A Decline Without Discipline

Two months after the listing, two things happened at SpaceX at once. The stock fell below the IPO price in late July, and the low sat roughly 20% under it. And in the second-quarter results released within that window, quarterly capital expenditure rose from $10.1 billion to $18.4 billion — 1.8 times the prior quarter. On the company's own reported figures.IPO price — the price at which a company first sells shares at listing. Here it was $135, and the listing raised roughly $75 billion. Capital expenditure — money actually spent on plant and infrastructure, kept distinct from the decision about where that money goes (capital allocation). SPCX is the ticker under which the stock trades on the Nasdaq.

A price was unmistakably paid, and the plan for spending did not move a single notch. In this piece, discipline means the force that actually makes a company change how it plans to spend. Think of a household budget. A shrinking balance does not by itself shrink next month's spending. Spending falls only when someone cancels the card, or the household rewrites its own plan. A price being paid and a plan being changed are two different events. So the bare fact that the price moved is not, in this piece, discipline.

Put it behind the wheel of a car. The needle on the dashboard can swing hard, but if the wheel does not turn the car keeps going where it was going. In this piece the dashboard is the market price and the wheel is capital allocation — the decision about where the company's own cash and its borrowed money go.

The needle swung, and the wheel stayed where it was. So who is setting conditions on this company now? We take up the parties that paid a price, one at a time. Start by re-measuring the decline. In late June we offered that decline as evidence of discipline.

A Fall the Size of the Climb

Shareholders did pay. The stock peaked at $225.64 on 16 June. Every drawdown measured in this section runs from that peak, not from the IPO price. Within three days $600 billion was erased from market value; the 22 June close was $154.60, a cumulative 23% across the three sessions.The same three days reach us measured two ways. Reporting as of 18 June tallies market value falling from roughly $3 trillion to $2.37 trillion, roughly 21%; the close-to-close run through 22 June is 23%. The measurement dates and the bases both differ, so the two are recorded side by side.

But that $600 billion is not measured from the IPO price. It is the giveback from the 16 June peak. Market value at that peak stood 69% above the value at listing. The number we read as the size of the punishment was the size of a round trip.

Nor were those three days entirely a verdict on the fundamentals. The catalysts cited alongside the drop are the Cursor acquisition announcement, the bond plan, and the opening of options trading in the stock. The bonds were still a plan at that point; the size was fixed at $25 billion the following week. Options trading opened on 16 June, the day of the peak.Options trading — a separate market in the right to buy or sell a stock at a set price. For this stock it opened on 16 June, after the listing. Which is one more reason a price being paid cannot, on its own, be read as "the market punished him."

The phase that actually took the stock below the IPO price arrived quietly in late July. The low sat roughly 20% under it. Sources split on the exact low, so this piece records only the magnitude.One source puts the low at $107.01 intraday on 28 July; another lists the 12-month low as $104.83. The two differ on whether the figure is intraday or closing, so both are recorded and neither is used.

The strongest objection is the early-August surge. The stock jumped 15.8% in a day, the argument runs, so "a price was paid" is a snapshot of one phase. Start with the date. That surge falls between the 6 August close and the 7 August close, not on 9 August, and reporting alongside it has the stock back near $115 on 8 August. We could not obtain a precise close for 9 August.

And the reason reported for the surge is not a decision by the company. It is relief that the first lock-up expiry passed without a wave of selling.Lock-up — an undertaking that bars those who held shares before the listing from selling them for a set period. So this piece builds no claim on August prices. The stage runs to the 12-month high and low, and the three days in June.

The needle swung hard twice in two months. Now look at the wheel.

The Wheel Did Not Turn

Whether the wheel turned is read off four items: quarterly capital expenditure, the survival of the acquisition in progress, launch cadence, and governance. Bend any one of the four and discipline was present. The four do not carry equal weight. The first two are the actual test. Launch cadence had no reason to fall in the first place, so its holding steady proves nothing, and governance is not the kind of thing that moves in two months. Read the back two as background.

An objection stands up immediately: that we narrowed the definition in our own favour. Count only a change in capital allocation as discipline and no market reaction can ever qualify — the conclusion is already inside the definition. It is a fair charge, which is why the definition sits at the top of this piece rather than hidden at the back.

And what is being marked is not the wording of our June sentence but what that sentence promised. In late June we wrote this: "The market was not applauding 165 launches. It was pricing how fast the money staked on loss-making businesses burns." The stock was collapsing ten days after listing. The sentence was plausible, and half of it still holds. The sentence in which we used the word discipline is this one: "What disciplines him right now, immediately and financially, is not a person but a single impersonal public market."

Had it meant only that a price had been set, "discipline" would be a restatement of "the market priced it" and would claim nothing. What we promised alongside the word was a testable prediction — that the rate at which the money burns actually changes. That prediction is what we are marking. The narrowing carries a cost. The moment the capital-expenditure plan bends, this piece is wrong.

Cut the scope in advance too. Observing that none of the four bent does not separate "nobody disciplined the company" from "discipline was present and the company held the right answer." What we measure is whether the decision changed, not whether the decision was correct. And if a plan was quietly shelved without being announced, this frame cannot see it. The table sets a credit-funding row alongside the four, as background.

The heaviest cell in the table is the right-hand end of the capex row. It is primary evidence landing inside the window, because it was said after the drop. At the 4 August results release the company's finance chief put it this way: "the next two quarters are very similar to the current quarter from a CapEx level perspective." That is not a report of spending already made. It is guidance on the plan ahead. Not a cut — a hold.

The plan does not live only on paper. Of the $18.4 billion spent in the quarter, $15.8 billion went to AI compute — chips and data centres. The remaining $2.6 billion is everything else. On the data-centre side, a permit for a third building and a capacity contract with an outside company are confirmed. That capacity contract, though, dates to 6 May and therefore precedes the drop. It must not be read as a fresh commitment struck afterwards. It goes only as far as evidence that a plan already laid down is still standing.

We went looking for evidence in the other direction on purpose. Across the same results release, buybacks, withdrawal of the acquisition, and any reduction in capital expenditure went unmentioned — not once.Buyback — a company purchasing its own shares to return money to shareholders. It is the first card played when a company answers shareholder pressure, which is why we searched for it deliberately.

The strongest objection is the results themselves: the company has already turned toward profitability, and was that not what the market asked for?

Granted. The loss did shrink. But the axis is a different one. Discipline is read off "did the spending plan bend," not "did earnings rise." In that same quarter the spending plan went to 1.8 times. The driver confirmed behind the smaller loss is not restraint on spending but a surge in revenue. Even the AI segment was reported to have turned positive only on the company's adjusted profit measure.Adjusted profit — a supplementary measure the company uses in its results, taken before several costs such as depreciation are subtracted. It is a different figure from profit or loss under accounting standards. For the same quarter the company reported adjusted profit of $3.5 billion against a net loss of $541 million under accounting standards — the roughly $4 billion between the two is attributable, broadly, to items such as depreciation and interest.

In its July ratings opinion, the credit rating agency S&P expects this company's free cash flow to remain negative through the end of the 2020s.Free cash flow — the cash left after subtracting what was spent on plant and investment from the cash brought in. When it is negative, the company has nothing left over and must keep pulling money in from outside. Sources split on the end point, 2029 against 2030, so this piece records only "through the end of the 2020s." The two axes that outlook cites are AI investment and Starship. That the Starship promise is still short of demonstration was handled separately in Starship's Reuse: Reuse Is Already Proven — Starship Isn't.

A price was paid, and the plan is unchanged. Which leaves room for the disciplining party to be someone else. In late June we sat the bondholders in that empty seat.

Creditors Who Speak Only in Prices

In late June we wrote that the bondholders had attached conditions — that financial covenants came with the $25 billion bond, and that they tied down spending. We withdraw that sentence. It had no basis.

Lenders commonly write conditions down. Do not take debt above this level; do not push capital spending past that one. No such language appears in the issuer's own disclosure. It goes as far as saying the notes are unsecured and rank equally with the existing senior notes. The stated use of proceeds is repayment of a bridge loan and general corporate purposes. What the issuance summary describes is only the standard provision that a breach accelerates the remaining debt.

That does not license writing "there are no conditions." Three of the contract documents themselves we failed to open. We could not make the determination either way. Nor is it that no condition exists anywhere. Away from the public notes, the separately drawn bridge loan did carry a covenant to hold a debt multiple below a set line — as of the May listing documents. But the facility carrying that covenant is the bridge loan the June bond was earmarked to repay, so whether it is still live is something we could not determine.Debt multiple — how many times a year's earnings the debt amounts to. The covenant holds the consolidated multiple at or below 3.75 times, with up to 4.25 times permitted for four quarters where there is a qualifying acquisition. A single source, so it is attributed.

In practice a covenant like that is the standard device for tying down spending indirectly. Borrow to buy plant or to buy a company and the multiple rises, so the closer the line, the more new investment catches on it. Which makes "it never asks where the money goes" too strong. Put precisely: the line protects the ability to repay, and inside it, where the next dollar goes is not asked. And for as long as it is observed, the covenant demands nothing. Our sources hold no record of a breach, and we could not obtain how far the company currently sits from that line.

The direction we can confirm ran the other way. Money crowded into the June issue. The target was $20 billion; orders came to $89 billion, more than four times over, and the size was upsized and fixed at $25 billion. All three agencies assigned an investment grade.Investment grade — the band of ratings for borrowers expected to repay on time. Below it is called speculative grade. The issue priced in five tranches maturing from 2031 to 2056, with coupons of 5.35% to 6.65%.

Yet early the following month, the price at which those same bonds traded pointed the other way. By a secondary tally reassembled from trade data, in early July the average spread on this company's bonds ran 1.62 percentage points over government bonds — wider than the 0.92-point average for BBB, the band it sits in, and wider even than the 1.55-point average for speculative-grade BB.Spread — the extra yield paid over a government bond of the same maturity. The wider it is, the riskier the market judges the borrower. This figure is a secondary tally reassembled from trade data, so the argument rests on its direction and not on the precision of the number.

Read that number with one caution. The average mixes in maturities out to 2056, and the longer the maturity the wider the spread runs to begin with. Take the 2031 maturity alone out of the same source and it widened from 1.06 to 1.12 percentage points — far closer to the BBB average. Read the 1.62-point average whole as "the market's distrust" and you bill the company for width the maturity mix produced.

Through July some outlets built a "junk in all but name" frame on those secondary prices. Against the ratings table alone it is an overreach, but it is also a claim with a measured basis in traded prices. It is half the measurement, though. As we just saw, part of that average width belongs to the maturity mix, and the frame bills that part to the company as well.

A spread, though, is a price and not a piece of language. For it to become discipline, the dearer cost of funding has to bend the spending plan. We have already seen what the company said at the first results release after the spread widened. The next two quarters run at the same level. A price was paid in the bond market too, and the decision did not change there either.

Why the two markets look at different things comes down to a single ranking. Think of a mortgaged house. Sell it and the lender is paid first; what is left is the owner's. The price can move up or down and the amount the lender receives is unchanged — the whole of the movement lands on the owner. That is claim priority: when what a company holds is divided up, who is paid first and who takes what remains.

What the bond buyers looked at is whether it can repay through maturity. On the company's reported figures, cash and securities stand at $100 billion and the order backlog at $47.5 billion.Order backlog — work already under contract but not yet recognised as revenue. Trace where that cash came from and the picture is clear. The listing raised roughly $75 billion, and the June bond brought in $25 billion more. Part of the bond proceeds is earmarked to repay the bridge loan, so that much does not stay as cash. Even so, the two raises are the same order of magnitude as quarter-end cash. This is not headroom operations built up. It is money taken in from outside.

Nor is Starlink filling that gap. Second-quarter communications operating profit was $1.7 billion, up 79% on a year earlier, but that is less than a tenth of the same quarter's $18.4 billion of capital expenditure. What that coverage is becoming as an asset inside the telecom industry was covered in Borrowed Spectrum — Coverage Is Sliding From Asset to Common Component, and Exclusivity Is the Transition.

So "free cash flow negative through the end of the 2020s" and "investment-grade BBB" arriving side by side from the same agency is not a contradiction. As long as the company sits inside the line where it can repay, then within what we could confirm, bondholders hold no documentary means of stopping the spending directly. When uneasy, they move the price instead of the language.

Put the events in order and one thing catches. In the same week $600 billion was erased from the peak, investment grades arrived from all three agencies (18–19 June), and the $25 billion issue was fixed the following week (23 June). The secondary spread widened after that again, in early July. Across the two months in which the stock went below its IPO price, this company's funding never stopped. The power to set a price and the power to open and close the money tap do not sit on the same layer. While the price fell, access widened.

Recall the definition once. Paying a price and changing a decision are different things. The bond spread falls outside that line too. Which leaves the shareholders.

Free to Sell, Powerless to Change

Shareholders cannot set a condition on this company for three reasons, and all three are arithmetic. First, in this company the stock is currency. xAI and Cursor were both all-stock acquisitions — paying by newly printing the company's own shares instead of handing over cash. It is a shop settling its bills in its own gift vouchers. As long as it can print more vouchers it can keep buying even as their price falls, and what thins instead is the share belonging to whoever already held them. Of the two, xAI was a merger absorbed into the parent before the listing. A deal with the controlling shareholder standing on both sides carries a name: self-dealing. What that name contests is conflict of interest, which is not this piece's axis, but we will not pass it by without attaching the name.

This is not to say a falling share price constrains the company not at all. Where the consideration is fixed as a dollar amount, buying the same thing takes more shares, so it gets dearer. Whether the two deals fixed an amount or fixed an exchange ratio is something we could not confirm. Either way, the party bearing that cost and the party making the decision are not the same. The cost comes out of the remaining shareholders' share, and whether to issue is settled by whoever holds voting power in the mid-80% range. So the constraint does not translate into an approval step. The Cursor deal was not withdrawn, and is reported to have run toward a close targeted for late August.

Second, control is locked twice over. The first lock is the arithmetic of votes. The company has two classes of stock, one carrying one vote per share and the other ten. On the listing documents, Musk's voting power is in the mid-80% range.The figure printed in the listing documents is 85.1%, against an economic stake of roughly 42%. Reporting that recalculates it at 80–82% after the listing exists alongside it, and because the timing and the method differ, this piece records only "the mid-80% range."

The second lock is time. The lock-up does not release at once — it is split into nine stages, and the first tranche was 911.5 million shares. That is a little over a sixth of the 5.3 billion shares that come free together in early December. Core insiders stay locked until mid-2027. In the event, the first release on 6 August passed without a wave of selling.

What a lock-up binds is selling, not voting. The two locks are separate facts drawn from different documents, and this piece claims no mechanism joining them. They are set side by side because the outcome is the same — the only thing a shareholder can do is sell, and selling does not touch control. We looked for a case of an outside shareholder exerting influence through the board or a shareholder proposal, and none came up. We searched along two tracks, activist investors and shareholder proposals, and went as far as opening the letters issued by an institutional-investor body; nothing in them named this company. Even the credit rating agency Moody's wrote "key-man dependence" into its ratings opinion.The governance this piece measures is one route to changing capital allocation. As with the self-dealing named but not pursued above, the axis that contests conflict of interest is not treated here.

Third, money that does not judge came in. On 7 July the company joined the Nasdaq-100, and funds that buy along with the index flowed in. This is passive money — money that buys the index composition without forming a view on the company. Think of a standing bank order. The same amount leaves on the same day each month, and nobody looks at whether it is dear or cheap that day.

The objection comes that the inflow is an estimate, and a small share against a company this size. Correct. So we claim nothing from its magnitude. What this fact carries is not size. It is character. Shares bought along with an index still carry votes — no trading view is formed, but the voting rights are exercised. They are one-vote-per-share votes, though, and in front of the mid-80% range the arithmetic does not change. Beyond that, the money did not arrive because the company performed. It arrived because the exchange rewrote its rules. The opposite of discipline is not selling. It is not judging.

Record the money running the other way as well. Fifteen actively managed Korean funds built around a space theme took positions in the stock on their own judgement before the index announcement, and at some the weighting ran to the 26% range. Not all of the money that came in came without judgement. That money, too, attached no condition to the company.

The most honest objection is time. Two months from listing is short, and most of the shares are still locked. Once the lock-ups run off and the shareholder register changes, the argument goes, discipline will stand up then. Granted. Releases run through 20 August, September, and October into the large tranche in early December. So this piece's "no discipline" is an observation about June to August 2026, not a permanent declaration. The dates to look again are early December and mid-2027, and why those are the dates is set out at the end.

Not the bondholders and not the shareholders — which makes it look as though nobody has set a condition on this company at all. That is not so.

The One Condition Left Standing

Plenty of parties did set conditions on paper. The bridge loan carried the debt-multiple line. The regulator, the US Federal Aviation Administration, grounded Starship test flights for 47 days from late May over a mishap investigation. The Cursor deal has a regulatory approval process attached. All of them conditions people made and wrote down. One more cut to the scope — performance conditions attached to this company's government procurement contracts sit outside the axis this piece measures.

Yet not one of those three is actually holding the company now. A single question sorts them — is the condition still unmet, and does releasing it depend on the company clearing it itself? The grounding was lifted on 13 July: a condition met, and gone. The acquisition review is still open, but the party that clears it is the regulator rather than the company, and being a condition attached to a single transaction, it dies the moment that transaction resolves — it is reported to be running toward a close targeted for late August. The debt-multiple line is different again. It is not a threshold to be crossed but a quiet maintenance covenant for as long as it is observed, and our sources hold no record of a breach — meaning the company is not currently in violation. What is more, the facility carrying it is itself the bridge loan the June bond was earmarked to repay, so whether it is still live is something we could not determine. Once all three drop out for their own separate reasons, one thing remains that no resolution erases and that only the company can clear itself. The threshold of an accounting profit.

Index eligibility criteria are a rulebook set in advance for which companies go into an index. They work like admissions eligibility. However good the grades, you cannot file until the required years of enrolment are in — not because an admissions officer dislikes you, but because the criteria say so.

Faced with the same wave of listings, two indices moved in opposite directions. Nasdaq rewrote its rules with effect from 1 May 2026. A newly listed company inside the top 40 by market capitalisation is admitted after 15 trading days. The free-float requirement that had gated eligibility was removed as well. This company benefited from that rule at once.

The S&P 500 went the other way.The names overlap, but the credit rating agency S&P mentioned earlier and the index company that sets these criteria do different jobs. One assigns credit ratings; the other builds indices. Four criteria bear on this company.

• The most recent quarter must show an accounting profit
• The prior four quarters summed must be profitable as well
• The shares must have traded publicly for 12 months or more
• At least 10% of the shares must actually trade in the market

Accounting profit — net income positive when calculated under accounting rules. It is a different figure from the free cash flow seen earlier, so one can be positive while the other is negative.

This company does not yet meet the first three. Mid-2027 gets quoted often, but that is only when the third of them, the 12-month trading period, comes due. Summing four quarters to a profit takes several profitable quarters stacked up, so on the arithmetic the profitability threshold falls later than that.

Put one premise on record. All of this stands on the assumption that the company wants to clear that threshold. Yet within what we could confirm, the company has never publicly set a turn to profit as a goal. If it does not want to, the threshold has no force at all — and that too is a disconfirmation of the judgement set out below.

The threshold did not survive by default. In May 2026 the index provider reviewed a proposal to relax it — cutting the public trading period from 12 months to six, and waiving the profitability criterion for very large companies. On 4 June it rejected its own proposal. Looking at the same wave, one lowered the threshold and one held it.

A threshold and a recommendation are different things. Analyst price targets diverged sharply, but no amount of divergence stops the company doing anything.The two banks that led the listing came in at $205 and $300, and that gap alone is worth roughly $1 trillion of market capitalisation. It is a background observation that even the experts cannot fix a price, not a basis for this piece. Aviation draws the same distinction. A safety recommendation from an accident investigator can go unfollowed and the aircraft still sells; without type certification from the regulator it cannot be sold at all. Type certification is the approval to build and sell that model. What makes a norm bite is not the authority of whoever issued it, but whether it blocks market access.

Hold that yardstick against the eligibility criteria and the grain runs the wrong way. The criteria are not on the type-certification side. Kept out of the S&P 500, this stock still trades on the Nasdaq, and the funding actually happened in June. A threshold that does not block access at all is, within what we could confirm, the only condition on this company still standing unreleased.

What happens when a metric moves into a place with nobody to do the disciplining, medicine went through first.

In the 1990s New York State and Pennsylvania published mortality report cards for cardiac bypass surgery. The party those report cards were meant to hand power to was the patient. But a patient has thin means of knowing what happens in the operating room, and thin means of demanding anything of the hospital. The position the patient stands in is, here, the shareholder's. Once results were published by hospital and by surgeon, hospitals operated less on the sickest patients and moved them to other treatments the report card did not capture. Outcomes for the severely ill group worsened as a result.An economics paper on the effects of publishing coronary artery bypass graft report cards in New York State and Pennsylvania in the 1990s. Authors and journal are listed in the sources below.

One thing has to be said alongside it. The metric the report card measured did improve. What broke was outside the metric. The shape of this case is not "measurement failed" but "the metric improves and what sits outside it gets worse."

In a 2003 survey of interventional cardiologists in New York State, 79% are reported to have said their decisions to treat high-risk patients were influenced by the report card. Eight in ten. The figure reaches us via a secondary summary, so it is attributed.

Note where the analogy stops, too. A report card is a public instrument built with discipline as its intent. Index criteria are private design meant to protect the quality of an index product. Which makes this case the stranger one. A rule that never had any intention of disciplining this company is, within what we could confirm, the condition still standing unreleased.

Come back. What the criteria demand is an accounting profit. There is a soft spot. An accounting profit does not bend capital expenditure directly. Money spent on plant is not booked as cost all at once in that year; it falls out over several years as depreciation.Depreciation — the accounting treatment that spreads the cost of long-lived plant across the years it is used. Which is why a large capital investment does not cut that year's profit by its full amount. If anything, the incentive runs toward parking spending on the balance sheet as an asset. The adjusted profit seen earlier is precisely the measure taken before that depreciation charge. But that supplementary measure does not clear the criteria — what they look at is net income under accounting standards.

A force pushes the other way too. Spending parked as an asset does not vanish; it returns as depreciation in the quarters that follow. Keep spending $18.4 billion a quarter as now and that tail lengthens with it, leaving the $4.0 billion gap between adjusted profit of $3.5 billion and the accounting net loss of $541 million as the distance still to be covered. Capitalisation protects this quarter's profit exactly to the extent that it cuts the profit of the quarters after. So what gets managed is not the business but the metric that clears the threshold. The threshold itself is not a person, but the method of clearing it is in human hands — the accounting discretion that sets depreciation lives and the scope of capitalisation sits with a management installed by that mid-80% voting block.

And the threshold takes a long time. From listing to inclusion, Uber took four and a half years and Airbnb two years and nine months, and Uber's trigger was its first quarter of accounting profit.

Two Months in Which Only Prices Moved

Prices moving and the company moving are two different events. Shareholders did pay, and the bond market paid too, but nowhere in two months is there a trace of that price returning as a decision by the company. So there is one sentence we are correcting — in "the market disciplined him," what is wrong is not "the market" but "disciplined."

In a company whose capital structure is split into layers, judgement splits into layers too. Some layers only set a price and demand nothing. And the power to set a price and the power to open and close the money tap sit on different layers — across the two months in which the stock went below its IPO price, this company's access to capital widened.

Within what we could confirm — and setting aside the acquisition review, which waits on its own resolution — exactly one of the conditions written down against this company can be released only by the company clearing it: an accounting profit. The party that set it is not a person. It is a rulebook. More precisely, that rulebook was not built with this company in mind — the 4 June decision to leave the threshold where it stood, that one a person made. The seat of discipline is still empty.

Carry this frame to another company and four questions come with it. When the price moved, is there a party that can convert that price into a documented demand? If such a demand is attached, is it still unmet, and does releasing it rest in the company's own hands? What the remaining demand measures — is it the business, or a metric? And can the company still take in money from outside — while the money tap stays open, the plan holds even as the price falls.

Any product tracking the Nasdaq-100 has held this company since 7 July. It is not there because anyone judged the company. It is there because the exchange rewrote its rules. If the opposite of discipline is not judging, then that opposite flows through our accounts too.

We make one call, and we hang low confidence on it. The eligibility criteria will not fill that empty seat; they will summon metric management instead. As we just saw, the criteria cannot bend capital expenditure directly. So the order we expect runs like this — a quarter of accounting profit arrives before the cash position improves, and capital expenditure holds where it is. Write down the observable form as well: at the results release for the quarter in which the trailing four-quarter total turns positive, is a change to depreciation lives or to capitalisation treatment disclosed alongside it? Write the null hypothesis too — revenue growth at the current rate could deliver the profit first on its own. The signal separating the two cases is not the profit itself, but whether an accounting change arrives next to it.

There are two dates to come back to. Mid-2027 is when the 12-month trading requirement comes due and, at the same time, when core insiders' lock-ups release. It is not a projected inclusion date. The precedents are four and a half years and two years nine months. And what comes first on the calendar is the 5.3 billion shares released in early December. The chance for pressure from the shareholder side to stand up comes round once more then. The arithmetic seen earlier is unchanged, though — even given the chance, the only thing a shareholder can do is sell, and selling does not touch control.

If between the second half of 2026 and the first half of 2027 the quarterly capital-expenditure plan actually bends, or the all-stock acquisitions stop, or a buyback begins, our correction was premature. If the public bond contracts open up and a condition aimed at spending is in fact confirmed, this piece's account — that bondholders speak in prices rather than in language — collapses; as it stands we failed to open them, so the determination itself is unmade. And if the S&P 500 admits this company without an accounting profit, or the company puts out a profitability target independent of inclusion, this judgement is wrong.

For two months we looked at the dashboard and talked about discipline. The needle swung hard twice, and the wheel never moved.

In a company where answering for enormous promises has narrowed to a single person, what holds the last condition that only the company itself can release is not a person. It is a rulebook.