SpaceX, Palantir, Anduril & Onodrim | The Return of Conglomerates

1. Key Themes

SpaceX Is Becoming a Vertically Integrated Hyperscaler, Not Just a Launch Company

Garrett (137 Ventures, a 16-year SpaceX investor whose largest position it is) argues the market underappreciates that SpaceX is committed to AI infrastructure as a durable business, not a one-off. He cites Q3 numbers and the hyperscaler logic: "They grew top line like 160%. They, you know, posted $4.4 billion of net income, $8.8 billion of EBITDA." 00:00:02 He adds: "the company is clearly going the route of a hyperscaler. They're building out more software and services around that compute stack in the infrastructure. They're committed to this business, this is not just a one-off thing, and they are the best in the world at building this capacity and building it the fastest." 00:02:11 The valuation framing: "consensus EBITDA, just based off of Q3, it's a business that's trading anywhere from 20 to 30 times. 20 to 27 EBITDA. For a $2 trillion company that is a dominant provider in multiple critical industries." 00:04:22

Vertical Integration Gives Risk-Adjusted Exposure to AI Regardless of Who Wins

Garrett's investment logic for owning integrated players: "if you're going to invest anywhere within this space, you probably want to own the vertically integrated players, which they are, right? From the ability to monetize at the app layer, the model layer, all the way down to monetize compute. That gives you some bit of a risk-adjusted... investment because whether you have a shift towards open-source... or you continue to see the closed-source models win out, they kind of benefit either way." 00:02:55 And the downside protection: "you also have the ability to shift capacity towards your own products if demand slows down in the out years." 00:03:33 He also notes fast build cycles mean "your CapEx investments are on shorter duration cycles, which means you're going to get a payback period on the sort of outlay, and then if the markets shift, you can shift quickly." 00:03:54

Sub-12-Month Payback and Cost of Capital Decide the Neocloud Race

On the financing squeeze, Garrett says: "in order to really build out this capacity... speed and payback period is everything. And then cost of capital is a huge part of it, because a lot of this is funded by the debt markets." 00:14:18 He agrees with Elon's point about financing advantages: "there are certain companies that have a huge advantage, right, and a lower cost of capital and they're also going to suck up a lot of that demand at, you know, better rates." 00:14:42 He also flags SpaceX reiterating "they're getting sub-12-month payback periods, right, which is obviously a signal for demand." 00:02:42 He admits the pricing difficulty: "How to price a company is going to be entirely driven on by your ability to believe in some out years of earnings. And that's really hard right now. It's hard to believe in the next 2 years, right, of earnings, right, let alone the next 20." 00:15:45

Launch Cost Collapse Creates Entire Downstream Industries (Starship, Orbital Compute, Varda)

Garrett explains that driving down the cost of critical infrastructure is the mechanism that spawns new businesses: "for launch, we saw that even with Falcon 9... the sort of exponential increase in the number of launches and how much capacity is delivered to space because of SpaceX driving costs down." 00:10:11 On Starship: "a lot of the business models that we're talking about today within SpaceX were not even thoughts just, you know, 1, 2, 3, 4, 5 years ago." 00:10:36 He claims Starship is underpriced: "that launch on just a handful of Starlink V3s delivered like 1% of the world's internet capacity... And they're going to deploy tens of thousands of these." 00:06:34 On timing for orbital compute: "they're trying to launch some of the first orbital compute satellites, you know, within the back half of next year, I believe it is." 00:09:17

The Return of the Conglomerate (Modern, Platform-Based)

Garrett's core thesis: "I have this like, return of the conglomerate sort of thesis right now." 00:00:26 Old conglomerates failed because of "no shared common operating platform" and diseconomies of scale; new ones differ because "they actually have a core, common operating tech platform... or... they're actually vertically integrated." 00:29:23 He cites Apple and SpaceX as examples of vertical integration across disparate industries, and Anduril as a "lowercase c conglomerate" where, via Lattice, "every marginal product sold, right, makes the value of the other ones go up." 00:30:42

Dual-Use Scale Is the Only Way to Sustain Defense Industrial Capacity

Garrett argues commercial dominance is a prerequisite for defense capacity: "to have the type of capacity that we really need from a defense perspective, you're going to need to be dominant in the commercial industry, or else it just won't work." 00:31:55 The evidence he offers: "the defense primes of China, right, like 60 to 80% of their revenue comes from the commercial sector," and shipbuilding as the prime example: "China's leading prime there, right, is so dominant because they're a dominant commercial ship manufacturer in the world, and then they also do defense products." 00:31:55 He names "Palantir and SpaceX" as the only current capital-C conglomerates, and says "the best success stories probably are those that are in the commercial sector that are dominant that are then crossing over in the defense industry." 00:32:37

The Neglected Supplier Layer: Onodrim's Roll-Up of European Defense and Aerospace Suppliers

Garrett describes Onodrim as "an aerospace and defense scale-up platform. We want to use technology to buy and acquire businesses all across Europe within critical industries, deploy technology into them, scale them up, and grow them." 00:33:49 The gap: "In Europe, you have not seen that level of consolidation... definitely not a HEICO of Europe." 00:34:08 And where value is: "now there's 4,000 drone companies in Europe. It's clearly that this is overfunded, right, and too crowded. What has not received as much support is the layer below, right? Who's supplying all these companies?" 00:36:21

Duration of Compounding Matters More Than Growth Rate

On what 137 underwrites: "SpaceX, you know, compounded for like 40% for 20 years. Now, 40% on a stand-alone year is good, right, but it's not incredible... What's really amazing about that is the length, right, the duration of that." 00:19:34 And the math: "9% over 30 years, right, is like a 13x." 00:20:00 Because they are minority holders, "we want to be excited about it 10 years from now, 20 years from now." 00:20:24

Power-Law Math Justifies Capital Flooding Into Hot Categories

Garrett uses fintech as proof that apparent over-capitalization is rational: "There was a no shortage of capital that was flooding into fintech... a ton. And there's a lot of companies that are on... the graveyard... Except if you look at Stripe, right, Revolut, right, $100-plus billion opportunities... So that capital was not wasted, right? It's just it's a power law." 00:26:05 He applies it forward: "That's how this will play out across many categories, definitely in AI, definitely in aerospace and defense." 00:26:25 His godfather Teddy Forstmann's lesson: "you would much rather own a quality asset that's a 5X and pay a 25% premium to buy it than to own a... shitty asset, right, and pay a 25% discount to buy it." 00:24:39

Deploying Software and AI Into Legacy Manufacturing Is the Re-Industrialization Playbook

Garrett cites Applied Intuition ("touching the electronics of these vehicles, right, and simplifying all the processing units... and the wiring and... improve the cost structure"), General Matter ("applying sensors to things where there wasn't sensors... most of this capacity with state-owned actors that are still using Cold War era technology"), Hadrian, and Palantir's pioneering of "deploying software into legacy manufacturing, from shampoos to planes." On AI specifically: "Your ability to start integrating in PLCs is significantly faster, which allows you to do a lot of interesting things with machines." 00:49:40

Preserving Tech's Non-Zero-Sum Culture and Resisting Political Weaponization

Garrett closes with a warning: "our industry is unique, that it's not zero-sum." He adds: "consumer decisions and... policy-makers' decisions shouldn't be influenced by the politics of the people involved in these companies. They should just be influenced by the merits of the business." 00:54:02 And a cautionary prediction: "tearing this industry from outside because of political reasons, it will backfire." 00:54:02

2. Contrarian Perspectives

The Best Way to Survive AI Uncertainty Is to Own the Vertically Integrated Spender, Not Pick Model Winners

While most investors debate open vs. closed models, Garrett argues SpaceX/xAI wins either way: "whether you have a shift towards open-source, right, and that changes, right, the economics on inference and compute, or you continue to see the closed-source models, you know, win out, they kind of benefit either way." 00:03:14 His evidence is that capacity can be redirected to own products and CapEx cycles are short enough to throttle. This runs against the instinct to pick the "best lab."

Europe's Overfunded Drone/Neo-Prime Layer Is a Crowded Trade; the Boring Supplier Layer Is the Opportunity

Contrarian to the dominant venture narrative: "there's 4,000 drone companies in Europe. It's clearly that this is overfunded, right, and too crowded. What has not received as much support is the layer below, right? Who's supplying all these companies, right? Where are they getting their key components, right? Where are they getting an aerostructure manufacturing, right? Where are they getting their engines?" 00:36:21 He confirms the neo-prime winners "are established," so incremental dollars chase diminishing returns.

Conglomerates, the Failed '80s Model, Are Back and Will Command Premiums

The consensus view is that conglomerates are discounted and unfocused. Garrett inverts it: "I think we need to return to the like dual-use conglomerates. You know, General Mills, right, made cereal and weapon systems." 00:00:26 He says the market "will value the diversification" and that "scale allows you to subsidize and maintain capacity and capabilities that are critical to defense sector, and also throttle up and down as needed." 00:31:26

Defense Capacity Can't Be Built by Industrial Policy Alone; You Need Commercial Dominance

Against the Washington consensus that budgets and contracts fix production: "industrial policy can only do so much... to have the type of capacity that we really need from a defense perspective, you're going to need to be dominant in the commercial industry, or else it just won't work." 00:31:55 He also says in Europe the problem "was not going to be how much money Europe spends," it was "solving this production issue." 00:42:48

Overpaying for Quality Is Rational, and Fintech's Graveyard Doesn't Invalidate the Capital Flood

Rather than calling current prices irrational, Garrett argues the power law pays for the mispricing: "things will be mispriced cuz there's too much capital chasing too few opportunities. The power law will probably pay for that amount of capital. As an investor and as an LP, you probably want to make sure that you're in the power law of outcomes." 00:26:48

3. Companies Identified

SpaceX: Launch, Starlink, and now AI/compute company; 137 Ventures' largest position (16 years). Mentioned as "the most important company in the world," growing 160% with $4.4B net income and $8.8B EBITDA in Q3. "It's still the company we are most excited about and I think the most important company in the world." 00:12:21

Starlink: SpaceX satellite internet; V3 satellites and direct-to-cell. "Starlink, and that's moving way faster, particularly on the D2C side." 00:01:56 V3 launch "delivered like 1% of the world's internet capacity." 00:06:34

Starship: SpaceX's next-gen launch vehicle. "The iteration of Starship that we see today is a fully-functioning launch platform compared to everyone else." 00:11:18

xAI / Grok: AI products under SpaceX. "Whether it's from Cursor or Grok, Grok bots, right? They're showing their ability to productize and monetize their AI products." 00:01:56

Cursor: AI coding product referenced as monetized within the SpaceX complex. Mentioned alongside Grok. 00:02:11

Anduril: Defense neo-prime founded by Palmer Luckey; $90B valuation per Molly, a 137 holding. Called a "lowercase c conglomerate" built on Lattice, and predicted as "by far the power-law outcome in this category." 00:32:37

Lattice (Anduril): Anduril's common software/AI backbone. "They've actually built the same core back-end infrastructure... with Lattice, which powers kind of everything." 00:30:42

Palantir: Data/AI software company; 137 invested "almost 20 years." Cited as one of two modern capital-C conglomerates, with majority commercial revenue, and as pioneering deploying software "into legacy manufacturing, from shampoos to planes." 00:47:22

Onodrim: European aerospace/defense supplier roll-up co-founded by Garrett, Isis (ex-Palantir Europe), and Alex. "We want to use technology to buy and acquire businesses all across Europe." 00:33:49

Cognition: AI software engineering company (Devin); 137 invested heavily, at $48B and rumored higher. Cited as positioned at the harness/app layer with third-party structural advantage. 00:17:00

Hex: Data/analytics workflow platform; 137 just co-led its round. "We just co-led Hex's round, which is a company we've been investing in for a long time." 00:16:36

Ramp: Corporate card/finance automation; ~$60B valuation. "A case study of just continuing to not just expand the purview of the business... they're still underlevered on a lot of parts of the business, meaning they're underpricing on things." 00:24:59

Applied Intuition: Software for automotive and autonomous vehicles. "Touching the electronics of these vehicles... improve the cost structure of these vehicles." 00:50:16

General Matter: Uranium enrichment startup (Scott Nolan). "Bringing back an important, critical capacity in uranium enrichment, but they're also reinventing the playbook for how to do it." 00:50:16

Varda: Space manufacturing co-founded by Delian Asparouhov and Will Bruey. "Built a successful business on top of the ability to have a lower cost of launch... pharmaceutical industry and... defense industry." 00:10:11

Starcloud: Orbital compute startup (CEO Philip). Visited by Molly; Starcloud 1/2/3 in development, with 3 destined for Starship. 00:08:25

Armada: Edge compute company founded by Dan Wright and Jon Runyan, incubated from an idea 137 partners Alex and Justin pitched to SpaceX. "It's done extremely well." 00:42:48

Hadrian: Advanced manufacturing automation. Cited as adopting software in industrial manufacturing. 00:49:40

Amca: Cited alongside Valinor as building awesome businesses in the defense supplier layer. 00:36:21

Valinor: Founded by Jay and Julie; "awesome" defense-supplier-layer business. 00:36:21

Stripe / Revolut / Nubank: Fintech survivors from the COVID-era capital flood, offered as proof of the power law. "Stripe, right, Revolut, right, $100-plus billion opportunities." 00:26:20

Apple: Example of a vertically integrated conglomerate across consumer products and services. 00:29:47

TransDigm / Teledyne / HEICO / Honeywell: US supplier consolidators that Onodrim emulates for Europe. 00:34:08

Airbus: Dominant European aerospace OEM underpinning a mature supply chain. 00:35:11

Stark / Quantum / Helsing / KNDS: European neo-primes with sovereign advantages, "established" winners. 00:36:21

OpenAI and Anthropic: "Two of the most impressive businesses they've been built in a generation." 00:18:30

Google (orbital compute test; Google Cloud) and AWS: Google "just obviously tested a version just recently deployed in orbit" 00:09:29; Google Cloud and AWS cited as hyperscalers seeing growth regardless of model mix. 00:03:14

CoreWeave: Neocloud with financing issues referenced by Molly. 00:13:04

Rheinmetall, McKinsey: Employer backgrounds of Onodrim's team alongside Palantir and SpaceX. 00:39:03

Brex, Harvey, Mercor, Zone, Public, Deel, VCX by Fundrise: Sponsors; not investment insights.

4. People Identified

Christian Garrett: Founder/partner at 137 Ventures; co-founder of Onodrim. Provides the SpaceX thesis and conglomerate framework. "It's a company we've invested in for 16 years, it's our largest position." 00:00:02

Elon Musk: SpaceX founder. "I think he's always been right depending on what timeline you pick." 00:04:22 Also noted for flagging financing advantages of low-cost-of-capital players. 00:14:42

Palmer Luckey: Anduril founder. "Palmer's been obsessed with for quite a long time... they tend to be early and on the ball." 00:27:28

Shyam Sankar: Palantir CTO. "Shyam Sankar from Palantir talks about this all the time" (commercial-dominance-for-defense thesis). 00:31:26

Isis: Onodrim co-founder, second employee at Palantir in Europe. "Extremely well-known, well-respected, and well-regarded." 00:42:48

Alex (Onodrim co-founder): Co-builds Onodrim. 00:42:48

Shaun Maguire: Sequoia partner; published a SpaceX financial model on Twitter. "I love Shaun." 00:12:19

Delian Asparouhov: Varda co-founder and Founders Fund partner; close friend of Garrett. 00:10:11

Will Bruey: Varda co-founder. 00:10:11

Scott Nolan: General Matter founder; "first person I met in venture." Observed "how incredibly smart people were back in the '60s." 00:50:16

Jay and Julie (Valinor): Founders. "I'm big fans of Jay and Julie." 00:36:21

Dan Wright and Jon Runyan: Armada founders. 00:42:48

Alex and Justin (137 partners): Originated the Armada idea. Justin's line: "A bad investor can destroy your company; a good investor won't make your company." 00:47:22

Teddy Forstmann: Garrett's godfather, 1980s private equity figure, source of the "quality at a 25% premium" quote. 00:24:19

Shardul (Index Ventures): Quoted on paying a 25% premium. 00:23:12

Philip: Starcloud CEO. 00:08:25

Ashley: 137's operating partner who works embedded in portfolio companies. 00:44:43

Trae (Stephens): Garrett's mentor; "has co-founded a company" and talks about the intellectual depth of operating. 00:46:09

Josh (Kushner): Thrive partner who co-founded a company. 00:46:09

Jacob: Co-organizer of Hill & Valley with Delian and Garrett. 00:59:53

5. Operating Insights

Build the Common Platform First; Every Added Product Increases the Value of the Others

Garrett's reasoning on why Anduril's multi-product model compounds: "they've actually built the same core back-end infrastructure, right, from a software and AI perspective, right, with Lattice, which powers kind of everything. And so every marginal product sold, right, makes the value of the other ones go up." 00:30:42 The operating lesson is to invest in a shared platform before diversifying, so each new line has increasing returns instead of the diseconomies that sank old conglomerates.

Identify Bottlenecks, Not Demand, When Picking Where to Operate

Garrett's acquisition screen: "Where are the bottlenecks for this build-out? The contracts have been going to all the primes, but you've been seeing that production is the big bottleneck right now, and a lot of that's because of the supplier level. Whether it's their ability to scale, right, who's going to take the risk on CapEx, right? How do we modernize their operations to deliver faster?" 00:40:34 Funding follows contracts, but constraints sit upstream.

Hire a Deliberately Mixed DNA: PE, Government, Tech, and Operators

Onodrim's team composition: "folks that come from private equity, traditional aerospace and defense, government, and the technology sector... people from Palantir, SpaceX, Rheinmetall, McKinsey... nine nationalities." 00:39:03 The point is combining capital-allocation discipline, domain credibility, and software-deployment skill in one team.

Tell Investors What You Actually Need: Revenue and Talent

After seeing both sides, Garrett says founders really want two things from investors: "I need revenue, and I need talent, right? And it's like, can you help with these? And a lot of investors promise that they'll help with things, and they really don't." 00:47:22 Make explicit asks about customers and hires, and judge investors on delivering those.

Screen Acquisition Targets on Revenue Quality, Not Just Growth

Garrett's criteria for suppliers: "What's the quality of the revenue, right? Have they been doing this for a long time? Is there low churn, right? Which just shows that this is a high-quality business. Do they have capacity that's critical... even if majority of the revenue is coming from aerospace or industrials?" 00:39:46 He also wants a defense-growable business line within an otherwise commercial company.

6. Overlooked Insights

Sub-12-Month Payback on AI Compute Combined with Fast-Build Capacity Quietly Changes the Risk Profile of the Entire AI CapEx Debate

Garrett mentions it in a single run-on stretch, but it reframes the bear case. If payback is under 12 months and build cycles are short, then "if you build these things fast, you can pull back on these things really fast, right? And so your CapEx investments are on shorter duration cycles." 00:03:33 The standard fear is stranded multi-year assets; this implies capacity can be recouped (or redirected to internal products) before demand uncertainty bites, a structural advantage that slower-building neoclouds and debt-dependent players lack.

The Supplier Layer Is Where Both European Rearmament and US Capacity Fragility Live, and It's Been "Repatriated Away"

In a throwaway line, Garrett notes that the issue isn't just funding: "the sad part is a lot of the capacity is repatriated away, right? And it no longer exists in Europe. And you see the same dynamic in the US, right? They both exist." 00:40:34 This implies some critical capabilities are physically gone, not merely under-scaled, so rebuilding them requires acquiring and modernizing what remains. That turns the roll-up from a financial arbitrage into a scarcity play, with the surviving dual-use suppliers becoming increasingly strategic assets for both the European and US defense industrial base.