Image credit: Image via TechCrunch — AI. Used under fair use for news commentary. · source
Is this just a smarter way to do venture capital, or is it the same old game with a new coat of paint?
I've been watching money move through tech since the dial-up era, and I'll tell you, the structure changes but the instinct never does. Find the hot deals. Get your check in. Figure out the paperwork later. That's been the Silicon Valley playbook since before most of the people currently running AI funds were in middle school. What Justin Ernest did with Sabertooth VC is interesting, genuinely interesting, but let's not pretend it's without precedent.
Here's what actually happened. According to TechCrunch, Ernest invested nearly $400 million into a portfolio that includes some of the most talked-about names in the current tech cycle, Anthropic, Anduril, SpaceX. He did it without raising a traditional venture fund. No year-long roadshow, no LP meetings at the Four Seasons, no waiting around for a first close. Instead, he built what TechCrunch describes as a captive network of LPs, essentially a pre-assembled group of investors he could call on deal by deal.
The number, by the way, is worth flagging. TechCrunch's own reporting has it at two different figures depending on which URL you hit, one version says nearly $400 million, another says nearly $500 million. The company didn't clarify the discrepancy publicly, at least not as of when I'm writing this. That's not nothing. A hundred million dollars is not a rounding error. It remains unclear which figure is accurate, and it's the kind of detail that matters when you're evaluating the actual scale of the operation.
Related coverage
More in Autonomy
Researchers want large language models to rewrite the cost functions that govern how self-driving cars move. Bob Macintosh has some thoughts.
Robert "Bob" Macintosh · 13 mins ago · 4 min
A pair of fresh arXiv preprints tackle the tension between real-time planning and honest uncertainty in self-driving systems. Neither is a silver bullet, but the ideas are worth examining carefully.
Aisha Patel · Yesterday · 8 min
A new framework from arXiv claims to give monocular cameras the spatial precision of LiDAR. The approach is technically interesting, but the real test is whether it holds up outside a lab.
James Chen · Yesterday · 7 min
But back to the model itself. The traditional venture fund structure is, honestly, a bit of a dinosaur. You spend twelve to eighteen months raising capital, you pay lawyers to set up a Delaware LP, you negotiate management fees and carry with people who want to renegotiate management fees and carry, and by the time you've closed the fund, the best deals you wanted to lead in Round One have already been priced up by three other firms who moved faster. Ernest apparently looked at that process and decided it wasn't worth it. So he built something more like a syndicate on steroids, a standing group of limited partners who trust his deal flow enough to wire money when he calls.
I've seen this movie before. The SPV model, the rolling fund, the scout fund, the founder-friendly micro-fund, every few years someone announces they've reinvented venture capital and the press covers it like it's a moon landing. Some of those experiments worked. AngelList's rolling funds genuinely changed how smaller managers got off the ground. Some didn't. The graveyard of "VC 2.0" concepts is longer than people remember.
What's different here, or what appears to be different, is the size. Getting $400 million (or $500 million, pick your number) deployed through a non-traditional structure is not a small feat. That's not a scout writing $25,000 checks into seed rounds. That's real money getting into Anduril, which is a defense tech company now valued in the tens of billions, and Anthropic, which has raised somewhere north of $10 billion total. To get allocation in those deals without a traditional fund behind you, you either have extraordinary relationships, extraordinary timing, or both.
The LP network question is the one I keep coming back to. A captive network sounds elegant in a press write-up, but what does it actually mean in practice? Who are these LPs? Are they family offices? High-net-worth individuals? Other funds co-investing? The sourcing and structure of that network matters a lot for whether this is a repeatable model or a one-time trick that worked because Ernest happened to know the right people at the right moment. The TechCrunch piece, good as it is, doesn't fully answer that question. This is based on limited public disclosure, and the mechanics of how the LP relationships are structured, legally and economically, aren't spelled out.
That matters for anyone thinking about copying the playbook. And there will be people trying to copy it, there always are. Some young founder or aspiring fund manager is reading the same article right now and thinking, great, I'll just build a captive LP network too. Good luck! Building that kind of trust-based capital network takes years of relationship work, probably a track record from somewhere else, and the kind of access to top-tier deals that is itself a product of already being in the room. It's a bit circular, is what I'm saying.
Let's talk about the portfolio for a second. Anthropic, Anduril, SpaceX. That's a deliberately impressive sentence. It's also a sentence that covers a lot of ground ideologically. Anthropic is an AI safety company, founded by former OpenAI researchers who were, in theory, worried about moving too fast. Anduril is a defense contractor that makes autonomous weapons systems and surveillance technology and is very explicitly not worried about moving fast. SpaceX is SpaceX. Putting all three in the same portfolio says something about Ernest's investment thesis, which seems to be less "I have a specific view on the future" and more "I want to be in the deals that matter, whatever they are."
That's not a criticism, necessarily. Some of the best investors I've watched over the years were generalists who bet on founders and momentum rather than sector theses. But it does raise questions about, well, multiple things, including how an investor without a formal fund structure does the kind of due diligence and governance work that institutional LPs usually demand. When you're in a traditional fund, there are compliance structures, advisory boards, quarterly reports. With a captive LP network, the accountability mechanisms are less obvious, at least from the outside.
Now, is any of this relevant to the autonomous vehicle and robotics world specifically? More than it might seem. The companies in Ernest's portfolio aren't just abstract tech bets. Anduril's autonomous systems work is directly adjacent to the robotics and AV space, they're building autonomous platforms for military applications that use the same sensor fusion and decision-making architectures that civilian AV companies have been developing for a decade. Anthropic's models are increasingly being evaluated for robotics applications. SpaceX's Starlink is infrastructure that matters for connected vehicle systems in rural and remote environments.
The money flowing into these companies shapes what gets built and how fast. And the way that money flows, through traditional funds with institutional oversight or through more informal captive networks, shapes who has accountability for the outcomes. Call me old-fashioned, but I think that question matters more as the systems being funded start making real-world decisions at speed.
The broader point, and I'll wrap up here because I've been going longer than my editor likes, is that the venture capital industry is genuinely in a weird moment. The traditional fund model is under pressure from multiple directions. Rolling funds democratized access for smaller managers. Mega-funds from Andreessen and Sequoia absorbed more and more of the institutional LP capital. Corporate venture arms from the big tech companies outcompete traditional VCs on valuations because they don't need returns in the same way. And now you've got people like Ernest building captive networks that sidestep the whole structure entirely.
It's too early to say whether the Sabertooth model is the future or a footnote. The track record is real, those are genuinely consequential companies in that portfolio, but one cycle of good deals doesn't make a durable institution. The test is whether the model holds up when the market turns, when LPs get nervous, when the hot deals of 2024 and 2025 start showing their complications. Every structure looks smart in a bull market.
I've been wrong before about which experiments stick. I was skeptical about rolling funds for longer than I should've been. So take my skepticism here with appropriate salt. But I'd want to see a lot more transparency about the LP network mechanics before I'd call this a new model for the industry rather than a smart individual finding a path that worked for him specifically.
If you want to argue about it, my email's on the about page.
New research from NASA JPL and university labs shows reinforcement learning can teach rovers to handle loose soil without getting stuck, cutting energy use by 37% on sandy slopes.
