The Number That Arrived Without a Tape

One hundred thirty-one dollars became one hundred forty-two dollars last week, and nothing was filed, settled, or cleared.

The revision came in a research note from MoffettNathanson, a firm whose reputation was built in telecom, cable, and media coverage rather than aerospace. The instrument it moved is the equity of a company that has never held a public offering. There was no opening print. No closing auction. No consolidated tape, no dark pool print, no exchange record that anyone outside a small circle of funds and family offices could inspect. There was a number inside one document, then a different number inside another document, and the difference between them is 8.4 percent.

Crypto has spent the past several weeks chopping sideways — majors pinned inside a range, funding rates drifting toward neutral, the sort of tape that produces essays about boredom. Meanwhile, in the private market, the graph goes up. It goes up most quarters. It goes up because the only people who publish a number for SpaceX are people who have a reason to publish a number, and nobody is required to publish one at all.

When the graph spikes, the soul remains quiet.

That quiet is what I want to write about. Not whether SpaceX is worth $142 a share — I genuinely cannot tell you, and neither can you, and that is not a rhetorical flourish. It is the finding. What I can tell you is that a target price is a genre convention of public equity, and somebody transplanted it into an asset class that has no closing price, no float, no order book depth, no settlement finality, and no obligation to disclose the inputs. The number is real in the sense that a number is real. It is legible in the sense that a headline is legible. Whether it is true is a different question, and it is a question the market structure actively prevents anyone from answering.

What The Note Actually Said, And What It Refused To Say

Start with the inventory of facts, because the inventory is short.

MoffettNathanson, an independent equity research firm, raised its target price on SpaceX from $131 to $142. SpaceX is not listed on any exchange. Its shares change hands through periodic company-run tender offers and through private secondary venues that broker transactions between existing holders and approved buyers. Because the company has no public float, it carries no SEC reporting obligation — no 10-K, no 10-Q, no 8-K, no proxy statement, no audited quarterly segment disclosure. Everything the public knows about its financials arrives voluntarily, through reporting, leaks, tender-offer circulars that circulate informally, and the aggregate behavior of a thin secondary market.

When I ran this through my own diligence framework — the seven-axis pass I use on any protocol or platform, covering regulatory posture, technical architecture, business model, competitive position, financial risk, macro policy exposure, and user scenarios — only three axes had enough material to score at all. Business model, competition, and policy. The other four returned insufficient data. Not bad data. Insufficient data. That distinction matters enormously, because insufficient data does not mean the risk is absent. It means the risk is unpriced by anyone who cannot afford to build their own model from scratch.

The original note evidently carries more. Nobody publishes a target price without a spreadsheet behind it, and the spreadsheets behind sell-side notes are usually competent. But the inputs — subscriber counts, ARPU by region, satellite depreciation schedules, launch cadence assumptions, Starship capital commitments, terminal discount rates — live behind a wall that only the firm's clients and the company itself can see. What reaches the public is the output: a two-digit number with a dollar sign.

I have spent enough of my career staring at half-published systems to recognize the pattern. The most consequential number in the room is almost always the one nobody is required to explain.

This Is Not A Rocket Valuation. It Is A Broadband Valuation Wearing A Rocket Suit.

Here is the observation that reframes the whole event, and I have not seen it made cleanly.

Craig Moffett built a career at Bernstein and then at his own firm by understanding one industry extremely well: connectivity. Telecom carriers, cable operators, satellite operators, media distributors. His analytical reflexes are the reflexes of a person who values subscriber bases. He thinks in ARPU, churn, net additions, capex per subscriber, spectrum position, and the terminal value of a recurring revenue stream with a high fixed-cost base.

That is not a rocket mindset. It is an ISP mindset.

So the $142 is best read as a broadband subscription valuation attached to a company that happens to own a launch business. The $142 target is an ISP in a rocket suit. The launch division is not the product in this model; it is the capex line. It is the reason the ISP can put capacity in orbit at a cost per kilogram nobody else can match. The revenue that justifies a target price like that comes from people paying a monthly fee for a dish and a beam.

Which means the four numbers that determine whether the thesis is correct are these: how many Starlink subscribers exist as of a specific date, what those subscribers pay on average across regions with wildly different pricing power, what the per-satellite capital cost and replacement schedule look like once the constellation reaches steady-state replenishment, and how much of the free cash flow gets consumed by Starship development rather than returned or reinvested into the network.

None of those four numbers is public. Some approximation of the first has appeared in press reporting at various points, always with a date stamp and always moving. The other three are essentially invisible. An analyst with access can model them. You cannot. And the gap between those two positions — analyst access versus public silence — is not a minor asymmetry. It is the product.

The Arithmetic That Does Not Reconcile

Let me do something slightly uncomfortable and show my work, including the part where the work fails.

A per-share target price is meaningless without a share count. SpaceX's capital structure includes multiple share classes, preferred instruments, and instruments that convert on various triggers. On an as-converted basis, reasonable guesses for the total land somewhere in the neighborhood of 1.2 to 1.6 billion shares, and I want to be explicit that I have not been able to verify that figure from any primary source. It is a guess built from secondary reporting and it could be wrong by a wide margin.

Run the multiplication anyway. At $142, a 1.2 billion share count implies roughly $170 billion of equity value. At 1.6 billion, roughly $227 billion. That is already a band wide enough to fit a mid-sized public company inside. Now set it against the valuations that have circulated around tender offers in recent years, which have been reported well north of $300 billion and in some accounts approaching $400 billion.

Those numbers do not obviously reconcile.

There are at least four honest explanations, and I cannot rank them because I lack the inputs. The target price might apply to a specific share class with different economics than the class used in the tender. The share count might be substantially higher than my estimate, which would push the implied value up. My recollection of the tender valuations might be imprecise or reflect pre-money, post-money, or headline figures that were never defined carefully. Or the target price and the tender marks simply reflect different methodologies operating on different time horizons, one of which is not a mark at all.

Two competent observers can sit two times apart on the value of the same company, and neither can check the other's work. That is not a scandal. It is the ordinary condition of a market with no price.

I am not accusing anyone of error. I am demonstrating that I cannot tell, from the outside, whether there is one. And if I — someone who does this professionally, who has audited prototype contracts line by line and argued with developers about reward curves at two in the morning — cannot resolve it in an afternoon, then the retail buyer reading a headline about a raised target has no chance at all.

The Oracle With One Reporter

Private markets are, structurally, a blockchain with a single validating node.

I mean that precisely. On-chain, the entire discipline of oracle design exists because a price that feeds into settlement is a point of attack. The mature designs share a common shape: multiple independent reporters, an aggregation function, a dispute window during which a wrong value can be challenged, an economic bond that gets slashed if the challenge succeeds, and a fallback path when the primary feed fails. The cost of lying is explicit and denominated. That is what makes the number usable.

Strip all of that away and you get a system where one reporter publishes a value, no mechanism exists to challenge it, no bond is at risk, and no settlement depends on it being right — until a transaction happens, at which point the transaction price becomes the new truth and the old report becomes a footnote.

The $142 target is an oracle with one reporter and no dispute window. It is not malicious. It is simply not falsifiable, which is a different problem and a harder one.

The downstream mechanism deserves attention, because this is where the damage tends to accumulate. Funds that hold SpaceX on their books need a quarterly net asset value. They need a defensible number for their limited partners. The tender offer provides one observable anchor, and the research note provides another, and both are more comfortable than a written-down model. So everyone marks toward the most recent observable anchor. Nobody writes down voluntarily. The anchors ratchet upward and never reset.

I watched that machine run in 2022 from close enough to smell it. Terra and Luna were not killed by a code bug in the narrow sense. They were killed by a system where the price that validated the system was produced by the system, and the participants had spent a year treating that price as independent because it had never yet been wrong. The firms that fell hardest in the contagion — the lenders, the funds, the platforms that froze withdrawals — had all marked their books to numbers they could not have sold into. The number was not a price. It was a belief with a chart attached.

When the graph spikes, the soul remains quiet.

The quiet, in 2022, was the sound of people realizing the tape had never been independent of them.

What Actually Drives The Value, And Why You Will Not See It

Let me be fair to the company, because the reflexive anti-private-market take is lazy and I am not writing it.

SpaceX's business is genuinely extraordinary in ways that are documented and verifiable. Launch cadence that no competitor approaches. Reusable first stages that land routinely, which was science fiction when I was first reading about the industry. A vertically integrated model that builds the rocket, the satellite, and the customer terminal in-house, which eliminates three layers of vendor margin. Government contracts of the kind that provide multi-year revenue visibility. A satellite constellation that is, as of this writing, the only one operating at meaningful commercial scale.

Now the other side. The value of that business depends on things that are knowable in principle by management and disclosed to nobody. Subscriber counts at a specific date. Revenue per user segmented by country, because Starlink pricing in a wealthy market and pricing in a developing market are different products with different margins. Satellite design life versus actual observed life, because the gap between those two figures determines whether the constellation is a capital asset or a consumable. Ground infrastructure spend. Spectrum acquisition costs. The cost per kilogram that Starship is supposed to reach, against the cost per kilogram it currently achieves, against the cost per kilogram competitors achieve.

Every one of those is an operational metric that a competent management team tracks in a dashboard. None of them is a legal obligation.

That is the design of the asset class, not a failure of this particular company. The disclosure regime we take for granted in public markets exists because retail money enters through public markets and retirement accounts, and a century of securities law concluded that money deserves protection. Where retail money is legally excluded, the duty to disclose evaporates. Private companies are permitted to be opaque, and most of them are.

The interesting question arrives when retail money enters anyway — through a secondary platform, through a special purpose vehicle, or through a token. Then you have retail exposure without retail protection, and the structure that produced the protection never engaged.

The Wrapper Economy

As widely reported during 2025, several retail-facing brokerages began offering tokens that track private companies, SpaceX and OpenAI among them, issued through special purpose vehicles and distributed to non-US users under local rules. At least one of the underlying companies pushed back publicly on the implication that it had endorsed the arrangement.

I want to walk through the mechanics, because the mechanics are the story.

An SPV is formed. The SPV acquires shares in the private company, typically through a secondary transaction brokered by an intermediary with access. The SPV signs a limited partnership agreement with its investors. A platform then issues a token that represents a claim on the investor's interest in the SPV. The token trades continuously on a public chain.

Count the layers. Platform. SPV manager. Custodian holding the underlying shares. The underlying company, which has no relationship with any of the first three and no obligation to honor any claim arriving through them.

Now count the things the token does not have. It does not have a redemption channel at net asset value. It does not have a published attestation of the underlying share position at a verifiable on-chain address. It does not have a transfer restriction regime that maps cleanly to the securities law underneath it. It trades on weekends and at 3 a.m., which is a property, not a virtue, because the thing it supposedly represents trades never.

A wrapper is not a rail. This is the distinction that years of crypto product design keep blurring. Wrapped Bitcoin works, to the extent it works, because there is a mint and a redeem path, an identifiable custodian, published attestations, and a market where the wrap and the underlying can be arbitraged against each other. Remove redemption and the token stops being a claim on anything. It becomes a prediction market on a valuation, with the added feature that the resolution date is unspecified and may never arrive.

And the securities law does not dissolve. Rule 144 holding periods, Rule 144A qualified institutional buyer requirements, Regulation D accredited investor verification, Regulation S offshore distribution — a token changes who gets told about the restriction. It does not change the restriction.

Verification Is A Budget Line, Not A Virtue

There is a reason nobody checks.

I spent a meaningful stretch of my career inside zero-knowledge proving cost models, and the lesson from that world transfers directly. Proving is expensive. Verification is cheap in theory and expensive in practice, because verification requires a verifier with the right inputs, the right tooling, and the right incentive. The entire research arc of the ZK industry — recursion, aggregation, proof compression, hardware acceleration — exists because the naive assumption that verification is free turned out to be false at scale. Operators burned real money discovering that.

Private company diligence has the same shape. To verify a SpaceX mark, you need a data room, a management conversation, access to the capitalization table, and enough context to know which share class you are looking at. That is a five-figure exercise at minimum, before you get to the part where access is gated by relationships. Nobody purchasing a hundred dollars of a token gets any of it. They get a webpage, a chart, and a name.

The cost of verification is the price of trust, and when the cost is high, trust is what people buy instead.

I learned this the hard way in 2017, when I was building the vote-weighting machinery for a quadratic funding mechanism and manually auditing more than fifty prototype contracts to make sure the code enforced democratic intent rather than just paying lip service to it. The problem we kept hitting was not the math. It was sybil resistance. Making a fake identity expensive enough that gaming the mechanism cost more than the subsidy it captured was the entire engineering problem, and it was never fully solved — it was only made expensive enough to be uninteresting. Verification was a budget line, not a principle, and everyone in the room eventually admitted it.

The Layer Nobody Prices

Here is a piece of structure that most buyers of private company exposure never see.

Preferred shares carry liquidation preferences. A 1x preference means the holder gets their money back before common shareholders see anything. A 2x preference means twice their money. Participating preferred gets the preference and participates in the upside. Ratchets adjust conversion ratios downward if a later round prices lower. Anti-dilution provisions come in full-ratchet and weighted-average flavors with materially different consequences.

Now stack special purpose vehicles on top, each with its own fee load, its own vintage, its own share class, and its own carry arrangement.

Two people who both say they own SpaceX can be exposed to two economically different companies. One might hold common acquired in a secondary transaction at a premium to the last primary round. Another might hold preferred acquired at the primary, sitting senior in the stack with a preference that guarantees a return before the common sees a dollar. Same headline valuation. Radically different outcomes in any scenario other than a very large exit.

Crypto has built this structure twice already. DeFi lending tranches, and the pre-token points markets, where notes purchased at five or ten cents resolved at wildly different values depending on the eventual terms — and where the people holding the senior claims understood the stack while the people buying points on a discord server did not.

The headline price is the least informative number in the entire structure, and it is the only one that travels.

The Subsidy Parallel I Cannot Stop Seeing

In 2020, in the middle of DeFi Summer, I sat in a series of increasingly unpleasant meetings about liquidity mining rewards.

The mechanism is simple. A protocol pays token emissions to whoever deposits liquidity. Deposits rise. Total value locked rises. The chart goes up and to the right. Everyone in the room knows the chart is a subsidy, and everyone in the room also knows that if you stop paying, the deposits leave within days, because most of the capital that arrived was attracted by the yield and has no opinion about the protocol whatsoever.

I refused to sign off on a distribution that rewarded speculation over usage, and it cost me. Three months of negotiation with core developers, a boardroom where I was the only woman and my concerns were routinely described as soft, and investors who wanted the TVL number because the TVL number was what they needed for their own fundraising. I argued that a number produced by paying for it is not evidence of anything. I lost some of those arguments.

Now look at the tender offer.

The company arranges a transaction at a fixed price. That price becomes observable. The observation propagates into fund NAVs, into research notes, into headlines. New capital arrives, attracted by the observable price rather than by the underlying cash flows, because the observable price is the only input most of that capital can access. The price ratchets upward on the strength of its own observability.

A price you cannot sell into is a press release with a decimal point.

The distortion compounds the same way it does in DeFi. LPs who entered at the top of the incentive curve eventually exit at a discount, and the discount is socialized across everyone who stayed. In DeFi we call this impermanent loss and post it in a forum thread. In private markets we call it a down round and discover it three years later, when retirement money has already been restructured twice.

What I remember most from that summer is how easy it was to defend the subsidy. The spreadsheet was right there. The argument for patience required explaining a mechanism, and the argument for emissions required pointing at a line that went up.

Legibility Is Not Transparency

Here is where the ETF analogy earns its keep, and where I want to draw a line that I think the industry keeps blurring.

I spent part of 2025 helping translate cryptographic concepts into language that regulators could actually evaluate, ahead of the ETF approvals, sitting between engineers who did not understand policy timelines and lawyers who did not understand cryptographic guarantees. The experience forced a distinction on me that I had previously held loosely.

Bitcoin was never opaque. Every block, every output, every input, every coinbase reward since 2009 is verifiable by anyone with an internet connection and the willingness to run a node. Transparency was never the missing property. What was missing was legibility — the translation of a transparent system into instruments that institutions with rules are permitted to hold, price, custody, and report. A pension committee cannot mark to a twenty-four-hour venue with no closing print. It cannot custody a private key. It cannot explain a liquidation cascade to its board. The ETF did not add transparency. It added a translation layer.

For SpaceX, the honest ask is not disclosure of everything. The honest ask is a legible instrument. A registered security with standardized periodic reporting, a defined share class, a disclosed capitalization structure, and a market where the price is produced by buyers and sellers rather than announced by a research note.

That is a real ask, and it has a real cost, and the cost is the subject of the next section.

The Uncomfortable Case For Opacity

I am a decentralization advocate. I have spent most of my working life arguing that verifiability is a precondition for legitimate systems. So let me make the argument I find genuinely difficult.

Starship took more than a decade. Reusable boosters took years of destroyed hardware and public failure. Satellite constellations require capital deployed years before the first dollar of revenue. This is the shape of the work: long engineering clocks, high capital intensity, uncertain timelines, and repeated visible failure.

Now put that on a public market with a daily price.

A public company pursuing this program gets punished at every cratered test flight. The price falls. The board faces pressure. The program gets stretched, descoped, or cancelled. The capital that has a ten-year horizon gets diluted by the capital that has a ten-day one, and the ten-day capital votes with the share price.

Opacity is a subsidy for patience. The illiquidity is not an accident of the structure. It is the mechanism that stops capital from fleeing at the first bad quarter. Patient capital gets compensated for its patience with an information advantage and an exit advantage, because that is the only way to attract capital that would otherwise go somewhere liquid.

If you tokenize the equity and give it a continuous price and a light exit, you import reflexive volatility into a ten-year engineering clock. The flight fails, the token drops, the program gets funding pressure, the timeline stretches, the token drops further. That is a doom loop that private structure currently prevents. As someone who believes in public verifiability, I do not have a clean answer for it.

The harder version is worse. Information asymmetry is not only a bug. It is part of the compensation package for the capital that funds the risk before anyone else will.

And yet.

Look at who actually collects the subsidy. Not the patient capital alone. The subsidy flows to the people closest to the information and closest to the exit — the ones who know what the tender price was before it was published, who know the subscriber count before it appeared in a journalist's inbox, who can sell into a tender while everyone else reads about it afterward. The patient capital gets compensated. So does the adjacency.

The retail buyer of a headline gets a chart and a name. Nothing else.

When the graph spikes, the soul remains quiet.

That quiet is the sound of a subsidy being collected by people who did not fund the risk.

What I Would Watch, Concretely

Signals, not predictions. This market is chopping sideways and the discipline that pays in chop is positioning against specific triggers rather than against a direction.

Starship orbital cadence and payload deployment. Not test flights — payload deployment, repeatably. The entire cost-per-kilogram thesis for the next decade rests on it, and a research note cannot make it happen.

Amazon Kuiper reaching commercial service. This is the clock that matters, and it is running. A competing constellation with a parent company that can absorb years of losses changes the pricing power of the incumbent. Also worth watching: Vulcan and Ariane 6 cadence, because launch competition compresses margins on the division that currently funds everything else.

Starlink subscriber count and ARPU, disclosed with a date. If the target price is an ISP valuation, those are the only two numbers that matter, and every other input is downstream of them.

The spread between primary tender marks and secondary prints. Secondary transactions in private company stock frequently clear at a discount to the last headline valuation, and that discount is the most honest opinion the market produces. When the spread widens, the headline is under pressure regardless of what any note says.

Employee tender participation. This one is underrated and I have not seen it modeled. Employees hold the best per-dollar information advantage available, because they see the operations daily without paying for a data room. How many of them choose to sell into a tender — and what fraction of their vested holdings they sell — is a sentiment signal with better inputs than anything in the public record. When participation surges, the people closest to the work are taking chips off the table.

And redemption. Whether any tokenized wrapper ever publishes a redemption channel at net asset value with verifiable attestations of the underlying position. If redemption appears, the product changes category and becomes genuinely interesting. Until then it is a trackable claim with a price chart, and the chart is not the asset.

Regulatory treatment of tokenized private equity, particularly in jurisdictions that have built actual regimes for it rather than tolerance-based enforcement. A regime with disclosure requirements and a redemption obligation would be the single most important development in this entire story, and it is more likely to arrive from Brussels than from Washington.

What This Actually Means

The number is $142 now. It was $131 before. In six months it will be a different number, and someone will write about it, and the process will repeat with the same opacity and the same asymmetry and the same inability of any outside observer to check the work.

I would rather leave you with a question than a conclusion, because I think the conclusion is genuinely unavailable.

If a valuation cannot be disputed by anyone outside a small circle, if the price is produced by the same structure that benefits from the price, and if the only channel through which ordinary people can access it removes their legal protections without removing their exposure — what exactly is being sold, and to whom, and on whose behalf?

When the graph spikes, the soul remains quiet. It is worth asking why the quiet never appears in the research note.

This is analysis, not investment advice. Positions in private companies and their tokenized representations carry risks that cannot be evaluated from public information, which is rather the point of the preceding several thousand words.