Forensic mode: Activated.
SpaceX just beat Wall Street by a margin that should embarrass the consensus. Second-quarter revenue came in at $7.814 billion, up 92% year over year, against expectations of roughly $6.8 billion. Adjusted EBITDA reached $3.538 billion โ 191% above the year-ago figure and far ahead of the ~$2 billion the street expected. Net loss narrowed to $541 million from $1.008 billion. Operating loss collapsed from $970 million to $143 million. On paper, this is a clean beat.
None of that is the lead.
The lead is a flat line buried inside the cash-management section of the new SEC filing: 18,712 Bitcoin. Cost basis: $661 million. Identical to Dec. 31. Identical to June 30. Zero net purchases. Zero net sales. This is a company that converted an $85.675 billion IPO โ the largest in market history โ into a liquidity stack approaching $100 billion. And during a six-month period in which the fair value of its Bitcoin position fell by roughly one-third, the treasury did not add a single satoshi nor sell a single coin.
Let me state the implications plainly. A large cohort of market participants priced SPCX as a Bitcoin proxy. The first quarterly report now shows that the post-IPO treasury's risk appetite stopped at money-market funds and government securities. It did not reach for Bitcoin at what the unfiltered price chart called the low. It did not trim Bitcoin to harvest the $539 million unrealized loss for tax purposes. It did nothing.
Data doesn't lie. It does require cleaning before it becomes evidence. That is the purpose of this report.
What follows is a five-part forensic reading of the filing: the earnings beat, the $100 billion treasury, the static Bitcoin ledger, the AI capital engine, and the $105 billion unlock. Each section is built on numbers from the report or derived directly from them. The conclusion will not comfort anyone who bought the proxy narrative.
Context: What This Filing Actually Is
Let me first establish what this document is. SpaceX's first quarterly SEC report since its May 2026 record IPO is no longer a launch-company financial statement. It sits at the intersection of four separate markets: crypto treasury policy, AI infrastructure spending, equity derivatives, and the largest lockup expiration in stock-market history. Each of those markets has a direct stake in three specific line items: the Bitcoin balance, the AI compute contract book, and the tradable float.
The IPO created a specific trade. Within days of the S-1 filing, crypto analysts presented a new framework for pricing Bitcoin exposure, X payments, and AI compute through a single ticker. The notable early finding was that a $2 trillion listing would eclipse Tesla in market value while holding less Bitcoin than the largest public treasury companies โ which challenged, even then, the idea that SpaceX would be a workable Bitcoin proxy. Crypto derivatives platforms did not wait for a resolution. SPCX futures were listed before the stock's Nasdaq debut, with pricing that implied a first-day pop. Traders wagered over $1 billion in crypto-based positions ahead of the listing.
That framing was always fragile. It treated a non-core treasury asset as the center of gravity for a company whose long-run value depends on launch margins, Starlink subscriptions, and now AI infrastructure. The quarterly report is the first instrument that lets us cleanly separate the proxy narrative from the actual ledger.
On methodology: I have been auditing this industry since before quarterly reports were part of the standard toolkit. In early 2021, OpenSea's volume was breaking records and the market called the NFT surge structural. I ran my own SQL queries across more than 450 collections and discovered that 30% of apparent volume was self-cleared โ wash trades between the same parties inflating dashboard totals. The market was less healthy than it looked. I published a "Real Volume" dashboard that became the reference standard for more than 500 analysts. That experience fixed my permanent filter: raw reported numbers require counterparty verification before they can be treated as evidence.
The same filter applies to a public-company filing. The headline revenue beat is real income-statement data, but it does not tell you where the company is going. The balance sheet does. The cash-flow statement does. The cost-basis line on a digital asset does. When I traced the UST de-peg for 72 hours in May 2022, the lesson was clinical: emotional positioning is a lagging indicator; the ledger is leading. Everyone on social media was declaring the peg fine while the Curve pool balances showed the exit. When I built the 11-issuer ETF inflow tracker in 2024, I found that institutional money moved on a schedule โ Tuesday at 10 a.m. Eastern, matching pension rebalancing โ with about 80% short-term prediction accuracy. Institutions follow rulebooks. Treasuries follow rulebooks. The rulebook is visible in the filing.
I measure. Then I write what the measurement shows. That is the whole method. Now let me apply it to each section of this report.
Core, Part 1: The Earnings Beat, Cleaned of Hype
The operating numbers deserve a standardized table, because they are the ground truth from which every other analysis flows:
| Metric | Q2 2026 | Q2 2025 | Change | | --- | --- | --- | --- | | Revenue | $7.814 billion | $4.07 billion (implied) | +92% | | Adjusted EBITDA | $3.538 billion | $1.216 billion (implied) | +191% | | Net loss | -$541 million | -$1.008 billion | -46% | | Operating loss | -$143 million | -$970 million | -85% | | Total capex | $18.369 billion | โ | โ | | AI capex | $15.828 billion | $749 million | +2,013% |
The beat is concentrated in revenue and EBITDA. Revenue growth of 92% year over year on a base of $7.8 billion is acceleration, not maturation. The EBITDA swing from roughly $1.2 billion to $3.5 billion shows the launch and satellite businesses converging on cash generation. Meanwhile the GAAP net loss, at $541 million, is the most honest number in the table. It is what remains after paying for the future.
Now clean the beat. Adjusted EBITDA is not the same as operating cash flow. The standard adjustments for a hypergrowth company exclude non-cash compensation, one-time items, and mark-to-market effects. The $3.538 billion EBITDA figure cannot be used as a direct cash proxy without checking the cash-flow statement for the same quarter. The operating loss of $143 million is the GAAP bridge between EBITDA and the net loss line, and it confirms that the legacy business has essentially reached operating breakeven. The remaining losses come from the investment budget and from the AI segment, which I will examine in Part 4.
The market sold the stock after the beat. That, in itself, is not a contradiction. Revenue beats are frequently sold when the market believes the operating beat is retroactive โ already priced โ and the forward budget contains the real information. The real information here is $15.8 billion of AI capex in one quarter and the largest lockup expiration in history eight days after earnings. The beat validated the past. The budget and the float schedule defined the future. An efficient market prices the future. That is why the short interest sits at 34% of the public float.
One additional number deserves emphasis: AI revenue of $2.561 billion represents roughly 33% of total revenue in the quarter. A company that generates one-third of its revenue from compute contracts in its first year of public disclosure is not a satellite company with an AI side project. It is a hybrid infrastructure company. The market's valuation models are still catching up to that classification.
Core, Part 2: The $100 Billion Liquidity Stack
The post-IPO balance sheet is a fortress. Cash, cash equivalents, and marketable securities closed the quarter near $100 billion. The breakdown from the SEC filing:
- Money-market funds: $65.625 billion at June 30, more than triple the $21.339 billion reported at year-end 2025.
- Government securities classified as cash equivalents: $4.011 billion.
- Government securities classified as marketable securities: $6.487 billion.
- Combined money-market and government-securities position: $76.123 billion.
The remainder of the near-$100 billion total sits in conventional cash and other permissible instruments. The structure matters. A treasury that parked $65.6 billion in money-market funds is not holding cash because it lacks ambition. It is harvesting yield at institutional scale while preserving the highest-grade liquidity. Money-market funds yield what the short end of the curve offers; government securities with matching durations add basis points and safety. This is the allocation of a company that wants maximum optionality, not maximum return.
Run the yield math, because the income statement only tells part of the story. At a conservative 4% annualized rate on the $76.123 billion of cash-like instruments, quarterly interest income lands around $760 million. Compare that to the reported net loss of $541 million. Even after the AI buildout, the liquidity stack alone covers the loss and leaves roughly $220 million of quarterly surplus. Consider the rate scenarios:
| Annualized Yield on $76.123B | Quarterly Interest Income | Net Loss (Q2) | Surplus | | --- | --- | --- | --- | | 3% | $571 million | $541 million | +$30 million | | 4% | $761 million | $541 million | +$220 million | | 5% | $952 million | $541 million | +$411 million |
The "loss" in the income statement is a chosen, financed loss. The company could fund a meaningful portion of its operations forever from the cash buffer. The treasury exists to buy time โ and in this case, to buy compute infrastructure without leverage dependency.
For crypto readers, this is the structural context for the Bitcoin question. A company with $100 billion in liquid assets that held its 18,712 BTC flat during a 33% drawdown did so while fully able to add hundreds of thousands of coins. The constraint was never capital. It was rulebook. The filing shows a treasury that standardizes on government-grade instruments and treats Bitcoin as a separate line item โ likely under a distinct authorization, with its own disposition rules.
Standardization is the reason we can run this analysis at all. SEC disclosure requires a consistent quarterly presentation of cash, cost basis, fair value, and segment reporting. That is the difference between an auditable system and an opaque one. When I audited the Terra collapse in 2022, the core problem was the absence of standardized, comparable data across Curve pools and de-pegging strategies. When I built the L2 Efficiency Index in 2023, the differentiator between Arbitrum and Optimism was not marketing โ it was that Optimism provided a more standardized smart-contract compatibility surface, which shifted 15% of developer activity toward better documentation. Standardization creates value because it allows measurable comparison. SpaceX's filing is a model of that discipline. The line "18,712 BTC, cost basis $661 million" is comparable across every future quarter. That is the foundation of trust.
Core, Part 3: The Bitcoin Ledger โ Static While the Pool Tripled
Now to the line that interests the crypto market most.
SpaceX reports 18,712 BTC with a cost basis of $661 million. That works out to approximately $35,300 per coin at acquisition. At June 30, the fair value of the position stood at $1.098 billion, down from $1.637 billion at Dec. 31. The difference is a $539 million unrealized loss.
Derive the implied marks. The year-start fair value implies roughly $87,500 per coin. The June 30 fair value implies roughly $58,700 per coin. Over six months, the market cut the per-coin mark by about one-third. The treasury's response was inaction โ and the cost basis confirms the inaction is not a reporting artifact. If the company had sold any meaningful tranche, the weighted-average cost basis would have moved or a realized loss would have appeared in the income statement. Neither happened.
A note on accounting method: under the FASB's 2023 crypto-asset measurement rules, companies must mark digital assets to fair value each quarter, with changes flowing through net income. That rule is why we can see the $539 million loss as a discrete line instead of buried in an impairment footnote. The transparency is an improvement over the old model, where a holding could sit at cost until it was impaired permanently. The fair-value regime turns the Bitcoin table into one of the most informative disclosure points in the entire filing.
Now the disclosure trap, which any honest analyst has to flag. A net-zero change over six months does not prove zero transactions. The filing does not disclose transaction-level activity. Purchases and sales could have occurred during the period and offset each other before quarter-end. My 2021 wash-trading audit taught me to distrust flat ledgers with opaque counterparty logic. However, the binding constraint here is the cost basis: any sale above the ~$35,300 average would have realized a gain; any sale below it, a loss. The reported cost basis matches the year-end figure exactly. To preserve that average through internal churn, the treasury would have had to buy and sell at the same average price in offsetting amounts. Possible. Not plausible. The economically sensible reading โ and the one the market has adopted โ is that SpaceX simply held.
The allocation ratio tells the structural story. At year-end 2025, Bitcoin represented approximately 6.6% of cash, cash equivalents, and marketable securities. After the IPO, that ratio fell to approximately 1.1%. The decline was not caused by selling. It was caused by dilution: $85.675 billion in IPO proceeds entered the treasury while the Bitcoin position stayed constant. The asset was mathematically diluted into insignificance. A Bitcoin "proxy" stock cannot have a 1.1% treasury weight, zero net allocation, and still be called a proxy. The data reframes the trade.
For comparison, consider the broader treasury-company cohort. Many public Bitcoin treasuries entered 2026 with cost bases near the cycle top and found themselves "millions in the red" at $78,000 per coin, before the further drawdown through June. Strategy โ formerly MicroStrategy โ continued its leveraged accumulation model, raising equity to buy more Bitcoin into weakness. Tesla went the other direction: after accumulating, it sold 75% of its position in July 2022, near a cycle low. SpaceX's behavior is different from all three. It holds a legacy position at a very low cost per coin. It does not add. It does not sell. It does not use options to hedge at the treasury level. This is a strategic reserve, not a trade and not an accumulation program. The $437 million of remaining profit above cost gives the position its own buffer โ the unrealized loss on the year is real, but the position as a whole is still more than 1.6 times its cost basis.
The question every crypto analyst will ask next: will they buy the dip? The data offers a provisional answer: no. The company watched the mark fall one-third and did nothing. It had $100 billion in liquidity and chose money-market funds. There is no evidence of a Bitcoin authorization expansion. If anything, the IPO diluted the decision-maker's percentage โ Musk's equity stake is locked, but the treasury now operates over a much larger, more fiduciary-driven balance sheet. Public shareholders did not buy a Bitcoin treasury vehicle. They bought an AI infrastructure company with a legacy BTC bag. Those are different securities, with different risk profiles, and the market is slowly adjusting.
Core, Part 4: AI Capex โ Follow the Gas, Not the Hype
Follow the gas, not the hype. That is the rule for this section, and it is the rule for the entire AI-crypto convergence trade.
SpaceX directed $15.828 billion into AI infrastructure during the second quarter. That is more than 21 times the $749 million spent a year earlier, and roughly double the first-quarter deployment. AI infrastructure absorbed 86% of the company's total quarterly capital expenditure of $18.369 billion. Wall Street had estimated AI spending at $13.09 billion; SpaceX ran 21% above that number even as total capex landed slightly below the $18.58 billion consensus. The informative miss is on composition: the company is funneling virtually all incremental capital into compute.
First-half totals confirm the trajectory. AI capital expenditure reached $23.551 billion from January through June, against $3.316 billion in the same period last year. That is 83% of the company's total $28.476 billion investment for the half. The quarter-over-quarter acceleration โ from roughly $7.7 billion in Q1 to $15.8 billion in Q2 โ indicates the buildout is compounding. The CFO's guidance that capital spending will remain at a similar level in the final two quarters means another $30 billion-plus of assets entering service by year-end. This is not a project. It is a permanent line item.
The scale is best understood by comparison. The largest technology companies in the world now run AI capex in the tens of billions per quarter, each. SpaceX is a first-year public company spending at that level while still growing its base business at 92%. That makes the capital-allocation decision the most important economic fact about the company. Everything else โ the launch cadence, Starlink subscriber growth, even the Bitcoin position โ is secondary.
Now the revenue side. AI revenue reached $2.561 billion in the quarter, driven by compute agreements with Google and Anthropic plus subscription revenue from Grok and X. That is roughly 33% of total revenue, as noted. The contracts that guarantee this revenue are significant: $14.1 billion in cloud-services agreements signed by quarter-end, with another $6.7 billion added after the quarter. Total contracted AI revenue commitments: $20.8 billion and growing.
The CFO, Bret Johnsen, claims contracted compute deployments produce payback periods of less than one year โ allowing the company to recover equipment costs faster than it does on launch sites and satellite infrastructure. That claim requires forensic scrutiny, and this is where the data begins to depart from the narrative.
First, scope. The payback statement applies to specific compute contracts, not to the AI segment as a whole. A deployment that is pre-sold to Google or Anthropic, with contracted utilization, can plausibly return its equipment cost in under twelve months. Estimate the math: if the $14.1 billion contract book covers computing capacity that cost, say, $10โ12 billion to build, the <1-year payback is credible โ assuming zero utilization risk, counterparty performance, and no renegotiation. The two anchor customers are among the strongest balance sheets in the industry. The credit risk is minimal.
Second, the segment as a whole. The AI division reported an operating loss of $1.257 billion. It booked $1.885 billion in depreciation and amortization and spent $2.178 billion on research and development during the quarter. Run the coverage ratio: segment D&A of $1.885 billion against segment revenue of $2.561 billion means depreciation alone consumes 74% of AI revenue. Add the R&D bill, and the segment's cash conversion is deeply negative by design. The company is buying a future annuity with current losses.
Third, the contract-to-capex ratio. The $20.8 billion in signed contracts, including post-quarter additions, must be measured against cumulative AI capex. Through H1, cumulative AI capex is $23.551 billion and growing by $15.8 billion per quarter. The signed contract book is roughly 0.88x cumulative capex today. For the CFO's payback claim to remain true at year-end, the contract book must grow at least as fast as the capex line. If the ratio falls below 1x, the depreciation wall will start to overwhelm the segment. If it grows above 1.5x, the market is underpricing the annuity.
Here is the crypto-native connection. Bitcoin miners spent 2025 pivoting toward AI as a bear-market escape, betting that electricity scarcity and data-center locations give them a structural edge. The SpaceX-Anthropic deal validated that thesis while introducing a lethal new competitor. A miner's edge is power-land and co-location speed. SpaceX's edge is balance-sheet scale and contract-negotiation speed. When one entity can write $13 billion checks per quarter to buy compute, it can outbid the entire mining sector for future capacity. Follow the gas physically โ the electricity and the GPUs โ and you find SpaceX building its own hyperscale data-center fleet, independent of cloud providers, signing anchor tenants before the concrete is poured.
The hash-rate analogy is exact. When I analyze Bitcoin miners, the first metric is the cost of newly mined coins versus spot price. The equivalent here is the cost of newly deployed compute versus contracted revenue. In both cases, capital intensity extends for years while revenue recognition lags. The only question that matters at the end is whether the contract book covers the asset cost within the depreciation window. Today, the answer is borderline. That is the risk the market is pricing on the sell side.
Core, Part 5: The $105 Billion Unlock and the Crowded Short
Now the supply side of the SPCX trade.
On Aug. 6 โ two days after this earnings release โ insiders become eligible to sell approximately 900 million shares worth about $105 billion at current prices. Tom Dunleavy, Head of Venture at Varys Capital, describes the release as among the largest lockup expirations in market history. That is not hyperbole; it is arithmetic. For scale, the Uber lockup that set records in 2020 was roughly a quarter of this size.
The Aug. 6 expiration is stage one of a broader float expansion. A second block becomes eligible after the company reports third-quarter earnings. Further restrictions expire Dec. 8. By the end of that sequence, roughly 40% of SpaceX's outstanding shares could be freely tradable. Elon Musk's personal stake is locked until June 2027, which means the largest holder cannot participate in any of the 2026 supply waves. The supply event is therefore a test of employee and early-investor conviction, not the founder's.
Positioning data tells the demand side of the story. S3 Partners estimates that 95% of SPCX shares available to borrow are out on loan. Short interest has reached 34% of the public float. In plain terms: nearly every share the market is willing to lend has been lent. The bearish trade is fully loaned out. There is almost no marginal share available for new shorts without a lender stepping in.
The crypto derivatives market shows a parallel surge. CoinGlass data reviewed by CryptoSlate shows SPCX futures volume and open interest at the highest levels since contract launch. Over the last 24 hours of the observation window, trading volume reached approximately $6.85 billion and open interest approached $720 million, across the crypto venues that list the contract.
These two data sources are frequently conflated in media coverage, and that conflation is an analytical error. Equity short interest is a one-sided, confirmed liability: 34% of the float is sold short, and the borrow is nearly exhausted. Futures open interest is a two-sided ledger: every contract contains both a long and a short. Rising open interest tells you new money is entering the market โ it does not tell you whether that money is net bullish or bearish. The $720 million of open interest against a market capitalization in the trillions is a rounding error in the pricing mechanism. It is temperature, not confirmation. In my 2021 NFT audit, I learned to separate reported volume from directional evidence. I separated wash trades from genuine collections and found that 30% of the reported volume was fake. The same discipline applies here: do not read $6.85 billion of futures volume as $6.85 billion of bearish conviction. Half of that open interest is longs who believe the unlock is already priced and the short is overextended.
The interplay between the unlock and the short crowd is the most volatile variable in the trade. A lockup expiration expands supply only if insiders choose to sell. If they do, the float grows, borrow availability increases, and the short thesis plays out mechanically. If they hold, the 95% borrow utilization becomes fuel for a squeeze. The late addition of the third-quarter block and the Dec. 8 release creates three separate windows of uncertainty. The market is already trying to front-run each one. That front-running is visible in the derivatives OI and in the equity short book.
There is one nuance the short crowd often misses. When a lockup expires, shares that were previously restricted often become available to borrow. A busy unlock can increase the lendable supply, which lowers the borrow rate and reduces squeeze pressure. In that scenario, the expiration actually helps the short thesis by relieving the utilization constraint. The conventional "squeeze" story assumes insiders both hold and lend. The data so far shows heavy borrowing already; the available-borrow pool at 95% utilization suggests the market is at the limit of that mechanism. The unlock may simply reset the equilibrium at a higher float.
Contrarian: The Beat Didn't Cause the Selloff, and Neither Did Any Single Pressure
The consensus read of this week's price action: "The earnings beat failed to ease concerns about SpaceX's AI spending and the imminent increase in tradable SPCX shares."
That sentence is directionally true and analytically incomplete. The selloff has two causes, and they operate on different time horizons. The AI spending concern is a multi-year question about margins, depreciation schedules, and the contract book. The unlock is a liquidity event measured in weeks. Credit must be apportioned carefully. Claiming that either one alone caused the selloff is like blaming one counterparty for a two-sided trade. The market is not distinguishing the two pressures, and my job is to separate them in the data.
Here is the contrarian proposition: the unlock is the most visible event in the filing, and the most visible event is the most priced-in event. A lockup expiration that has been public knowledge since the IPO, flagged in every sell-side note, hedged by a derivatives market at record open interest, and shorted to 95% borrow utilization is not a secret. It is a scheduled transaction. The market has had six months to position. The short crowd has positioned. The futures market has hedged. The question is not whether the event happens; it is whether the crowd's position is correct.
And crowded positions are data anomalies. When 34% of the public float is short and 95% of the lendable supply is on loan, the bearish trade is not high-conviction capital at a low-fee equilibrium. It is a crowded trade. Crowded trades resolve violently in the direction of the correction. If the unlock produces less selling than the short book expects, the buy-to-cover demand will collide with thin supply. If it produces heavy selling, the short book is rewarded equally violently. Either resolution is sharp. The open-interest data in the crypto derivatives market โ two-sided and at record levels โ suggests the market itself does not know which way that resolves.
The deeper contrarian point is for crypto holders specifically. The "Bitcoin proxy" trade is structurally dead. The first filing offers three pieces of evidence. One: Bitcoin is 1.1% of the liquidity stack and falling relative to the corporate balance sheet. Two: the cost basis is unchanged, and the treasury bought no additional BTC during a 33% drawdown. Three: the company's primary capital allocation is an AI compute buildout whose valuation will be set by infrastructure returns, not by the price of Bitcoin. Correlation is not causation. SPCX futures moved in sympathy with BTC during H1, but that is the derivatives market attaching a narrative to the stock, not the company's operations generating Bitcoin exposure. On-chain volume says otherwise: the flow of capital is from treasury into compute, not from treasury into Bitcoin. Anyone holding SPCX for Bitcoin exposure is holding the wrong instrument and relying on the wrong correlation.
Let me recall the Terra lesson as the relevant heuristic. In May 2022, the most dangerous sentence in all of finance was "the UST peg is fine." It was a narrative claim contradicted by the ledger โ the Curve pool balances were bleeding hundreds of millions per hour. The equivalent sentence this week is "the unlock is already priced." It may be true. It is also precisely what every crowded-shorts holder says before the squeeze. My rule from that 72-hour forensics session: when the narrative and the ledger diverge, bet the ledger, and document the divergence. The ledger says the event is scheduled, the position is crowded, and the underlying business is structurally repricing. The narrative says the worst is over. One of them is wrong. Analyze the unhedged side carefully.
There is one more piece of data that should concern Bitcoin treasury bulls. A company with $100 billion in liquidity, watching its crypto mark fall one-third, chose to do nothing. That is not a failure of nerve. It is a statement about capital structure. Public companies with large, fiduciary-bound treasuries do not rotate into volatile assets on drawdowns; they rotate into money-market funds on schedule. The IPO converted SpaceX from a founder-controlled accumulator into a governance-constrained institution. The 1.1% allocation ratio is the new ceiling for Bitcoin's role in this balance sheet, absent an explicit authorization change. Every Bitcoin treasury company that models itself on "buy the dip" should read that distinction carefully. The standard that matters in a drawdown is not whether you can wait. It is whether your capitalization structure allows you to allocate to volatility at all.
Takeaway: The Signals to Watch
The first quarterly report resets the terms of the SPCX trade. Let me summarize the positions in a Risk vs. Reward matrix, because this is the lens that matters for position sizing:
| Factor | Reward Scenario | Risk Scenario | | --- | --- | --- | | AI contract book | Grows above 1.5x cumulative capex; payback confirmed | Falls below 1x; depreciation wall hits segment | | BTC treasury | Cost basis unchanged; stable reserve, no realized loss | Any quarterly movement interpreted as policy shift | | Aug 6 unlock | Insiders hold; crowded shorts squeeze violently | Insiders sell; float expands; short thesis confirmed | | Cash stack yield | Interest income fully offsets net loss | Rate cuts compress the treasury profit center |
Three signals now decide the next chapter.
First, the Bitcoin cost basis. I will run the comparison the moment the Q3 filing drops. Any movement in the 18,712 BTC position โ up or down โ changes the treasury narrative instantly. Down means realized losses are being locked in. Up means the drawdown finally triggered purchases. Flat for a second consecutive quarter means the 1.1% allocation ratio is permanent, and the proxy trade is conclusively dead. I will be watching that line the way the bond market watches a payroll print.
Second, the unlock execution. Track the delta between eligible shares and actually sold shares after Aug. 6. If insiders dump, the short crowd wins and the stock finds a lower floor. If they hold, the 95% borrow utilization converts into the tightest squeeze setup in recent market memory. The $105 billion headline is noise; the ratio of sold-to-eligible shares is signal.
Third, the contract-to-capex ratio. The company spent $15.8 billion in Q2 and signed $14.1 billion in contracts before quarter-end, with $6.7 billion added after. H2 capex at the CFO's steady-state guidance adds more than $30 billion in assets. The single number that validates or invalidates the AI thesis is whether the signed contract book keeps pace: above 1.5x cumulative capex and the market is underpricing the annuity; below 1x and the depreciation wall becomes the dominant story of 2027.
Here is the forward-looking question I want every SPCX futures trader to answer by Dec. 8:
The ledger shows a $100 billion cash stack, an unchanged Bitcoin book, an AI engine consuming 86% of capex, and a fully scheduled unlock sequence. Which part of that data justifies pricing SPCX as a Bitcoin proxy? And if none of it does, what exactly is the open interest betting on?
I do not know the answer. The data will tell us by the third quarterly report. It always does.