New SPVs open AI startup investing to smaller investors
Michael, a retired dentist in Ohio, had spent decades building a comfortable nest egg. He wasn’t a Wall Street wolf, just a man who wanted his savings to grow. For years, his financial advisor had steered him toward blue-chip stocks and municipal bonds. Then, in early 2025, a friend mentioned something called a “special purpose vehicle” that could get him into companies like SpaceX and Anthropic. Michael was intrigued but skeptical. He wasn’t a venture capitalist, and these deals seemed reserved for the Silicon Valley elite. Yet, within weeks, he was a shareholder in a quantum computing startup, his money parked alongside that of billionaires and university endowments. He felt lifted—part of an exclusive club. But he didn’t fully grasp the machinery that made it possible.
For decades, the world of late-stage startup investing was a gated community. Only the largest venture capital firms, sovereign wealth funds, and a handful of ultra-high-net-worth individuals could buy shares in companies like SpaceX or Databricks before they went public. The rest of the financial world—family offices managing a few hundred million, smaller institutional investors, and certainly individuals like Michael—were locked out. The gatekeepers were the VC funds themselves, which raised capital from limited partners (LPs) over 12 to 18 months, then deployed it over a decade. This system was slow, hierarchical, and exclusive by design. It rewarded patience, connections, and a tolerance for risk that most small investors couldn’t afford.
Then came Justin Ernest, a former investor at Playground Global who saw a gap that was both a problem and an opportunity. Family offices—wealth management firms for ultra-rich families—and smaller institutions were desperate for access to AI and defense-tech startups. But the traditional VC fund model was too slow for these fast-moving companies. Ernest didn’t launch a fund. Instead, he used his network to secure allocations of stock in high-profile, later-stage companies like Anthropic and Anduril. He then offered these individual deals to a group of about 30 smaller institutional investors using special purpose vehicles, or SPVs. These are essentially single-deal funds: investors buy shares in a vehicle that owns the stock, bypassing the need for a formal fund structure. Over the last 12 months, his firm, Sabertooth Capital, has invested nearly $500 million into 10 companies, including Anthropic, Anduril, Base Power, Databricks, PsiQuantum, and SpaceX. [2] He’s writing checks ranging from $10 million to $275 million, always participating in official, company-approved funding rounds.
The New Middleman
This is where AI lifts. The technology itself isn’t the direct tool here—it’s the demand for it that has created a new kind of financial intermediary. AI companies, especially those building foundational models, require enormous capital. They also need to control who holds their shares. Unauthorized SPVs—where shares are sold without company approval—have become a headache for startups like Anthropic and Anduril, which have cracked down on them. But Sabertooth operates differently. It secures allocations directly from the companies, meaning the startups themselves vet and approve the investors. This gives smaller LPs like Michael peace of mind: they know their money is going through a legitimate channel, not a “fly-by-night” operation.
Ernest’s approach is a masterclass in network leverage. “I’ve always found that my sort of superpower is being the nucleus of my network,” he told TechCrunch. [2] He can obtain investor capital for a new SPV on a tight timeline—often just four or five phone calls. “I have a captive set of LPs,” he said. “I can usually make four or five or six phone calls, and I know exactly what my LPs will commit.” This speed is crucial in a market where AI startups raise rounds in weeks, not months. Sabertooth has already had one major return: chipmaker Groq, which was licensed and acqui-hired by Nvidia for $20 billion late last year. [3] Next up is SpaceX’s highly anticipated IPO, along with Anthropic’s expected public listing. These could deliver windfalls for his investors.
But this system also deceives. The SPV structure, while efficient, obscures risk. Investors like Michael see a chance to own a piece of the AI revolution, but they may not fully understand that they are buying into a single company’s stock, not a diversified fund. If that company fails, their investment is gone. Traditional VC funds spread risk across dozens of startups. SPVs concentrate it. Moreover, the fees can be opaque. SPV managers typically charge a carried interest—often 20% of profits—on top of management fees. For a $10 million check, that’s a significant bite. And because these vehicles are not regulated like mutual funds, investors have less recourse if something goes wrong. The very speed and exclusivity that make SPVs attractive also make them dangerous for the unwary.
The Unseen Hand
The deeper deception is in how AI itself becomes the product being sold. These startups—Anthropic, Anduril, PsiQuantum—are not just companies; they are bets on a future shaped by artificial intelligence. Investors are buying into a narrative of inevitable progress, where AI will solve everything from climate change to cancer. But that narrative is curated. Founders and VCs have a vested interest in painting a rosy picture, downplaying the technical hurdles, regulatory risks, and ethical dilemmas. The SPV structure, by bypassing traditional due diligence processes, can amplify this information asymmetry. A family office CIO like Benjamin Wagner, who manages wealth for 50 individuals, might have the expertise to evaluate a quantum computing startup. But Michael the dentist? He relies on the reputation of the intermediary. “Justin is authentically an investor,” Wagner said. “He has judgment, he has expertise, he’s very technical.” That trust is essential, but it’s also a single point of failure.
This is where AI also makes us superfluous. The SPV model, powered by the demand for AI, is essentially a disintermediation of the traditional VC fund. It cuts out the general partner who would have managed a diversified portfolio, replacing them with a deal-by-deal approach. But it also creates a new kind of gatekeeper—the networker who can aggregate capital and secure allocations. Ernest’s success depends on his personal relationships, not on a systematic process. If he were hit by a bus, the SPVs would likely dissolve. The model is fragile because it’s built on one person’s reputation and connections. In a world where AI could theoretically match investors with startups algorithmically, why do we need a human middleman at all? The answer is trust, but that trust is a brittle thing.
The Risk of the Golden Ticket
The historical arc is clear. Venture capital began as a cottage industry in the 1950s, with firms like American Research and Development funding early tech companies. It evolved into a professionalized asset class, with LPs committing capital for a decade. The SPV model represents a regression to an earlier, more informal era—one where deals are done on handshakes and phone calls. But it’s also a response to the speed of AI. Traditional funds take too long to raise and deploy. In a market where a company like Anthropic can go from zero to a $60 billion valuation in a few years, speed is everything. Ernest’s approach is a pragmatic hack, but it’s also a gamble.
The caution is this: what happens when the market turns? SPVs are illiquid. Investors cannot sell their shares until the company goes public or is acquired. If the AI bubble bursts, those shares could become worthless. And because the SPVs are not diversified, a single failure could wipe out an investor’s entire stake. The system works brilliantly in a bull market, but it has never been tested in a downturn. Regulators are not watching closely. The SEC has focused on unauthorized SPVs, not the legitimate ones. The startups themselves are too busy growing to police their cap tables. The family offices are eager for returns. And the intermediaries are making money on every deal. Everyone has an incentive to keep the machine running—until it does not.
Michael, the retired dentist, might not care about any of this. He’s excited to own a piece of SpaceX. He feels lifted. But he’s also been deceived—not by a fraud, but by a system that makes risk invisible. And he’s been made superfluous: his money is welcome, but his voice is not. He has no say in how the company is run, no board seat, no information rights. He’s a passenger on a rocket he can’t steer. The question is not whether AI will change the world. It is whether the financial machinery built around it will lift everyone, or just those who already know how the game is played. And who is watching when the music stops.
Sources
2. TechCrunch
3. Nvidia