The number that should stop you is not $30 million. It is $7 billion.
Erebor Bank, National Association, received its national banking charter in February 2026. By the time Robinhood Ventures Fund I (NYSE: RVI) disclosed a $30 million purchase of Erebor preferred stock, the bank's deposit base had crossed $7 billion. That is roughly a billion dollars a month for a de novo institution.
De novo banks in the United States typically spend three to five years grinding toward their first billion in deposits. Erebor cleared seven in under a year. When a growth curve that steep shows up in something I am auditing, I stop reading the press release and start looking for the funding source. Deposits do not materialize. They are placed. And placements leave traces.
Erebor was founded in 2025 by Palmer Luckey and a group of backers including Valor Equity Partners, Lux Capital, and 8VC. Its stated mandate is the innovation economy โ defense, robotics, artificial intelligence, hard tech. Not consumer deposits. Not crypto custody. Operating accounts for companies that build things with long procurement cycles and classified adjacency.
One precision note before anything else, because the coverage has been sloppy: the Office of the Comptroller of the Currency grants the charter. The Federal Deposit Insurance Corporation grants the insurance. Two agencies, two statutes, two sets of examiners. Reporting that collapses them into federal deposit insurance from the OCC is the kind of error that tells you the writer never opened the approval order. Small thing. Structural thing.
Robinhood Ventures Fund I is the first vehicle under Robinhood Ventures, listed on the New York Stock Exchange in March 2026. Its pitch is exposure to private companies through a public wrapper โ private equity for people who do not qualify for private equity. The Erebor position is preferred stock, not common.
Robinhood's interest is not mysterious. Its retail base wants private-market access it cannot legally reach, and its crypto arm has spent years learning how much of that demand is really a demand for early entry. A chartered bank with a hard-tech deposit base is a supply source for that demand.
Start with the instrument, because the instrument is the argument.
Preferred stock in a de novo bank is not a growth bet. It is a capital compliance instrument. Newly chartered national banks operate under heightened capital and leverage expectations through their early years. Regulators want a buffer before the balance sheet runs. Preferred equity, depending on structure, can count toward regulatory capital. So when a fund buys $30 million of Erebor preferred, the honest read is that it is funding the ratio, not the roadmap. Growth capital buys engineers. Capital-stack capital buys the right to keep the deposit door open.
That reframes the deal. Robinhood did not invest in a bank's future. It invested in a bank's present compliance position, and it took a senior claim to do it.
Then the deposits. Seven billion dollars in operating balances from defense, robotics, and AI firms is a specific kind of money. Treasury money โ payroll, procurement, contract milestones. Sticky in the way that switching a primary operating account is sticky. It is also, almost certainly, concentrated. A de novo bank does not accumulate $7 billion across ten thousand accounts in a year. It accumulates it across a handful of very large ones. Five contractors relocating treasury operations is not a deposit franchise. It is a warehouse with a charter bolted to the door.
I have seen this shape before. In 2022 I spent three weeks reverse-engineering Anchor Protocol's incentive loop, and the tell was never the headline yield. It was the concentration of the funding source behind it. The mechanism looked distributed. The money was not. Yield is a symptom, not the cure โ and deposit growth is a symptom too. The question is always who is placing the funds, and why they chose this venue over the eleven other places they could park a treasury.
The answer here is legible. Hard-tech firms in defense and robotics have a banking problem that crypto solved conceptually and banks solved legally: fast settlement, programmatic treasury operations, and a counterparty that understands the sector. Erebor sells the second and third, and inherits the first from the pressure crypto put on settlement rails.
Which brings the stablecoin question into focus. That era taught every treasurer in America that a dollar can move in seconds, settle on a weekend, and be programmed. Banks watched the lesson and answered it inside the perimeter โ instant internal ledgers, API-driven treasury, insured balances. Erebor is not competing with stablecoins. It is selling the compliance wrapper around behavior that stablecoins proved out. The float stays inside the insured system. The yield accrues to a balance sheet, not a token holder.
Now the wrapper. RVI is a NYSE-listed fund holding private positions. That structure creates a mark. A listed fund's share price and its net asset value are two different numbers, and for private holdings the NAV is a model output, not a market print. Same problem I audited in 2020 when I forked Compound's source to understand its interest rate curve: the mechanism was legible because every input sat on-chain and every state transition could be replayed. A private mark is the opposite. You cannot fork it. You cannot replay it. You trust the valuation policy and the auditor.
That is not a scandal. It is an oracle problem wearing a suit. The entire tokenized-RWA narrative โ tokenized treasuries, tokenized private credit, tokenized fund interests โ hits the same wall. The token is trivially verifiable. The claim it represents is not. Code does not lie, but it does leave traces, and the trace here leads back to a committee setting fair value on an illiquid position. Trust is verified, never assumed. Inside a listed private-equity wrapper, almost nothing is verifiable from the outside.
In 2026 I audited the zero-knowledge circuits on a verifiable compute layer for an AI oracle. The lesson was not that proofs are magic. A proof tells you the computation ran correctly against inputs someone else supplied. The circuit is the easy part. Input attestation is the war. Banking settled that war with insurance and examiners. Crypto is still fighting it, one oracle committee at a time.
Which is precisely why the charter matters more than the token.
Here is the part my own side of the industry will not like.
The bull market narrative says crypto is absorbing traditional finance. Stablecoins eat deposits, tokenization eats settlement, DeFi eats the yield curve. Erebor points the other direction. A federally chartered, deposit-insured, OCC-supervised bank raised seven billion dollars to serve the most capital-intensive sectors in the economy โ without a token, without a DAO, and without a single on-chain governance vote.
Charters solve the one problem tokens structurally cannot: legal finality with recourse. If Erebor's treasury product misbehaves, there is an examiner, a statute, and a court. If a DeFi protocol misbehaves, there is a forum thread. In the red, we find the structural truth, and the structural truth of 2026 is that capital is flowing toward permission, not away from it. Permission is scarce. Scarcity is where the margin lives.
Governance is the art of managing disagreement, and the disagreement inside this deal is quiet. When I tested a quadratic voting mechanism against 500 simulated voters in 2024, the finding that mattered was not the lift in minority participation. It was that participation structures only hold when someone can enforce the outcome. Erebor has an enforcer. Most protocols have a snapshot.
The hierarchy is worth naming too. Robinhood's retail-facing fund holds preferred stock, senior to the common equity of the founders and the earlier backers. The vehicle marketed as access sits ahead of the people who built the thing. Legitimate structure. Still a structure.
Watch two disclosures, not the price. Deposit concentration in Erebor's next call report: if seven billion resolves to five accounts, the number is a warehouse and the franchise is unproven. And the next capital action โ preferred today tends to mean common tomorrow, and the terms of that raise will tell you whether the charter was the product or the platform.
The question I keep landing on: when the scarcest asset in the system is a permission, what exactly is being decentralized?
