Sabertooth Capital has deployed close to $500M into 10 hot startups in 12 months using SPVs and nominee structures, with no traditional fund. The architecture that gives it speed also concentrates fee layering, allocation discretion, and oversight risk inside a single firm.
TechCrunch's reporting on Sabertooth Capital surfaces a word that the people running the hottest AI cap tables use for the alternatives: "shady." Justin Ernest, who founded the firm after roughly five years at Playground Global, treats that description as a market opportunity. Over the last 12 months his firm has deployed close to $500 million into 10 high-profile startups, according to TechCrunch's Marina Temkin, with no traditional fund in the middle.
The portfolio cuts across the AI and deep-tech names that the open market has effectively closed: Anthropic, Anduril, Base Power, Databricks, PsiQuantum, and SpaceX. Check sizes reported by TechCrunch run from roughly $10 million to $275 million. Each deal is its own vehicle, and Sabertooth serves as the gatekeeper for a group of about 30 smaller institutional investors who want a seat at tables they cannot reach on their own. The structure is the story. It is also the part of the story that warrants the most scrutiny.
Sabertooth was built to solve a specific access problem. Family offices and smaller institutional allocators have wanted exposure to the fastest-growing private companies for years, but the rounds that matter most are typically filled by a small group of lead investors before anyone else can ask. The traditional response, raising a fund, takes a new manager 12 to 18 months to close and a fund cycle to deploy, by which time the names that justified the raise have moved on. Ernest's pitch is that he can move faster because he already had the relationships.
The mechanism Sabertooth uses, as described to TechCrunch, combines two distinct vehicles. For some deals, Sabertooth packages a special purpose vehicle, an SPV, that holds the single asset and passes exposure to the participating investors. For others, Sabertooth takes a different path. The firm holds shares in its own name on behalf of the participating investors, who do not appear on the cap table at all. In TechCrunch's framing, the second path is the nominee structure. The distinction matters. An SPV gives each investor a clean ownership line to the asset. A nominee structure does not. The investor's claim runs through Sabertooth, not through the company they think they own a piece of.
That load-bearing difference is also the part of the story that has had the least independent pressure-testing so far. The deal count, the dollar volume, the named cap tables, and the size of the investor base are all Sabertooth's own figures, repeated by TechCrunch on the firm's description. They have not been confirmed by the companies Sabertooth has invested in, by the LPs in its investor base, or by any filing. For a story about access and trust, the access numbers being self-reported is worth saying out loud.
The risks in the model are structural, not personal. They sit in the architecture.
Fee stacking is the first one. With single-asset vehicles, each deal can carry its own management fee and carry arrangement. If an LP writes ten separate checks across a year, the LP is paying Sabertooth's overhead and incentive layers on each check, not on a diversified fund. The TechCrunch reporting does not disclose Sabertooth's fee or carry terms. The fact that the firm is also the gatekeeper to the deals means there is no third party repricing the manager for the LP.
Allocation discretion is the second. The TechCrunch reporting describes Sabertooth as the source of deal flow into the structure, with the named participating investors receiving what Ernest can negotiate. A firm that both picks the deals and holds the nominee rights to those deals is the only party that can move positions, handle secondaries, or vote on corporate actions. The LP does not see the cap table, and the LP does not have a direct relationship with the company. Whether that discretion is exercised well is a judgment call the LP has to take on Sabertooth's word.
Nominee conflicts sit underneath the discretion problem. If Sabertooth holds shares on behalf of multiple investors across a single company, and the company later does a down-round, a structured secondary, or an acquisition, Sabertooth is the one negotiating the outcome on behalf of parties whose interests may not align. The LP has no seat at that table. The LP often has no visibility into it.
The regulatory gap is the part the source flags without fully pressing. The "shady" framing in TechCrunch's reporting is the source's own word for what happens when allocators move outside the regulated fund structure: nominee accounts at multiple brokers, transfers that may not be reflected in the company's records, and KYC and AML procedures that vary by vehicle. Some of those risks are well documented across the broader SPV market. Others are overstated. The question Sabertooth raises is whether the structural protections that come with a registered investment adviser, audited books, and a fund's fiduciary overlay are features the model is replacing, or friction the model is designed to remove.
The demand side, at least as TechCrunch reports it, is not in doubt. A participating family-office CIO is quoted endorsing the structure and the speed of access. The buyer being happy is the easiest part of the story to source, because the buyer got what they wanted. Whether they got a good price relative to the lead investors in the same round, whether the fees they paid compounded against a hypothetical fund return, and whether the nominee structure has held up under any stress event are questions that require more reporting than one trade-press article.
For now, Sabertooth's model is a useful stress test of how much structure investors will trade for speed and access. The market for hot AI cap tables is large enough that there is room for both traditional funds and Sabertooth-style allocators. The open question is whether the LP end of the deal has the same information about the structure that the gatekeeper does. On the public evidence, that gap is the part of the trade the buyer is not being quoted on.