder dry. They insulated themselves from
- 9:01any market multiple contractions.
- 9:03Right. And they just waited for the
- 9:04real bottle neck to emerge. Which brings us to the second
- 9:08quarter of 2026, the energy rebuild.
- 9:11So they finally come back to the market.
- 9:12They. Rebuilt the public equity book
- 9:14to over $400 million, but the composition was entirely
- 9:18different. Roughly 300 million of that
- 9:20capital was aggressively concentrated across regulated
- 9:23electric utilities, merchant power generators and advanced
- 9:26nuclear reactor development. Plus a massive anchor position
- 9:30in Amazon. They put around $120 million
- 9:33just into Amazon. What's fascinating here is the
- 9:36secret they discovered to use Thiel's terminology.
- 9:39The underlying thesis of this massive pivot is the realization
- 9:43that the primary constraint on digital intelligence has
- 9:46officially shifted. It is no longer about
- 9:48semiconductor availability. Exactly.
- 9:50Getting enough silicon chips is no longer the problem.
- 9:53The true bottleneck is the continuous physical delivery of
- 9:57base load electrical power. So the game is no longer about
- 10:00silicon. It's about electrons.
- 10:03It is Like imagine a company spending billions to build a
- 10:07massive fleet of $1,000,000 next generation sports cars.
- 10:10OK. I'm with you.
- 10:11These cars are absolute engineering marvels, but right
- 10:14as they are rolling off the assembly line, everyone looks
- 10:17around and realizes nobody remembered to pave the roads or
- 10:20build any gas stations. Oh wow, you have all this
- 10:22immense horsepower and absolutely nowhere to run it.
- 10:25That is the AI industry right now.
- 10:27The data centers are the sports cars and the electrical grid is
- 10:30the unpaved Rd. Exactly what's happening?
- 10:33Advanced data center clusters require gigawatts of
- 10:36uninterrupted 24/7 base lit electricity.
- 10:39Right. You can't run an AI training
- 10:42model on solar power that cuts out the second it gets cloudy.
- 10:45And you can't just plug a new GW data center into the local grid
- 10:49either. Municipal transmission networks
- 10:52have multi year interconnection backlogs.
- 10:54So if you want to connect a new massive load to the grid today,
- 10:58you might be waiting in a queue until what, 2030?
- 11:01Easily 2030. Furthermore, licensing
- 11:03conventional power plants takes incredibly long regulatory
- 11:07cycles, so operational existing base load generation assets have
- 11:11suddenly become the ultimate strategic economic bottleneck.
- 11:14OK, recognizing the power bottleneck is one thing, that is
- 11:17a brilliant macro thesis. But how do the specific
- 11:20companies they chose fit into a sustainable, risk averse
- 11:24investment framework? Well, this is where it gets
- 11:26really interesting. Because this is where the
- 11:28venture capital mindset of Peter Thiel suddenly crosses over with
- 11:31Warm Buffet, they start applying A strict value investing lens to
- 11:34these physical constraints. Right, when you look at classic
- 11:38value investing, the Graham and Dodd criteria refined by
- 11:41Buffett, you were looking for a few specific things.
- 11:44You want a clear, understandable business model.
- 11:47You want a wide economic Moat protecting the business from
- 11:50competitors. You want high returns on
- 11:52invested capital, substantial free cash flow, you know what
- 11:56Buffett calls owner earnings. And a robust balance sheet
- 11:59offering a clear margin of safety.
- 12:00Exactly. So let's look at VISTRA core
- 12:03ticker VST. This is a 14.1% portfolio weight
- 12:08for the fund. And as you mentioned earlier,
- 12:10they actually bought back in after selling it.
- 12:13They did. Vistra operates as an integrated
- 12:15competitive power generator and retail energy supplier.
- 12:19They have an asset fleet of approximately 44 gigawatts.
- 12:22That spans natural gas, nuclear, coal and battery storage, right?
- 12:26Yes. And they have a pending
- 12:27acquisition that will add another 5.5 gigawatts of highly
- 12:31efficient natural gas generation.
- 12:33But the real magic here is their Moat.
- 12:35Their Moat is essentially just physical reality.
- 12:37The Moat rests on physical asset scarcity and extreme regulatory
- 12:42barriers to entry. You cannot just spin up a new
- 12:45nuclear power plant in your garage like a software app.
- 12:48No, definitely not. Constructing new base load,
- 12:51nuclear or thermal generation capacity in the US requires
- 12:54years of environmental permitting, immense supply chain
- 12:57lead times for specialized parts, and massive, massive
- 13:01capital expenditure. And Vistra already operates 0
- 13:05carbon base load facilities like the Comanche Peak Nuclear
- 13:08Station. Because they already have these
- 13:10assets operational, they possess incredible pricing leverage.
- 13:14Because the big tech companies, the hyperscalers like Microsoft,
- 13:17Google and Amazon are absolutely desperate for that power to run
- 13:21their AI computing clusters. Yes.
- 13:23So they're going to Vistra and locking into long term power
- 13:26purchase agreements or PP as Vistra isn't just selling power
- 13:29day by day on a volatile open market.
- 13:31No, they are locking these desperate AI companies into
- 13:34multi decade inflation proof contracts.
- 13:37They're basically turning a fluctuating commodity into a
- 13:40guaranteed subscription like revenue stream.
- 13:43And the financials completely reflect that transformation.
- 13:46Vistra is generating incredible cash flow.
- 13:50They projected full year 2026 adjusted free cash flow of
- 13:54roughly 4 to $4.7 billion. That is staggering.
- 13:58And they are aggressively prioritizing shareholder
- 14:01returns, repurchasing hundreds of millions of dollars of their
- 14:05own common stock. Their credit quality is rising,
- 14:07too. They were upgraded to investment
- 14:09grade by both Fitch and S&P. Which lowers their borrowing
- 14:13costs and unlocks even more financial flexibility to expand.
- 14:17Exactly. So Vistra is the cash cow
- 14:19generating the raw power, but then you look at the other chunk
- 14:22of the portfolio, the regulated utilities, American Electric
- 14:25Power, DTE Energy, CMS Energy, FirstEnergy.
- 14:28These make up another 10 to 11.5% each.
- 14:31These companies don't necessarily generate all the
- 14:33power. They own the transmission lines.
- 14:35And to me this fits the Buffett model.
- 14:36Florida State IT. Really does.
- 14:38These utilities have state sanctioned geographic
- 14:40monopolies. It is legally and economically
- 14:43impossible for a competitor to come in and build a duplicate
- 14:46high voltage Interstate transmission grid right next to
- 14:49yours. It is a protected franchise
- 14:51monopoly and they sit right in the path of the AI build outlook
- 14:55at American Electric Power AEP. They operate the largest
- 14:59electricity transmission system in the US.
- 15:02AEP Ohio recently joined a federal partnership for a
- 15:05planned 10 GW AI data center complex.
- 15:08And to make that happen, they need a $4.2 billion transmission
- 15:13build out. Exactly.
- 15:14And the beauty of regulated utility rate making is that when
- 15:17they spend that $4.2 billion to build new lines, it expands
- 15:21their regulated rate base. Yes, the state utility
- 15:24commissions guarantee them a specific return on equity on
- 15:27that invested capital. It is honestly like owning the
- 15:30only toll bridge right outside the hottest nightclub in the
- 15:32world. It doesn't matter which DJ is
- 15:34playing inside. It doesn't matter which tech
- 15:35company has the best AI algorithm.
- 15:38If they want to get their data to the club, they have to pay
- 15:41your toll. That's a perfect way to look at
- 15:42it. These utility holdings offer
- 15:44defensive earnings stability, They are low beta, meaning they
- 15:48don't swing wildly with the stock market, and they offer
- 15:51deeply reliable cash generation. They function basically as bond
- 15:54proxies, but with inflation protected growth baked right
- 15:59into their brake base. Exactly.
- 16:01OK. I'm totally with you on Amazon,
- 16:03Vistra and the utilities. They fit perfectly into the Moat
- 16:07and cash flow value framework. But wait a second, what is it?
- 16:11I'm looking at the notes on X Energy in this portfolio, ticker
- 16:15XE. You cannot tell me this is a
- 16:17value play. X energy.
- 16:19They generated something like $40 million in revenue, they
- 16:23have negative operating cash flow, they are requiring massive
- 16:27capital outlays just to build fuel manufacturing facilities,
- 16:31and their near term survival basically relies on a billion
- 16:34dollars in federal funding from the Department of Energy.
- 16:36This is a cash furnace. Warren Buffett wouldn't touch
- 16:39this with a 10 foot pole. He absolutely wouldn't.
- 16:42And if you run X energy through a traditional Graham and Dodd
- 16:45value Screener, it fails catastrophically.
- 16:48It has 0 discernible margin of safety based on current cash
- 16:51generation. None at all.
- 16:53Then why on earth are they dedicating capital to it?
- 16:56Because this roughly 1% allocation is the classic
- 16:58thillion venture play. This is the zero to 1 bet hiding
- 17:01inside a portfolio of utilities. Oh, I see.
- 17:04X Energy is an advanced nuclear reactor and fuel design company.
- 17:08They specialize in small modular reactors, or SMRS. SMRS right?
- 17:12Specifically, they're developing a high temperature gas cooled
- 17:15reactor and crucially, they manufacture A proprietary
- 17:19uranium fuel called Trio X, which is physically engineered
- 17:23to resist meltdown under extreme temperatures.
- 17:26Ah OK. So the Moat isn't current cash
- 17:29flow. The Moat is deep intellectual
- 17:31property. Yes, intellectual property
- 17:33combined with massive regulatory licensing barriers.
- 17:37Getting a new reactor design approved by the Nuclear
- 17:39Regulatory Commission requires years of painstaking engineering
- 17:42validation. But once you have it, nobody can
- 17:44easily replicate it. Exactly.
- 17:46They have already secured A deployment partnership with
- 17:48Amazon and signed long term fuel fabrication agreements.
- 17:51It is a highly speculative high upside investment on the
- 17:55structural shift toward modular, decentralized nuclear power.
- 17:58That makes perfect sense. It's a small calculated bet with
- 18:02asymmetrical upside. If SMRS become the standard for
- 18:06powering localized data centers, that intellectual property is
- 18:10worth billions. Easily.
- 18:13And Speaking of outliers, there's one more here, Vista
- 18:15Energy, ticker VIST. This is an upstream oil and gas
- 18:20exploration company operating down in Argentina.
- 18:23It's about 2.1% allocation. Now as a commodity producer,
- 18:26they have absolutely 0 structural pricing power.
- 18:29They are totally exposed to the boom and bust of global oil
- 18:32prices. Plus, you know all the
- 18:34macroeconomic volatility of operating in Latin America.
- 18:37True, it completely lacks the impenetrable Moat of a regulated
- 18:41US utility, but the business reality here is about
- 18:44operational efficiency. How so?
- 18:46Vista Energy is drilling so cheaply in the Vaca Muerta Basin
- 18:49that even if global oil prices completely tank, they are still
- 18:53pulling cash out of the ground. Really.
- 18:54Yeah, they treated incredibly low cash flow multiples.
- 18:57They can add a low cost production footprint with very
- 18:59favorable lifting cost per barrel.
- 19:01So even during moderate commodity downturns, it
- 19:04generates positive free cash flow.
- 19:06Exactly, it provides a high immediate cash yield to balance
- 19:11out a portfolio that is otherwise making very long
- 19:14duration bets on infrastructure. Here's where it gets really
- 19:17interesting for you listening. When you zoom out and look at
- 19:21this entire portfolio rotation from dumping NVIDIA to buying
- 19:25nuclear plants and toll bridge utilities, you are seeing a
- 19:28master class in anticipating the second and 3rd order
- 19:31implications of AI. It's phenomenal foresight.
- 19:35The initial phase of generative AI was all about hardware CapEx.
- 19:39That was 2023 to 2025. Capital flooded into
- 19:43semiconductor designers because chip demand vastly outpaced
- 19:46fabrication capacity. But computing power is
- 19:48ultimately physically bounded by electrical current.
- 19:52Right. The macro migration we're seeing
- 19:53here is the realization that the primary constraint has moved
- 19:56from silicon chips to physical energy bottlenecks.
- 19:59If we connect this to the bigger picture, it really comes down to
- 20:01the difference in capital durability between these two
- 20:03sectors. Capital durability.
- 20:05Yeah, think about the mechanics of semiconductor hardware
- 20:07design. It is highly volatile.
- 20:09It is exposed to rapid product obsolescence.
- 20:12Because an AI chip you designed today might be completely
- 20:15obsolete in two years when a competitor shrinks the
- 20:18transistor architecture. Exactly, you have heavy customer
- 20:21concentration and you are constantly at risk of cyclical
- 20:25supply gluts once new fabrication plants finally come
- 20:28online. Right.
- 20:29Eventually TSMC or Intel builds enough factories to make all the
- 20:33chips everyone wants, the scarcity vanishes and the profit
- 20:36margins completely compress. And the treadmill of perfect
- 20:39competition starts up all over again.
- 20:41But you the cash flow compounding is an entirely
- 20:44different beast. Power producers with firm
- 20:47nuclear and natural gas capacity are converting this surging,
- 20:50desperate AI demand into multi decade agreements.
- 20:54You don't just replace a power plant in two years.
- 20:56No you don't. That physical infrastructure
- 20:58offers long term compounding that semiconductors simply
- 21:02cannot match. You lock in those inflation link
- 21:04contracts, you lower your structural cost of capital
- 21:07overtime, and you consistently use the cash to retire common
- 21:10shares. It is a profoundly more durable
- 21:12foundation for long term capital preservation and growth.
- 21:15It really is. It's the ultimate reality check
- 21:18for the digital age. We talk about the cloud like
- 21:21it's this magical, ethereal thing.
- 21:23It's not. No, you can't write all the code
- 21:26you want in the cloud. But the cloud still runs on a
- 21:29physical machine in a physical building, and that machine needs
- 21:31to be plugged into a wall. Exactly.
- 21:34And that leaves me with this final, honestly somewhat
- 21:37unsettling thought for you to Mull over.
- 21:39We've seen how computing power is absolutely bound by
- 21:42electricity, right? We've seen how domestic
- 21:45transmission grids and nuclear facilities are becoming the new
- 21:48gold mines for intelligent investors.
- 21:51But what does this actually mean for global geopolitics over the
- 21:55next decade? That is the $1,000,000 question.
- 21:57If a nation's absolute supremacy and artificial intelligence, its
- 22:01economic and military edge, is entirely dictated by the
- 22:04physical constraints of its domestic power grid, do local,
- 22:08regional utility companies eventually become matters of
- 22:11ultimate national security? Very possible.
- 22:14Could a. Boring regional electric utility
- 22:16in Ohio or Texas find itself more fiercely protected by the
- 22:20federal government than the cutting edge algorithms
- 22:22themselves. It really makes you wonder.
- 22:24It makes you wonder if selling those AI chips for power plants
- 22:27wasn't just a brilliant financial pivot, but a bet on
- 22:30the true architecture of future global power.
- 22:33Thanks for joining us on this deep dive.
- 22:35See you next time.