RIOT - Ultrathink: The Power Plant Paradox

A Buffett/Munger meditation on Riot Platforms and the AI infrastructure thesis

The Core Question: Asset or Liability?

Charlie Munger liked to say that the best way to think about a business is to ask: "What is the simplest explanation of what this company actually does?" For Riot Platforms, the answer has changed three times in five years, and that should give any serious investor pause.

First, Riot was a biotech company called Bioptio (one wonders what happened to that thesis). Then it became a Bitcoin miner -- essentially, a company that converts electricity into cryptocurrency. Now it wants to be a data center landlord -- a company that sells electricity to hyperscalers running AI workloads. The common thread is electricity. The uncommon thread is that management keeps changing what they do with it.

This is the opposite of what Buffett looks for. When he bought See's Candies, he knew it would still be selling chocolates in fifty years. When he bought Coca-Cola, the product was 100 years old and getting more popular every decade. Riot has reinvented itself twice in five years and is asking investors to trust that the third incarnation will be the profitable one.

But let us be fair. Sometimes a pivot is exactly the right move. The key question is whether the underlying asset -- 1.7 gigawatts of approved power capacity in Texas -- is genuinely valuable, or whether it is a commodity dressed up in narrative clothing.

Moat Meditation: The Power Portfolio

Here is where the Aschenbrenner thesis gets interesting, and where honest analysis requires setting aside one's biases.

Power is indeed becoming the binding constraint on AI development. Every hyperscaler -- Microsoft, Google, Amazon, Meta -- is scrambling for gigawatt-scale power sites. Getting a new gigawatt approved, permitted, and interconnected to the grid takes 5-7 years in most jurisdictions. Riot has 1.7 GW of approved power today, with land, with grid interconnection, in a market (ERCOT/Texas) that hyperscalers prefer.

This is an asset that cannot be easily replicated. In that narrow sense, there is something moat-like about it. It is similar to how Copart's 19,000 acres of salvage yards near population centers create an irreplaceable physical moat -- you simply cannot get these permitted anymore.

But here is the critical distinction Munger would make: having a scarce asset is not the same as having a moat. A moat requires the ability to earn superior returns on capital over time. Riot has spent $3.5 billion over five years and generated negative cumulative free cash flow. The asset is scarce, but the management has not demonstrated the ability to monetize it profitably.

Core Scientific -- Riot's most direct competitor -- has already signed meaningful HPC contracts and is further along the execution curve. TeraWulf has done the same. Riot has one 25 MW lease with AMD out of 1,700 MW of capacity. That is 1.5% execution. To use Munger's fishing analogy: they have an enormous lake, but so far they have caught one fish.

The Owner's Mindset: Would Buffett Own This for Twenty Years?

Absolutely not. And this is not even a close call.

Buffett's criteria are disarmingly simple: consistent earnings, high returns on equity, limited debt, honest management with skin in the game, and a predictable business. Riot fails every single one:

• Consistent earnings? Profitable in 1 of 5 years. Net loss of $663M in the most recent year.
• High ROE? Negative 23%. Five-year average: negative 13%.
• Limited debt? Went from $575M net cash to $633M net debt in two years.
• Management skin in game? CEO owns 2% and has been selling shares.
• Predictable? Revenue swings 72% year-over-year based on Bitcoin's price, a commodity no one can forecast.

Buffett has been explicit about avoiding businesses like this. In his 2014 letter, he wrote about gold (and by extension, Bitcoin): "If you own one ounce of gold for an eternity, you will still own one ounce at the end." Bitcoin produces nothing. It generates no cash flow. A company whose revenue is 89% dependent on mining Bitcoin is, at its core, a speculation on a non-productive asset.

The counter-argument is that Riot is transitioning away from this dependence. Fair enough. But transitions are uncertain, and the market is pricing partial success already. At $14/share ($5.3B market cap), the power portfolio is implicitly valued at roughly $3M per MW -- which is not cheap for uncontracted capacity.

Risk Inversion: What Could Destroy This Business?

Munger's inversion principle asks: "What would have to go wrong for this investment to be a disaster?"

The list is uncomfortably long:

1. Bitcoin drops below $50K -- Mining becomes cash-flow negative at all-in costs of $91K. The company would need to sell Bitcoin treasury to fund operations, creating a death spiral.

2. AI/HPC demand disappoints -- If the AI buildout slows (as happened with telecom fiber in 2001), Riot's power portfolio becomes stranded. The market would rerate from "AI infrastructure" back to "struggling Bitcoin miner" -- probably worth $3-5/share.

3. Continued dilution -- Another 50-100% share dilution (via ATM or convertible conversion) would destroy per-share value even if the business improves.

4. Competition wins the deals -- Core Scientific and TeraWulf are ahead. If hyperscalers sign with them first, Riot's "right market, right time" advantage may not matter.

5. ERCOT market changes -- Power curtailment credits ($57M in 2025) are a meaningful revenue source. Regulatory changes could eliminate this subsidy.

The bull case requires multiple things to go right simultaneously: Bitcoin stays above $80K, Riot signs major HPC leases in 2026, and the company stops diluting shares. The bear case only needs one thing to go wrong.

Valuation Philosophy: Is Price Justified by Quality?

This is where Klarman's thinking is most relevant. Seth Klarman's Margin of Safety teaches that the first question is not "What is the upside?" but "What am I risking to get it?"

At $14.01:

• You are paying 1.78x book value for a company that has destroyed book value through negative returns
• You are paying 8.9x sales for a business with negative operating margins
• You are getting no dividend, no share buybacks, and a history of aggressive dilution
• Your downside is real: the stock traded at $6.19 within the last year

The margin of safety is negative. You are paying a premium for optionality on an unproven pivot.

Compare this to the quality companies on our watchlist: Copart at 19,000 irreplaceable acres with 21% ROE. ASML with 100% EUV monopoly and 41% ROE. Costco with 92.9% membership renewal and 30% ROE. These are businesses where the moat is proven, the returns are consistent, and patience is rewarded with compounding.

Riot is not in this category. It is not even in the same universe.

The Patient Investor's Path

Here is the honest conclusion, uncomfortable as it may be for those attracted to the Aschenbrenner narrative:

The thesis may be correct. AI infrastructure may indeed be dramatically undervalued. Power assets may be the "picks and shovels" of the AGI revolution. Riot's 1.7 GW may someday be worth multiples of today's market cap.

But the vehicle through which you express that thesis matters enormously. Riot is a capital-destroying, dilution-prone, execution-lagging company with no track record of profitable operations. Even if you believe the thesis, there are better ways to play it:

• Core Scientific (CORZ) -- further ahead in HPC execution
• The hyperscalers themselves -- Microsoft, Google, Amazon benefit from AI compute demand with actual moats, cash flows, and dividends
• Power utilities -- companies like NextEra Energy with proven operations, regulated returns, and dividend histories

For the disciplined value investor, RIOT is a REJECT. It does not meet the minimum quality standards for capital deployment. If one must own exposure to this thesis, limit it to 1-2% of portfolio and buy at distressed prices (below $9, near book value), recognizing it is speculation, not investment.

Ben Graham's most famous lesson applies perfectly here: "An investment operation is one which, upon thorough analysis, promises safety of principal and an adequate return. Operations not meeting these requirements are speculative." RIOT, by this definition, is speculation -- regardless of how compelling the narrative may be.

"It is far better to buy a wonderful company at a fair price than a fair company at a wonderful price." -- Warren Buffett. Riot Platforms is neither a wonderful company nor available at a wonderful price.