BREAKING: The Return of Conglomerates | SpaceX, Anduril & Onodrim
1. Key Themes
Theme 1: SpaceX is evolving from launch provider into a vertically integrated hyperscaler
SpaceX monetizes AI across the full stack
Garrett argues SpaceX is building software and services around its compute stack, using capacity internally and selling it externally.
"The company is clearly going the route of a hyperscaler. They're building out more software and services around that compute stack, new infrastructure."
Vertical integration hedges against model-market uncertainty
Whether open-source or closed models win, the compute layer captures demand, as Google Cloud and AWS have shown. SpaceX can also redirect capacity to its own products if outside demand slows.
"If you build these things fast, you can pull back on these things really fast. And so your CapEx investments are on shorter duration cycles."
Starship and Starlink V3 open new growth curves, including orbital compute
Over 50% of SpaceX's revenue is tied to AI, but launch and connectivity are entering new growth phases. Orbital data centers are framed as the answer to terrestrial power constraints.
"That launch on just a handful of Starlink V3s delivered like 1% of the world's internet capacity."
"Starship will not just drive launch costs down an order of magnitude, which probably then will lead to an equal rise in the number of companies that can be built."
Theme 2: The return of the (tech-enabled) conglomerate
Modern conglomerates share a common operating platform
The 1980s conglomerates held unrelated businesses with no shared platform, which made them hard to value and inefficient. Today's versions are different.
"Today's conglomerates look different. One, they actually have a core common operating tech platform, and so there actually is synergies and operations between them."
Anduril as a "lowercase c" conglomerate; Palantir and SpaceX as "capital C" dual-use conglomerates
Anduril's products all run on Lattice, so each additional product raises the value of the others. Garrett sees Palantir and SpaceX as the only current dual-use conglomerates, and both now earn most revenue commercially.
"I think we need to return to the dual use conglomerates. General Mills made cereal and weapon systems."
China's industrial structure is the benchmark
Garrett notes that Chinese defense primes earn 60% to 80% of revenue from commercial sectors, and that China's leading naval shipbuilder is also the world's dominant commercial shipbuilder. Commercial scale lets companies flex defense capacity up or down, which he argues industrial policy alone cannot fund.
Theme 3: Production capacity, not budgets, is the binding constraint on European rearmament
The supplier layer is the underfunded bottleneck
Capital has flooded into neo-primes and drone makers, while the layer below (component suppliers, aerostructures, engines, services) has been neglected and remains unconsolidated, unlike the US.
"Now there's 4,000 drone companies in Europe. This is overfunded and too crowded. What has not received as much support is the layer below. Who's supplying all these companies?"
Onodrim's thesis: buy, hold with permanent capital, and modernize
Onodrim acquires controlling stakes in European aerospace, defense and industrial manufacturers, funds their capex, and deploys technology into operations. It favors dual-use businesses with durable, low-churn revenue.
"Does it solve a critical bottleneck? That's something we really focus on. Where are the bottlenecks for this build-out?"
"We think that supporting the reindustrialization rearmament in Europe is the most important mission right now."
Theme 4: Duration of compounding matters more than headline growth rates
137 underwrites how long a company can keep growing
As a minority investor with no control over exits, 137 concentrates in businesses it expects to hold for decades. SpaceX has compounded at roughly 40% a year for 20 years.
"9% over 30 years is like a 13X. It's incredible how really duration is what matters the most."
Cost of capital and payback speed shape the AI build-out
Most data center build-out is debt-funded, so companies with a lower cost of capital win demand at better rates, while unresolved variables make long-term earnings hard to underwrite.
"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 because it's the most efficient form of capital to go do it."
"It's hard to believe in the next two years of earnings, let alone the next 20."
Theme 5: Software and technology are moving into legacy industrial manufacturing
From Palantir to Hadrian to Applied Intuition
Garrett sees faster integration with PLCs opening up work beyond paperwork automation, and points to Applied Intuition simplifying vehicle compute and wiring across automotive, construction, mining and trucking, and General Matter rebuilding uranium enrichment capacity.
"Just because you deploy software into a business doesn't necessarily make it a technology business. But it has real impact."
2. Contrarian Perspectives
Investors overpay for quality, and that's fine
Garrett cites his godfather, private equity investor Teddy Forstmann: investors routinely overpay while claiming to pay fair prices. Rather than hunting discounts, Garrett would pay a 25% premium for a quality asset before buying a weaker one cheaply. The evidence cited is the portfolio's steep markups (Cognition at $48B and reportedly raising at $60B-$70B, Ramp at $60B, Anduril at $90B) alongside the claim that Ramp still underprices several products.
"9% over 30 years is like a 13X. It's incredible how really duration is what matters the most."
The fintech graveyard wasn't wasted capital
Against the narrative that COVID-era fintech investment was a bubble, Garrett argues power law outcomes cover the losses. Most companies shut down, but Stripe, Revolut, Nubank and Ramp produced or are on track for $100B+ outcomes.
"The Fintech category produced massive wins that have performed extremely well and that are all doing billions and billions of dollars of profit. So that capital was not wasted."
Tech's "non-zero-sum" culture may just be a function of high growth
Garrett offers an unusually pessimistic read of the industry's collaborative ethos, suggesting it is not virtue but circumstance.
"Probably the pessimistic take is the only reason we are not zero sum is because we're in a high growth industry, and many of these industries are not."
3. Companies Identified
SpaceX: Launch, connectivity and AI/compute company, valued around $2T. Why mentioned: 137's largest position after 16 years, and the central case study for hyperscaler and conglomerate evolution.
"It's still the company we are most excited about, and I think the most important company in the world."
Anduril: Defense technology company, valued at $90B. Why mentioned: Archetype of the "lowercase c" conglomerate built on the shared Lattice platform.
"Garrett calls Anduril a 'lowercase c' conglomerate, meaning a multi-product defense company comparable in structure to the largest primes."
Onodrim Industries: European defense and industrial platform that raised a $300M Series A at a $1.3B valuation, led by Lux Capital. Why mentioned: Garrett is co-founder; it targets Europe's fragmented supplier layer.
"Onodrim acquires controlling stakes in aerospace, defense and industrial manufacturing businesses across Europe and holds them with permanent capital."
Palantir: Software and AI company. Why mentioned: A "capital C" dual-use conglomerate, pioneer of software in legacy manufacturing, and the origin of Onodrim's founders.
"Palantir pioneered putting software into legacy manufacturing."
Cognition: AI coding company, last valued at $48B and reportedly raising at $60B-$70B. Why mentioned: Major 137 AI investment at the harness/app layer.
"Cognition was most recently valued at $48B and is reportedly raising at $60B to $70B."
Ramp: Finance automation platform valued at $60B. Why mentioned: 137 portfolio company that keeps consolidating the back-office finance stack.
"Ramp keeps consolidating more of the back-office finance stack and still underprices several products."
Hex: Analytics platform. Why mentioned: 137 recently co-led its round; part of the thesis that customers want to route across models.
"The firm has invested heavily in Cognition and recently co-led Hex's round."
Hadrian: Advanced manufacturing company. Why mentioned: Shows what modern tooling can do in industrial production.
"Hadrian has since shown what modern tooling can do in industrial production."
Applied Intuition: Vehicle software company. Why mentioned: Extending beyond automotive engineering software into vehicle compute simplification across industries.
"It is also simplifying vehicle processing units, controllers and wiring across automotive, construction, mining and trucking."
General Matter: Uranium enrichment company. Why mentioned: Rebuilding capacity currently held by state-owned operators on Cold War-era technology.
"General Matter, another portfolio company, is rebuilding uranium enrichment capacity that currently sits mostly with state-owned operators running Cold War-era technology."
Varda: Space manufacturing company serving pharma and defense. Why mentioned: Example of a company built on Falcon 9's lower launch costs.
"Companies such as Varda, a 137 portfolio company serving pharmaceutical and defense customers, were built on top of those lower costs."
Armada: Edge infrastructure company incubated with 137 partners' involvement. Why mentioned: Model for how Onodrim was formed.
"Alex and Justin pitched an idea to SpaceX... and Dan Wright and John Runyon went on to found Armada."
Amca and Valinor: US companies. Why mentioned: Working on other parts of the defense supplier problem.
"Garrett points to Amca and Valinor as US companies working on other parts of the same supplier problem."
StarCloud: Orbital compute startup. Why mentioned: Developing orbital data center satellites designed to fly on Starship.
"StarCloud is developing its own satellites, with StarCloud 3 designed to fly on Starship."
Helsing, Stark, Quantum Systems: European neo-primes. Why mentioned: Recipients of most new European defense funding.
"Most new funding has gone to neo-primes such as Helsing, Stark and Quantum Systems, along with a crowded field of drone makers."
TransDigm, Teledyne, HEICO, Honeywell: US supplier consolidators. Why mentioned: Evidence of the consolidation playbook Europe lacks.
"Supplier consolidators such as TransDigm, Teledyne, HEICO and Honeywell... have rolled up much of the defense supply base."
Stripe, Revolut, Nubank: Fintech winners. Why mentioned: Proof that fintech's power law outcomes justified the capital invested.
"Stripe, Revolut, Nubank and Ramp produced outcomes of $100B or more, or are on track to."
OpenAI and Anthropic: Frontier AI labs. Why mentioned: Garrett calls them among the most impressive businesses built in a generation.
"Garrett calls OpenAI and Anthropic two of the most impressive businesses built in a generation."
Google Cloud and AWS: Hyperscalers. Why mentioned: Evidence that compute businesses grow regardless of model mix.
"He points to Google Cloud and AWS, which have grown regardless of the mix between model types."
4. People Identified
Christian Garrett: Partner at 137 Ventures, 16-year SpaceX investor, Onodrim co-founder. Why mentioned: The episode's guest and source of the conglomerate and duration theses.
"If there's a great idea, most times you're not the person to go do it. If it's a weird case where you're the person to go do it, then by all means, go and do it."
Aistis Šimaitis: Onodrim CEO, Palantir's second employee in Europe. Why mentioned: Runs Onodrim day to day.
"CEO Aistis Šimaitis and CTO Alexander Blessing lead the company. Both come from Palantir."
Alexander Blessing: Onodrim CTO, ex-Palantir. Why mentioned: Technical leader of Onodrim.
"Šimaitis and Blessing run Onodrim."
Elon Musk: SpaceX leader. Why mentioned: Garrett endorses his point that lower-cost-of-capital companies will capture demand.
"Garrett agrees with Elon Musk's recent point that companies with a lower cost of capital will take a large share of demand at better rates."
Teddy Forstmann: Garrett's godfather and 1980s private equity investor. Why mentioned: Source of the "investors routinely overpay" insight.
"Garrett cites his godfather, Teddy Forstmann, a private equity investor in the 1980s: investors routinely overpay while claiming to pay fair prices."
Justin Fishner-Wolfson: 137 Ventures partner. Why mentioned: Origin of a founder-investor maxim Garrett repeats.
"A bad investor can destroy your company. A good investor won't make your company."
Alex (137 partner) and Justin: 137 partners who helped incubate Armada. Why mentioned: Precedent for Onodrim's formation.
"Onodrim was formed in a similar way to Armada, which Garrett's 137 partners Alex and Justin helped incubate."
Dan Wright and John Runyon: Armada founders. Why mentioned: Took the idea SpaceX passed on and built Armada.
"Dan Wright and John Runyon went on to found Armada."
Shaun Maguire: Investor. Why mentioned: Published his own SpaceX financial model.
"Shaun Maguire has published his own SpaceX financial model."
Brandon Reeves: Partner, Lux Capital. Why mentioned: Led the Onodrim round and articulated the investment case.
"Europe needs a stronger industrial foundation to meet the security challenges ahead."
Jeannette zu Fürstenberg: President & Managing Director, General Catalyst. Why mentioned: Onodrim investor, framing the sovereignty argument.
"Every country needs to be in control of its own destiny, and that starts with what it can build."
Alex Karp: Palantir CEO. Why mentioned: Source of a lesson passed on by Šimaitis (referenced in an embedded post).
"Focus on being important & valuable, & the money will come."
Delian Asparouhov and Jacob Helberg: Hill & Valley Forum collaborators. Why mentioned: Co-working on forthcoming announcements.
"Garrett, Delian Asparouhov and Jacob Helberg have additional announcements in the works."
5. Operating Insights
Build capacity fast so you can pull back fast
Short-duration capex cycles reduce risk, especially when you can redeploy capacity internally if external demand softens.
"If you build these things fast, you can pull back on these things really fast. And so your CapEx investments are on shorter duration cycles."
Remember what founders actually want from investors: revenue and talent
Many investors promise help here and fail to deliver. 137 embeds an operating partner in portfolio companies, and founders should be skeptical of generic "value-add" claims.
"A bad investor can destroy your company. A good investor won't make your company."
Target the bottleneck, and fund the capex risk others won't
For an industrial rollup, Onodrim's playbook is to find the critical constraint, take on the capital spending risk suppliers cannot, and add engineering and technology into operations to scale businesses pan-European and global.
"Does it solve a critical bottleneck? That's something we really focus on. Where are the bottlenecks for this build-out?"
6. Overlooked Insights
Direct-to-cell is growing faster than Starlink's core connectivity
Buried in the Starship discussion is a signal that SpaceX's mobile-network ambitions may be a bigger growth driver than the headline broadband product.
"Starlink's direct-to-cell business is growing faster than the core connectivity product."
Political attacks on tech companies are becoming a risk factor
Garrett flags boycotts and attacks tied to founders' politics from both sides, despite these firms being major employers and critical to national security and GDP growth.
"These companies are working on things that are nonpartisan. They're benefiting the entire economy in the world."