The Financial Times dropped a number on September 13 that should have moved more than one market: Vy Capital, a venture firm most people cannot name, holds roughly $40 billion of SpaceX stock — about 3.4% of the company. By Bloomberg's count, that makes it the fifth-largest holder on record, ahead of Sequoia Capital and Andreessen Horowitz on disclosed positions.
Four people decide what to do with it.
That is the detail buried under the headline. Vy's core investment team is four individuals. The firm has a few dozen employees in total. It stopped accepting outside capital last year. And it sits on top of an asset base that grew from $27 billion at the end of last year to $50 billion in June.
I have spent nineteen years watching markets price things they cannot see. Crypto taught me that the fastest way to find a defect is to ask who holds the keys and who holds the receipts. On that test, this is not a story about Elon Musk's loyal ally. It is a story about $50 billion in marks produced by an unaudited process.
Context: the firm that bought the worst headline
Vy Capital was founded in 2014 and lived below the media horizon for most of its existence. It first bought SpaceX in 2016, when the company was valued near $15 billion.
The timing matters. In September 2016, a Falcon 9 detonated on the pad at Cape Canaveral during a routine static-fire test, destroying the Amos-6 payload and grounding the fleet. That was the point of maximum perceived risk in the SpaceX narrative — the moment when a rational allocator re-underwrites or walks. Within weeks, John Hering, the firm's head, committed more than $100 million. This was not momentum capital. It was conviction capital deployed into a headline that read like a verdict.
Hering's relationship with Musk is operational, not merely financial. From 2019 he worked inside Starlink's early business development — hiring personnel, building financial models — and held a SpaceX employee badge. He sits on the board of The Boring Company, where Vy is the largest external investor and participated in the recent $3 billion round. Vy is also the largest outside holder in Neuralink, and in 2022 it committed $700 million to Musk's acquisition of Twitter.
The performance record is striking on paper: since 2014, a reported 41% internal rate of return, with $4.6 billion distributed back to investors. The firm now expects SpaceX to clear a $10 trillion valuation within five to seven years.
Set the disclosure against the size. A $50 billion asset base. A four-person decision core. A valuation thesis that depends on a single name compounding for another decade. There is no public ledger, no 13F equivalent, no daily NAV, no independent valuation committee visible from the outside. That is the structure. The rest of this is about what the structure implies.
A cap table is not a ledger
Public equities run on a disclosure regime: quarterly 13F filings with a 45-day lag, insider Forms 4 within two business days. Crypto rails settle in seconds and disclose in blocks. Private markets disclose when a reporter calls.
SpaceX shares carry transfer restrictions and rights of first refusal. The cap table is a private document maintained by a transfer agent and a general counsel. Every secondary trade needs company consent. There is no mempool. There is no block explorer. There is no state root to reconcile against. The ledger remembers what the market forgets — and where no ledger exists, the market simply remembers whatever the last round said.
I have audited token contracts where total supply was hardcoded and verifiable to the wei. I have never been able to verify a private fund's stated position in a single line item, because the only evidence is a mark produced by the party that holds it. In 2017 I tore apart the Parity multisig freeze by reading the state root discrepancy directly; the chain handed me ground truth in under four hours, which is why that breakdown beat the wire services. No equivalent exists for a SpaceX position. The closest thing to an audit here is a press leak.
SPV stacking is wrapped exposure without the reserves
Most allocators never touch a SpaceX share directly. They buy into a special purpose vehicle that holds shares, or a feeder fund that holds a stake in an SPV that holds shares. Sometimes there are three layers. Each layer charges management fees and carry. Each layer marks independently, on its own schedule, using its own valuation policy.
That structure should be instantly familiar to anyone who has used a bridge. Wrapped BTC on an L2 that bridges to an L1 that custodies with a multisig is not BTC. It is a claim on a claim on a claim. Every hop adds a trust assumption and a fee, and the end user's balance sheet carries the compounded risk. SPV stacking is the same machine with better lawyers. The difference is that a bridge at least publishes its reserve address.
There is a second-order effect that nobody prices. Layered SPVs create synthetic float. The same underlying share can support multiple economic claims across different vehicles — not by design, but through sloppy administration and mismatched valuation dates. When those vehicles mark on different schedules, the market gets contradictory prices for the same asset and no mechanism to arbitrage them. In a bull market nobody notices. Marks ratchet upward, subscriptions arrive, and the spread between vehicles looks like skill instead of noise.
The 41% IRR has no counterfactual
Let me put on the audit hat, because the headline number deserves it.
An internal rate of return is a time-weighted figure. It is dominated by the vintage of capital and the size of the largest position. A fund that deployed into a single asset at a $15 billion valuation and now carries it near $1.75 trillion on reported marks will print a spectacular IRR regardless of what happened in the rest of the book. IRR does not tell you the hit rate. It tells you the timing of the biggest winner.
The more informative figure is the distribution line: $4.6 billion returned to investors against a $50 billion asset base. Read that again. Most of the value Vy claims exists as unrealized marks on a single, illiquid, transfer-restricted position. That is not an accusation — it is a description of what the number actually proves. Marks are opinions. Settlements are facts. A private fund's NAV is a formula applied to a valuation policy, and the valuation policy is written by the same four people who decided to buy.
Crypto holds an uncomfortable advantage here, and I say that as someone who spends most of his working hours finding flaws in crypto. I can open an RPC endpoint and read a DAO treasury to the block — Uniswap, Aave, Arbitrum — and know precisely what it holds, when it moved, and what it paid in gas. The transparency is total and the analysis is instantaneous. Private markets offer the inverse: total opacity, delayed analysis, and a press cycle as the settlement layer. Anyone who believes crypto's remaining problem is transparency has never attempted to diligence a private mark.
The employee badge is the real disclosure
Hering held a SpaceX employee badge while his firm accumulated a stake that now ranks fifth. He helped construct Starlink's early financial models and hiring pipeline. This is the single most important fact in the story, and it is almost always reported as color.
Treat it as governance instead. A shareholder who also operates inside the company has access to cash flow projections, launch cadence data, and constellation economics months or years before the next external round is priced. That is not a scandal. It is the mechanism by which the round gets priced at all. Early capital buys information rights, and information rights are the actual product being sold.
Crypto builders will recognize this instantly. It is the private round at a fraction of a cent and the public round at two dollars. It is the advisor allocation. It is the pre-mine with a vesting cliff. The meaningful difference is that in token markets the schedule is published, the unlock is on-chain, and the distribution is visible to everyone in real time. In private equity the identical asymmetry exists with a nondisclosure agreement stapled to it.
Power lies in the code, not the community. In a token, that sentence is a warning about admin keys. In a cap table, the code is a transfer restriction clause, and the community is a syndicate of limited partners who cannot sell.
Ten trillion dollars is arithmetic, not a forecast
Vy's stated expectation is a SpaceX valuation above $10 trillion within five to seven years. Run the math the way a junior analyst would: 3.4% of $10 trillion is $340 billion, an 8.5x on a $40 billion mark. Clean, compelling, and wrong.
Three corrections. First, dilution. SpaceX raises on a cadence. Employee option pools expand. Any future capital-intensive program — Mars logistics, Starship cadence, direct-to-cell spectrum — gets funded with newly issued shares. A 3.4% position today is not a 3.4% position in 2032. This is the fully diluted valuation error that crypto spent an entire cycle learning the hard way. FDV is a marketing number. Market cap is a settlement number.
Second, liquidation preference. Later rounds typically carry preferences that get paid before common equity. In a downside scenario, common — which is what most employee exposure and most early-vehicle exposure actually is — absorbs losses first. Headline valuation and recovery value are different variables, and investor letters rarely separate them.
Third, liquidity. To convert a $40 billion mark into dollars, someone has to buy it. The pool of buyers who can write a nine-figure check for restricted SpaceX stock is small, and a bull market does not expand it — it flatters it. The $10 trillion target is not a return assumption. It is a liquidity event assumption wearing a return assumption's clothes.
The Musk stack is a single sequencer
Here is where my DeFi bias becomes analytically useful.
Starlink, The Boring Company, Neuralink, xAI, X. Distinct entities, distinct shareholder bases, one operator, one overlapping capital pool, and a small ring of people who sit across several of them simultaneously. Vy is the largest external investor in two of these companies, a board-level presence in a third, and a $700 million participant in a fourth.
The industry spent two years telling itself sequencers would decentralize. They did not. In nearly every production rollup, block production is a single operator with a privileged key, and the word decentralized describes a roadmap slide rather than a running system. That concentration works fine until it does not, and when it fails, it fails without a fraud proof.
The Musk stack has the same topology at the entity level. Every new company adds surface area rather than independence. Capital, attention, and key personnel are shared. A shock to one — a launch failure, a regulatory action, a spectrum dispute, a financing window that closes — propagates to all of them, and to every vehicle that holds them at a mark derived from a comparable.
That is correlation risk sold as diversification. Vy's book is not a portfolio. It is one thesis with several tickers.
Crypto's answer is worse than the problem
The obvious response is tokenization. If private markets are opaque, move the cap table onto a chain and let the market price it continuously.
The rails exist now. Tokenized equity products have launched in Europe. Pre-IPO perpetual markets have appeared on-chain, offering leveraged, cash-settled exposure to names that will never be delivered against. Every one of them will eventually list something that rhymes with SpaceX, because demand is enormous and the supply of genuine shares is zero.
Understand what that instrument actually is. It is not a claim on a share. It is a bet on where a private mark will be reported next quarter, settled in stablecoins, with no delivery obligation and no hedge. The issuer cannot acquire the underlying — transfer restrictions, rights of first refusal, and securities law see to that. So the perpetual trades against the rumor of a valuation rather than the valuation itself.
That is not transparency. That is leverage on opacity. Crypto did not solve the private market's verification problem. It financialized the problem, added twenty times leverage, and filed it under innovation.
The physical layer nobody is auditing
One more item, because it will matter in the next cycle.
Starlink's direct-to-cell service turns satellite bandwidth into a consumer product rather than a hobbyist experiment. Blockstream proved years ago that you can broadcast the Bitcoin blockchain from orbit; Starlink makes that economically trivial at scale. For regions without reliable terrestrial connectivity, a constellation is the only viable settlement layer for on-chain transactions.
But Starlink is a permissioned network with centralized terms of service, a single operator, and jurisdictional exposure in every market it serves. You cannot run a validator on a network you do not control. The physical layer of the next cycle is licensed, centrally governed, and priced by an entity whose shareholders include a four-person investment team. Nobody is writing fraud proofs for the constellation. Somebody should be.
The contrarian angle: opacity is the position, not the risk
The consensus read on this story is loyalty. Hering backed Musk after a rocket detonated on the pad, kept backing him through Twitter, and is now being rewarded. Warm, tidy, and analytically useless.
Read it instead as a governance defect with a bull-market alarm attached.
In a bull market, valuation is not a judgment — it is an ambient condition. Marks rise because comparable companies rose, not because anyone re-underwrote the cash flows. A firm with four decision-makers and no visible independent valuation committee can carry a $40 billion position through a full cycle without ever being forced to defend the number. The defect stays invisible while the direction is up. It becomes visible in the drawdown, when LPs ask who set the mark and the answer is: the same four people, in the same room, with the same spreadsheet.
The second blind spot is the exit, and the coverage never touches it. What is the path to distribution? An IPO, a tender offer, a secondary sale, or a merger. SpaceX has shown no urgency toward a listing, and Musk has a documented preference for private markets and retained control. Tender offers have occurred and are typically structured to favor the issuer. Secondary sales require consent, and consent is granted at the company's discretion and at a discount.
So the $10 trillion figure is doing something specific inside that investor letter. It converts a liquidity assumption into a performance claim. That is what investor letters are for. But anyone reading it as a forecast should notice the forecast contains no mechanism.
The third blind spot is the one crypto people will hate. Public blockchains did not fix any of this. Tokenized equities cannot touch restricted private stock. Pre-IPO perpetuals are synthetics with no settlement. On-chain treasuries are radically transparent, but almost nobody allocates to them at institutional scale because the vehicles are small and the wrappers are unfamiliar.
The private market's opacity is not a bug waiting for a chain. It is a feature that concentrates returns in the hands of the few participants who can tolerate illiquidity for a decade. Four people controlling $50 billion with no obligation to explain themselves is the most efficient structure ever devised for that purpose. Opacity is not the risk in this trade. Opacity is the trade.
Takeaway
Watch three things from here. First, the secondary bid for restricted SpaceX stock — that spread is the only honest price in the market, and it will tell you what the mark is worth before any auditor does. Second, whether a listing actually prices, and at what discount to the last private round, because a discount would reprice every vehicle that marked to it. Third, whether the tokenized-equity rails ever gain access to a name like this, or whether they remain permanently relegated to synthetics built on somebody else's cap table.
The $40 billion is real. The question that has never been answered is who verified it — and in a bull market, nobody has to.