Chase Lochmiller

1. Key Themes

Crusoe's Strategy: Build an "AI Super Major" Modeled on Exxon

Lochmiller frames Crusoe as the AI-infrastructure analogue of an integrated oil major. It sells three products across the value chain (data centers, GPUs, and tokens), and the vertical integration acts as a natural hedge because margin migrates between layers over time.

"there's really like three products that we ultimately sell to customers and where we're making money, we can sell data centers, we can sell GPUs, and we can sell tokens." (Chase Lochmiller) 00:34:24

"Exxon very famously doesn't hedge their oil exposure... they're sort of inherently hedged by being vertically integrated... when oil prices come down, Exxon makes less margin at the wellhead, but their margins that are downstream business actually go up." (Chase Lochmiller) 00:35:17

"our margins are going to move around across electrical data centers, chips, and services." (Chase Lochmiller) 00:36:14

The Real Constraint Is "Places to Plug In GPUs," and the Bottleneck Is Whack-a-Mole

Supply tightness is physical, not chip-level. The binding constraint rotates between components, energy, and skilled labor, so owning more of the stack is a speed advantage.

"there are not places to plug in GPUs. So that's ultimately the supply constraint." (Chase Lochmiller) 00:17:54

"The supply chain to support large scale AI data centers is sort of like a game of whack-a-mole. At different points, different bottlenecks come into play." (Chase Lochmiller) 00:17:54

"Energy is definitely a key constraint... Alongside that, labor is a very key bottleneck. There are a finite amount of skilled trade workers, electricians, welders, plumbers, construction workers in the United States." (Chase Lochmiller) 00:22:07

Vertical Integration as a Speed Weapon, Not a Margin Play

Crusoe integrated electrical manufacturing to hit deadlines and to design from first principles, not to stack margin. The Abilene build is the proof point.

"When we went out and canvassed the market for vendors that were supplying these medium voltage power distribution centers, the lead time for this one component was 100 weeks... we were able to do it in 28 weeks." (Chase Lochmiller) 00:18:49

"It's not the main reason we're doing it. It's like stacking margin. It's more... Availability, being able to deliver on time. And then also it gives us a critical innovation platform." (Chase Lochmiller) 00:20:18

Distributed, Energy-First AI Infrastructure (the Original Thesis)

Crusoe's founding insight was that AI inference doesn't need to sit in legacy hubs like Northern Virginia, because compute time dominates network latency. Bitcoin mining served as the early monetization engine for stranded or cheap energy while the AI platform was built.

"so much of the time to serve a neural network is actually the compute that's happening in the data center, not the time to get to the data center... we felt like the infrastructure to support AI wasn't going to be centralized. It was going to be distributed where energy was low cost and abundant." (Chase Lochmiller) 00:16:25

"at that scale compute becomes the bottleneck compute and data become the two bottlenecks and you know what is the bottleneck for compute energy so that was kind of the premise of starting the company." (Chase Lochmiller) 00:12:13

Time-to-Token and the Shift from Training Scale to Inference Utilization

2026 is the "era of the agent," and the winning metric shifts from gigawatt campuses to how fast capacity can be stood up. Crusoe is pushing small modular manufactured data centers to shorten that timeline.

"2026 is really about the scaling of utilization of AI infrastructure or utilization of LLMs and foundational models for useful tasks... For that, you don't need a gigawatt scale computer to serve these models... What matters is actually how quickly can I go from not having anything to being able to deliver tokens? Like, so what is my time to token?" (Chase Lochmiller) 00:21:13

"customers are being asked to forecast their demand further and further out because the timeline to bring online AI infrastructure is extending... early stage startups that are being asked to commit to compute in 2028, it's like that is an eternity from now." (Chase Lochmiller) 00:41:04

GPU Depreciation: Software Services Extend Chip Lives

Lochmiller argues the market underestimates how application ingenuity keeps old silicon valuable. Hopper pricing is the counterexample to the skeptics.

"Here we are three years later and the prices being charged for utilizing hoppers is higher than the rates that were being charged three years ago when they were brand new." (Chase Lochmiller) 00:40:04

"abstracting away some of the difficulties by actually, you know, having a suite of services that can monetize these chips for longer will ultimately, we believe, extend the depreciation cycle beyond six years." (Chase Lochmiller) 00:39:37

"I think people underestimate the ingenuity of applications and application developers of turning compute capacity into value." (Chase Lochmiller) 00:40:04

A Portfolio Approach to Contract Margin

Crusoe blends long take-or-pay contracts, shorter higher-margin contracts, and managed services to manage financing and renewal risk.

"We are renting capacity on a long-term basis, call it five-year contracts to credit quality customers... There are shorter term contracts that we'll do at higher margins, but they're riskier because at the end of the contract, it's like, is there a renewal?" (Chase Lochmiller) 00:33:12

"Those contracts are typically much shorter term in nature, and they end up being higher margin to Crusoe. And so you want a portfolio of those different margin profiles." (Chase Lochmiller) 00:34:08

Inference Efficiency Is a Memory-Management Game

The edge in serving tokens comes from keeping the GPU busy, which depends on moving KV cache across the memory hierarchy efficiently.

"the GPU is actually the most valuable thing in the entire data center, right? It is like the most expensive thing. So it's the thing you want to make sure you're keeping busy. If it's sitting idle, that's just, that's just money burning." (Chase Lochmiller) 00:45:30

"being able to really, really efficiently move data around is actually a critical aspect to being able to serve inference and be able to keep those GPUs busy." (Chase Lochmiller) 00:46:24

Data Centers Have a Narrative Problem the Facts Don't Support

Lochmiller lays out data on water, energy prices, jobs, and tax receipts, and argues the industry has lost the storytelling battle ahead of the US midterms.

"One of our giant buildings in Abilene, Texas, call it 140 megawatts... use about the same water annually as about 10 single family homes." (Chase Lochmiller) 00:25:04

"When you look at markets where data centers have made investments and have built big data centers, typically energy prices for communities have come down because it catalyzes more investment in energy generation... and you end up with more megawatts being amortized over the same transmission and distribution infrastructure." (Chase Lochmiller) 00:25:56

"we're going to account for more than a third of the tax revenue in Taylor County and Abilene... For the school system, we're more than doubling the tax receipts that are going to the schools." (Chase Lochmiller) 00:27:14

Open vs. Closed: Dollars Go to Closed, Tokens Go to Open

Crusoe's managed-inference vantage point gives an unusual read on the split.

"I do think that people are spending more money on closed source frontier models than they are on open source, but they are generating more tokens on open source than closed source." (Chase Lochmiller) 00:49:14

"there's far more private data that's untapped that can actually lead to massive performance gains." (Chase Lochmiller) 00:49:14

2. Contrarian Perspectives

"Most Moats Are an Illusion"

In the context of a hyper-capitalized infrastructure business, Lochmiller says he changed his mind on durable competitive advantage. Speed and adaptability beat defensibility.

"most moats are a illusion. Most moats don't exist. Most of them are ephemeral. And especially during a moment of accelerating technological progress... a lot of it just comes down to being able to move quickly and adapt efficiently to the ever-changing chess board that's in front of us." (Chase Lochmiller) 00:54:41

Harry's complementary view: "everything starts off as a moatless wrapper and it's the value that you build on top of it in segments." 00:55:22

Data Centers Lower Local Energy Bills and Barely Use Water

This is the opposite of the dominant public narrative. He cites measured outcomes in Abilene and in markets with big builds, not just claims.

"The data has shown that energy prices actually come down. It's actually the opposite of the narrative that's being told." (Chase Lochmiller) 00:26:42

"The number of people that have told me that data centers use all the water in the world is crazy. It's also just false." (Chase Lochmiller) 00:24:41

The Supply Bottleneck Is Energy Plus Blue-Collar Labor, and AI Is Creating Jobs, Not Destroying Them

Lochmiller sees the data center buildout as a re-industrialization event, and he thinks the public's AI job fears are misdirected.

"people are concerned that AI is going to take jobs, but yet it is so far only created tremendous amounts of jobs and tremendous amounts of economic development." (Chase Lochmiller) 00:31:32

"They're leading to this massive economic boom for the entire blue collar labor economy in the United States, this massive resurgence in re-industrialization of the United States." (Chase Lochmiller) 00:31:04

Wealth Concentration Isn't Worrying Because Value Created Dwarfs Market Cap

Pushed by Harry on the concentration of value in a handful of IPOs, he reframes the unit of measurement to value created for the world, not value captured by companies.

"if you look at the value that's being created by abundant intelligence that is going to drive massive productivity gains for the global economy, it's many fold of the market cap that's being created." (Chase Lochmiller) 00:54:21

Consumer Demand Softening Wouldn't Be the First Crack

When Harry argues that take-or-pay contracts plus a consumer demand crack is the key risk, Lochmiller counters that demand is broad-based across science, materials, and drugs, and not dependent on one consumer app.

"the thing about AI is it's transformative for so many areas of the economy. It's not, it's not a single thing that's benefiting." (Chase Lochmiller) 00:37:24

"it's the worst it's ever going to be for the rest of the future of humanity." (Chase Lochmiller) 00:37:46

3. Companies Identified

Crusoe

Vertically integrated AI infrastructure company (data centers, Crusoe Cloud GPUs, managed inference) that raised a $3.9 billion Series F at a $30.9 billion valuation (as introduced by Harry). Founded to monetize stranded energy with compute, it now builds gigawatt-scale AI campuses. Its Abilene campus, and the 200+ MW first phase delivered in a year when the next closest bid was two and a half years, is the signature case study.

"We had sort of committed to doing this in one year. And the next closest bid was two and a half years. There were 34 other data center developers, but speed was of the essence." (Chase Lochmiller) 00:18:21

"Crusoe is approaching 2000 people." (Chase Lochmiller) 00:57:20

NVIDIA

GPU maker and a long-term partner of Crusoe. Lochmiller used NVIDIA's chip power roadmap (150-200W to 300W to 600W per chip) to anticipate how data center design would have to change.

"we're long-term partners of NVIDIA and we'd been working with NVIDIA. And I was looking with them at the roadmap, you know, many years ago... as the power density goes up, it's going to change what the data center looks like." (Chase Lochmiller) 00:15:33

Fireworks AI

Inference platform. Harry's firm is an investor, and Lochmiller praises it as a specialized competitor and potential partner, describing the AI stack as "compete and collaborate."

"Fireworks is great. They've done an incredibly great job." (Chase Lochmiller) 00:47:33

Exxon / Chevron (as the model)

Traditional oil and gas super majors that serve as the strategic template for Crusoe: integrated upstream, midstream, and downstream, naturally hedged against commodity cycles.

"there are these super majors like Exxon and Chevron in the traditional oil and gas industry that are vertically integrated across upstream, midstream, downstream." (Chase Lochmiller) 00:35:17

Cognition

Cited as a leading example of a company post-training its own model on proprietary data.

"folks like cognition post-training their own model with a lot of their own data." (Chase Lochmiller) 00:50:13

Harvey

Legal AI company mentioned as possibly taking a similar domain-specific model approach.

"Harvey kind of maybe trying to take a similar approach from a legal perspective." (Chase Lochmiller) 00:50:13

Legora (transcribed "Lagora")

Legal AI product that Harry's girlfriend, a lawyer, uses; cited as evidence of how fast AI is changing professional work.

"My girlfriend's a lawyer. Six months ago, she didn't use Legora. Now she hasn't written a document in five months." (Harry Stebbings) 00:43:19

Compute exchange platforms (the Compute Exchange, Ornn, and others)

Emerging marketplaces creating tradable futures for GPU compute, supporting Lochmiller's view of compute as a commodity.

"there's a couple platforms like the Compute Exchange, Ornn, I think there's a handful of others that are like creating these tradable futures products." (Chase Lochmiller) 00:32:22

SGLang and vLLM

Open-source inference-serving projects; Lochmiller notes they are generalizable but that not all inference providers are equal.

"you have like great open source projects like SGLang and vLLM, but... those are very generalized or generalizable." (Chase Lochmiller) 00:47:15

MIT and Stanford research groups / Los Alamos

Early research relationships Crusoe used to learn what AI researchers value in infrastructure; Lochmiller also did physics research at Los Alamos.

"we were working with a lot of researchers at MIT and Stanford and different groups trying to understand what's actually valuable for AI researchers from an infrastructure platform." (Chase Lochmiller) 00:14:04

Eight Sleep, MongoDB, Framer (sponsors)

Show sponsors: Eight Sleep (smart mattress cover, Pod 6 from $1,999), MongoDB (AI agent data platform; cites ElevenLabs running 40 million agents and 75% of the Fortune 100), and Framer (AI website builder used by Perplexity and Miro).

"11 Labs run 40 million agents on MongoDB." (Harry Stebbings) 00:02:23

Data center security (unnamed category)

Harry flags it as a business he'd like to invest in.

"If anyone's doing data center security, by the way, would love to invest in that business." (Harry Stebbings) 00:31:58

4. People Identified

Co-founder and CEO of Crusoe; former quant trader, MIT-trained physicist, and mountaineer who climbed Everest in 2018. His background (physics research at MIT and Los Alamos, then quantitative finance) informs a Bayesian, probabilistic worldview.

"I think about the world in a very bayesian approach right i i don't i don't think the future's deterministic I think it's very probabilistic." (Chase Lochmiller) 00:13:12

Gavin Baker

Investor in Crusoe, quoted by Harry as saying the lowest-cost producer of intelligence wins.

"Gavin Baker, one of your investors, has said that the lowest cost producer of intelligence wins." (Harry Stebbings) 00:44:17

Elon Musk

Referenced for the "idiot index" framing, which compares the cost of a finished product to its raw materials. Lochmiller used it to explain Crusoe's end-to-end cost perspective.

"Elon has this very interesting framing of this. He calls it the cost of raw materials and then what is the end product relative to that cost of just the raw materials going in." (Chase Lochmiller) 00:19:43

Michael Dell

Named as the board member Lochmiller would most want, for his business instincts and character.

"I would say Michael Dell... I just have tremendous admiration for Michael just as a human being... He's got incredible business instincts. He's a great family man." (Chase Lochmiller) 00:56:12

Lee Jacobs

Referenced by Harry as having described Lochmiller's risk appetite in ordering millions of dollars of GPUs from NVIDIA "well before it was cool."

"I spoke to Lee Jacobs before, and he was talking to me about the risk mindset that you had to ordering millions of dollars of GPUs from NVIDIA well before it was cool." (Harry Stebbings) 00:40:48

Dario Amodei

CEO of Anthropic; Harry attributes to him and other AI leaders the "we're going to replace all jobs" messaging, which he argues fuels backlash against data centers.

"With respect to Dario and a lot of other leaders who've said for the last few years, we're going to replace all jobs... surprisingly, they hate you." (Harry Stebbings) 00:30:42

Zach (unnamed surname)

A fellow figure Harry spoke to before the episode; Lochmiller calls him "an awesome guy."

"Zach's an awesome guy. I love Zach." (Chase Lochmiller) 00:56:34

5. Operating Insights

Set Unreachable Windows and Take the Red-Eye to Protect Family Time

Lochmiller is explicit and structural about it, with specific rules the whole company knows.

"I take a lot of red eyes just so I don't miss a bedtime with my kids. I'm home almost every weekend... if I'm home in the Bay Area, everybody knows that I'm unreachable from 7 to 8 a.m. It's like I'm not taking any meetings at that point because I'm making my kids breakfast." (Chase Lochmiller) 00:52:09

Make the Operating Culture Explicit with a Plan A Through Plan D Discipline

He bakes "think like a mountaineer" into company values: redundancy in planning, endurance, and a safety-first mindset across physical construction and digital security.

"I have plans all the way down in case different things go wrong." (Chase Lochmiller) 00:05:57

"getting getting up to the top is optional getting down is mandatory." (Chase Lochmiller) 00:06:27

Win on Timeline by Vertically Integrating the Single Longest-Lead Component

Rather than accepting a 100-week supplier lead time, Crusoe built the component in-house in 28 weeks. The tactic is to find your one gating item and bring it inside.

"I went to our internal team and we had consciously vertically integrated electrical manufacturing. We said, how quickly could we make this ourselves?" (Chase Lochmiller) 00:18:49

Commit to a Date That Disqualifies Competitors

In a market where speed was the buying criterion, quoting a one-year delivery against a two-and-a-half-year next-best bid let Crusoe win the Abilene deal.

"speed was of the essence. So I said, sure, we can do it in a year." (Chase Lochmiller) 00:18:21

Build a Customer-Informed Demand Forecast and Reduce Customer Forecasting Burden

Rather than forecasting demand abstractly, Crusoe builds up from customer conversations and is designing modular products so customers don't have to commit years ahead.

"we build up forecasts from having conversations with customers... deploying small modular manufactured AI data centers that can deliver more just-in-time AI infrastructure." (Chase Lochmiller) 00:41:04

6. Overlooked Insights

Take-or-Pay Plus Managed Services Means the Real Hedge Is Customer Mix, and the Real Moat May Be the Abstraction Layer

Buried in the depreciation answer is the key mechanism: services that abstract away the underlying chip decouple monetization from hardware generation, which is how Crusoe argues old chips stay valuable and depreciation extends past six years. Combined with his "most moats are an illusion" comment, it implies the durable value is in software and services that sit on top of commodity hardware.

"when you look at that business where it abstracts away the actual underlying chip, the underlying compute from the service that people are receiving, it opens up new monetization engines that can persist for a much longer period of time." (Chase Lochmiller) 00:39:09

The Bitcoin Data Center Playbook Is the Template for AI's Cost Curve

Lochmiller briefly noted that Bitcoin mining shed roughly 98% of data center cost by dropping redundancy ("chicken coop" facilities), and that he expects a similar playbook in AI as power density rises. This is a quiet but major cost-curve thesis: AI inference doesn't need five-nines reliability or centralized hubs, which could reshape data center economics and favor modular, cheap-energy sites.

"they were able to eliminate something like 98% of the total data center cost with like these like chicken coop type data centers. You know, very, very low cost, no redundancy... we felt always that something similar was going to play out in AI." (Chase Lochmiller) 00:15:33