SpaceX's Shared Dependency

SpaceX went public at $135 a share, briefly touched $3 trillion in market cap, then fell 30% from its $225 peak to $156, erasing roughly $350 billion from Elon Musk's fortune in a week ^[1]. That decline matters beyond Musk's bank account. SpaceX used its peak valuation as currency: a $60 billion all-stock acquisition of Cursor (Anysphere), explicitly timed to the company's $2.97 trillion peak to minimize dilution ^[2]. Its shares fell 6% the day the deal was announced, with analysts warning of dilution ^[3]. SpaceX is spending a deflating asset to buy its way into AI. The same IPO wave that gave SpaceX that currency is also funding its compute customers. Three of them — Anthropic, OpenAI, and Reflection AI — are either filing for public offerings or operating on capital raised in the pre-IPO boom. Anthropic pays SpaceX $1.25 billion a month for access to the Colossus data center in Memphis ^[4]. Google pays $920 million a month under a deal running through June 2029 ^[5]. Reflection AI pays $150 million a month for Nvidia GB300 access at Colossus ^[6]. Anthropic's $15 billion annual payment to SpaceX alone is the largest single stream; the three together exceed $27 billion a year.

SpaceX's acquisition currency and its compute revenue both trace to the same source: a mega-IPO wave that is now contracting on SpaceX's own post-debut evidence. ^[2][4][7]

The dependency runs in one direction so far, and the evidence supports "at risk from," not "cannot pay without." Anthropic generates real revenue: one source reports a $47 billion annual revenue run rate ^[4], while another reports $30 billion in annualized revenue ^[8]. I use the higher figure because it reflects a run rate from a company whose revenue tripled from end-2025, but the discrepancy itself is worth noting — the numbers are moving fast enough that even the filings don't fully agree. Either way, $15 billion in annual SpaceX payments is a fraction of Anthropic's top line, and the company has diversified backers in Amazon, Google, and Microsoft ^[9]. Anthropic is not insolvent. But it has also committed over $100 billion to AWS over the next decade, plus 1 million Google TPUs and Nvidia GPUs ^[8], and it launched a price war ahead of its IPO, cutting token pricing to hold enterprise customers ^[10]. If Anthropic's public debut deflates, its capacity to keep paying $15 billion a year to SpaceX is at risk — a risk the underlying facts point to even if no source names it directly. The exposure is to Anthropic's growth trajectory and margin sustainability, not to its immediate solvency. What makes this a shared dependency rather than two separate stories is that SpaceX's own post-IPO decline is now constraining the next wave. OpenAI is weighing a delay of its IPO to 2027, citing SpaceX's volatile post-debut performance as evidence that retail investor enthusiasm is waning ^[7]. Sam Altman has rejected any valuation below $1 trillion despite private marks ranging from $730 billion to $852 billion ^[7]. If OpenAI delays, SoftBank's $40 billion bridge financing — due March 2027 — would come due before the IPO can convert its OpenAI stake into tradable shares ^[11]. SoftBank's attempt to secure a $6 billion margin loan against that stake has already stalled because lenders cannot price a privately held AI entity during market corrections ^[11]. One IPO's stumble is feeding into the next IPO's financing problem. The broader market complicates the story without refuting it. The SpaceX IPO was absorbed on day one without a crash — the S&P 500 and Nasdaq hit all-time highs that day, and Robinhood saw record traffic ^[12][13]. The Dow reached 51,562 in early June ^[14]. Robert R. Johnson argues that IPO raises are a minute fraction of the $79.4 trillion total U.S. market cap and that institutional dry powder can absorb the listings ^[15]. He is probably right about the market as a whole. But within AI, the drain is real and the market already discriminates. Nvidia reported $81.6 billion in first-quarter revenue with 211% earnings growth and controls about 90% of the AI data-center market ^[16][14] — proven AI revenue is being rewarded. SpaceX, whose AI segment lost more than $6 billion in 2025 against Starlink's $11.4 billion in revenue ^[3], is on the other side of that line. Its IPO filing projected a $28.5 trillion total addressable market with 93% tied to AI ^[5]. The market is pricing proven chips at a premium and speculative compute at a discount, and SpaceX is the speculative side.

Then. SpaceX's IPO peaked at ~$3T and the company immediately deployed all-stock currency for a $60B Cursor acquisition ^[2]

Now. SpaceX has fallen 30% to $156, and the same stock is the acquisition currency the market is repricing downward ^[1][3]

Tom Essaye of Sevens Report Research draws the parallel explicitly: while people connected to the internet, their connection wasn't nearly as profitable as quickly as everyone assumed ^[17]. Oracle declined about 25% after significant capex, and a broader stall could trigger order cancellations across Nvidia, Micron, Broadcom, and SanDisk — the same supply chain SpaceX's compute business depends on ^[17]. Morgan Stanley reports hyperscalers doubled gross leverage from 0.9x to 1.8x in two quarters, and UBS tracks $800 billion in additional U.S. credit funding AI projects over the past year ^[18]. The credit underpinning this capex is mounting at exactly the moment the IPO wave that replenishes equity is narrowing. SpaceX's compute contracts carry exit clauses that would turn market risk into contractual fact. Google's deal includes a termination clause if SpaceX fails to deliver the full GPU commitment by September ^[5]. Reflection AI has a 90-day exit window after three months ^[6]. These are contractual triggers, not financial ones — they depend on SpaceX's delivery performance, not on customer valuations. But if the IPO wave stalls and customers' own public-market valuations decline, the financial incentive to exercise those exits would increase. That second step is an inference from the underlying economics, not something the sources state: a customer whose stock is down and whose IPO proceeds are diminished has less reason to keep paying premium rates for capacity it might not need. No source closes the return path — customer payment failure leading to SpaceX stock decline leading to further IPO stalling. The evidence supports a shared dependency on the same contracting wave, not a proven feedback loop. SpaceX is not alone in this exposure. OpenAI filed for IPO with a $38.5 billion annual loss against $13 billion in revenue, planning to spend roughly $600 billion on infrastructure before seeing a return, potentially not until 2030 ^[19]. It pays Microsoft $17.2 billion a year and projects $600 billion in data-center commitments by 2030 ^[19]. Cerebras, the rare profitable AI infrastructure company with $87.9 million in 2025 profit, went public at $26.6 billion ^[20] — a reminder that the market is not rejecting AI wholesale. It is separating companies that make money from companies that spend it. The mega-IPO wave — SpaceX at $2 trillion, OpenAI at $1 trillion, Anthropic at $965 billion — adds up to roughly $4 trillion in intended market value ^[21]. Goldman Sachs CEO David Solomon described the pre-IPO environment as having more greed than fear ^[22]. The IPO wave drained capital from elsewhere: $5 billion left Bitcoin ETFs as speculative capital rotated into the SpaceX listing ^[23], and Vanda Research reported retail traders on track for a third consecutive day of net selling across single names — something not seen since March 2020 ^[24]. The IPO market was already showing strain before SpaceX: Ledger paused its $4 billion listing, Consensys delayed to autumn, Kraken paused its plans, and BitGo fell 36% since January ^[25]. State investment leaders from four U.S. states challenged Nasdaq and FTSE Russell over fast-track IPO rules that force passive funds to buy billions in shares of volatile, high-valuation companies ^[26]. Oregon Treasurer Elizabeth Steiner said she was deeply troubled ^[26]. S&P Dow Jones maintained traditional criteria, diverging from Nasdaq and FTSE Russell ^[26]. The institutional plumbing is under stress even if the surface market looks calm. SpaceX is spending deflating stock on a $60 billion acquisition at the same moment its compute customers' ability to fund $27 billion a year in combined payments is exposed to the same public-market sentiment that SpaceX's own decline is eroding. The company's AI thesis — 93% of its addressable market — sits on the speculative side of the line the market has already drawn. The window is narrowing — on SpaceX's own evidence.