SpaceX Holds Bitcoin Flat While $100B Flows Into Money-Market Funds: An Audit of the Post-IPO Balance Sheet

The data shows a mismatch. SpaceX reported $7.814 billion in Q2 revenue, a 92% year-over-year jump. Adjusted EBITDA came in at $3.538 billion, 191% higher. Wall Street expected $6.8 billion and about $2 billion, respectively. The earnings beat is real. But the balance sheet tells a different story: 18,712 BTC, unchanged since Dec. 31. Cash, cash equivalents, and marketable securities approached $100 billion. That is not a Bitcoin treasury. That is a cash pile. According to the SEC filing, SpaceX parked $65.625 billion in money-market funds at June 30. Government securities account for another $4.011 billion in cash equivalents and $6.487 billion in marketable securities. Together, $76.123 billion sits in yield-bearing, risk-free instruments. The Bitcoin position? Fair value fell from $1.637 billion to $1.098 billion. That is a $539 million unrealized loss. The cost basis stays at $661 million. No net purchases. No net sales. Over six months, SpaceX did nothing with its Bitcoin. This is the first quarterly report since the record IPO in May 2026. The company raised $85.675 billion in net proceeds, lifting its cash position to nearly $100 billion. Now the market faces a new problem: how to price a company that holds Bitcoin but does not accumulate it, spends aggressively on AI infrastructure, and is about to face the largest lockup expiry in history. The market has treated SpaceX as a Bitcoin proxy since its IPO filing. That framework is now obsolete. The data shows the company is not adding to its Bitcoin position. It is not hedging with Bitcoin. It is not using Bitcoin for yield. The only active treasury decision is the allocation into money-market funds. From an economic security perspective, the treasury is conservative, not aggressive. Bitcoin has been relegated to a non-operational asset. The question for crypto investors is whether they still need to price in SpaceX at all. Start with the cash. The post-IPO liquidity went into money-market funds and government securities. A textbook institutional custody move. In my MPC vault audits, I see the same pattern: when a company is unsure of its investment thesis, it defaults to treasuries. Bitcoin is not a treasury asset; it is a remnant. At year-end, Bitcoin was 6.6% of cash assets. Now 1.1%. The dilution is arithmetic, not a sell-off. SpaceX did not reduce its position. It just grew everything else around it. That raises a question: why hold 18,712 BTC at all? The filing does not disclose transaction-level activity. Purchases and sales could have offset each other before quarter-end. But the matching cost basis suggests no realized gains or losses. From an audit view, the position is inert. It provides no earnings benefit. It produces no yield. It is a one-way liability on the balance sheet. Yes, it remains $437 million above cost. That is mark-to-market noise. The fair value dropped $539 million in six months. Losses are the only active trend. Now the AI capex. SpaceX spent $15.828 billion in Q2 on AI infrastructure. That is 21 times the $749 million spent in the same quarter last year. It is roughly double Q1. It accounts for 86% of total quarterly capital expenditure. Wall Street estimated $13.09 billion for AI spending, but SpaceX overshot even that. Across the first half, AI capex reached $23.551 billion, or 83% of the $28.476 billion total investment. The company is building data centers, leasing compute capacity to Google and Anthropic, and selling Grok and X subscriptions. AI revenue hit $2.561 billion. That revenue is real. Let me stress-test Johnsen's payback claim. A one-year payback requires monthly cash flow equal to the upfront cost divided by twelve. For a $10 billion deployment, that is $833 million per month per contract. The AI division generated $2.561 billion in quarterly revenue. That is $854 million per month across the entire portfolio. The math does not reconcile. The payback claim applies to a subset of contracts with favorable terms. The rest of the portfolio is unproven. But the segment still loses money. The AI division recorded a $1.257 billion operating loss. Depreciation and amortization: $1.885 billion. R&D: $2.178 billion. Together, those costs exceed the segment's revenue. CFO Bret Johnsen says contracted compute deployments produce payback periods of less than one year. I have audited similar claims from mining companies pivoting to AI. The contract-level IRR and the segment-level P&L are different instruments. Using one to justify the other violates the constraint-based approach I rely on. The equipment costs are front-loaded. The depreciation will continue for years. Johnsen says capital spending will remain at similar levels in Q3 and Q4. That implies an annualized AI investment exceeding $60 billion. The balance sheet can absorb it. The income statement cannot. Now the supply event. The earnings beat failed to calm the market. SPCX shares face a lockup expiration on Aug. 6. Insiders become eligible to sell approximately 900 million shares worth $105 billion at current prices. Tom Dunleavy, Head of Venture at Varys Capital, says this is among the largest lockup expirations in market history. The shares already exist. The expiration just makes them tradable. Employees and early investors will decide whether to sell. Short sellers have already positioned. S3 Partners estimates 95% of SPCX shares available to borrow are on loan. Short interest is 34% of the public float. That is a crowded trade. The crypto derivatives market confirms the tension. CoinGlass data shows SPCX futures trading volume climbed to $6.85 billion in the last 24 hours. Open interest approached $720 million. Rising open interest does not tell you whether traders are long or short. Each futures contract includes both sides. But it tells you that leverage is building into the announcement. I have reviewed similar derivatives books for private lending protocols. The open interest spike before a known event is a warning sign. It does not matter whether the crowd is right. What matters is the exit capacity. Short squeezes are two-sided. The crypto derivatives data does not break down long vs. short open interest. The $720 million could include hedges. In my audit experience, OI spikes before unlocks are volatility plays, not directional bets. The shorts are in the stock market. Futures are catching up. The unlock distribution will determine the direction. Here is the contrarian angle. The consensus says the earnings beat matters. It does not. The consensus says the Bitcoin position is a signal. It is not. The real vulnerability is the interaction between the unlock and the short interest. A $105 billion supply event does not respond to EBITDA. It responds to forced sellers. If employees sell into a market with 95% borrow utilization, the stock price gaps down. The shorts may have a valid thesis, but they are crowded. Any positive catalyst can trigger a squeeze. That is speculation, not analysis. I do not speculate. I audit. The DAO was a warning we ignored. I spent six months forensically decomposing EVM opcode execution, mapping the reentrancy vulnerability to 12,000 lines of assembly. The root cause was not the function call. It was the interaction between two mechanisms: the fallback function and the gas stipend. Here, the interaction is between the lockup expiration and the borrow rate. Code doesn't lie; audits do. The financial engineering behind the unlock is the code. The narrative around the earnings beat is the audit. The narrative is wrong. What comes next? Watch Aug. 6. The Bitcoin balance is a footnote. The AI capex is a commitment. The lockup is the test. If SPCX holds support, it tells you that institutional demand can absorb a $105 billion supply shock. If it does not, the short trade wins. Trust is a bug, not a feature. The market should not trust the unlock schedule. It should verify the supply. The evidence will arrive when the first insider 13D lands. Zero knowledge, maximum proof. The proof will be in the volume, not the press release.