⚠ PRODUCTION CONSTRAINT NOTE, SECTION 1: Section 1 (trading performance) requires Wire amendment with June 12 first-day trading data before publication. The placeholder below marks the exact insertion point. Sections 2–5 are complete and verified.
Section 1, What Actually Happened on Day 1
[WIRE AMENDMENT REQUIRED: Insert first-day trading performance data here, opening price, intraday range, trading volume, closing price. Do not publish this deep-dive without this section completed. Once the Wire amendment arrives, replace this placeholder with the verified facts and remove this notice.]
Section 2, The SpaceX Infrastructure Thesis: Compute Landlord as Public Equity
The SpaceX equity case doesn’t rest on rocket launches or Starlink subscriptions. It rests on a simpler proposition: SpaceX built data centers, Google needs GPUs, and the contract is in the S-1.
The Google agreement, disclosed in SpaceX’s S-1 filing, runs from October 2026 through June 2029. Google pays approximately $920 million per month for access to at least 110,000 GPUs at SpaceX facilities. The math is $30.36 billion over 33 months. That’s not projected revenue. It’s contracted revenue – committed under terms that include, per prior analysis of the S-1’s termination clauses, conditions that limit Google’s ability to exit early.
For investors accustomed to evaluating SaaS companies on ARR multiples, this structure is unfamiliar. It looks more like commercial real estate than software. A single anchor tenant on a long lease. Predictable cash flow. Limited upside flexibility. That’s not a criticism, it’s a description that requires a different valuation framework. The $1.75 trillion post-money valuation implies roughly $53 per contracted dollar of Google revenue. Whether that multiple is rational depends on what investors believe SpaceX can lease to tenants beyond Google, and on the question the S-1 doesn’t fully resolve.
The xAI/Colossus overlap is the structural ambiguity. The precise allocation of compute capacity between SpaceX’s commercial data-center rentals and xAI’s Colossus cluster has not been publicly disclosed. Elon Musk controls both entities. Investors in SPCX are, on some level, trusting that the infrastructure serving Google isn’t also serving xAI’s competing AI development workloads on undisclosed terms. That trust may be warranted. But it isn’t documented in the public filing.
Section 3, Oracle’s $40B Raise: The Same Trade, Higher Cost
Oracle reported Q4 FY2026 revenue of $19.2 billion, per its official earnings release, with Cloud Infrastructure (OCI) revenue of $5.8 billion, up 93% year-over-year. Its Remaining Performance Obligations reached $638 billion, up 363% year-over-year. On paper, Oracle’s AI infrastructure buildout is working.
Then the stock fell roughly 10% in extended trading.
The market sold the beat because Oracle announced plans to raise approximately $40 billion in additional capital to fund further AI data center expansion. The terms, debt-to-equity ratio, coupon rate, timeline, haven’t been disclosed. What investors heard was: the backlog is enormous, and we need $40 billion more to build what’s required to service it.
Infrastructure vs. Model Layer, Key Financial Comparisons
| Entity | Capital Event | Revenue Visibility | Concentration Risk |
|---|---|---|---|
| SpaceX (SPCX) | $75B IPO | Google: $30.36B contracted | Single anchor tenant disclosed |
| Oracle (ORCL) | $40B raise planned | $638B RPO (backlog) | OpenAI: >50% est. (BofA, unconfirmed) |
| OpenAI | Confidential S-1 filed June 9 | Revenue undisclosed | Enterprise/API customer mix undisclosed |
| Anthropic | Confidential S-1 filed June 1 | $10.9B net rev. est. (WSJ) | Google + Amazon strategic investors |
Evidence
That’s the double edge. A $638 billion RPO is a promise of future revenue recognition. It’s committed contracts, not delivered work. Oracle’s OCI capacity today isn’t large enough to fulfill those commitments on its own timeline without the capital raise. So the backlog is simultaneously the bull case and the justification for the dilution.
Bank of America analysts have estimated that more than half of Oracle’s $638 billion RPO may originate from OpenAI contracts, though this figure is an analyst estimate and hasn’t been independently confirmed. If accurate, that level of customer concentration introduces a credit and operational dependency that matters enormously for enterprise buyers evaluating OCI as a long-term infrastructure partner.
Section 4, Infrastructure Equity vs. Model Equity: An Investor Framework
SpaceX and Oracle are different expressions of the same infrastructure trade. One is an IPO built around contracted compute. The other is a mature public company raising debt to build more of it. Both are pricing events that will shape how investors think about the OpenAI and Anthropic windows.
Here’s the framework that matters before those windows open:
*Revenue visibility vs. revenue recognition.* SpaceX has contracted revenue (Google deal). Oracle has committed revenue (RPO). These aren’t the same thing. Contracted revenue flows when service is delivered and Google doesn’t exit. Committed RPO converts to revenue as Oracle delivers work under contracts that span years. Investors who conflate backlog with current earnings will misprice both assets.
*Capital intensity.* SpaceX raised $75 billion at IPO. Oracle is raising $40 billion more. CoreWeave raised roughly $12 billion before its listing. The infrastructure layer requires continuous capital. That’s different from OpenAI or Anthropic, where the primary input is compute purchased from third parties like Oracle and SpaceX, not built. Model companies sit above the capital intensity curve. Infrastructure companies are the curve.
*Concentration risk.* Google is SpaceX’s disclosed anchor tenant. OpenAI may represent the majority of Oracle’s backlog per BofA’s estimate. Anthropic’s Series H investors include Google and Amazon. These relationships are load-bearing for multiple companies simultaneously. A strategic shift by one hyperscaler cascades across the infrastructure stack.
*Valuation comparables.* SpaceX at $1.75 trillion has no clean peer. CoreWeave’s valuation was roughly $23 billion at listing. Oracle’s market cap has historically traded at 5–7x revenue. The SPCX debut is the first attempt to price pure-play AI compute infrastructure at frontier-lab scale. Whatever multiple it establishes will anchor OpenAI and Anthropic negotiations, even though those companies are selling model capability, not compute capacity.
What to Watch
Verification
Partial SpaceX S-1 (SEC), Oracle earnings release, WSJ, BofA analyst note, no source pages fetched this cycle Section 1 (trading performance) is unpublished pending Wire amendment. BofA RPO estimate is analyst projection, not confirmed Oracle disclosure.Section 5, What to Watch Before the OpenAI and Anthropic Windows
Three data points will clarify the infrastructure equity story before OpenAI and Anthropic file publicly.
First, watch Oracle’s capital raise terms when they’re disclosed. The debt structure, coupon rate, maturity, covenant package, will tell investors whether Oracle is financing a profitable buildout or covering a cash-flow gap. A high-yield raise at wide spreads is a different signal than investment-grade debt at narrow spreads.
Second, watch for any SpaceX S-1 amendment or 8-K in the first 90 days of trading that addresses the xAI/Colossus compute boundary. If that disclosure comes with hard numbers, the infrastructure thesis gets a cleaner foundation. If it doesn’t come, that ambiguity will persist as a valuation discount.
Third, watch the Anthropic S-1 when it moves from confidential to public filing. Anthropic’s compute arrangements, who it buys from, on what terms, with what commitments, will tell you exactly how the model layer sits on top of the infrastructure layer. OpenAI’s relationship with Oracle is partially documented in the BofA analyst estimate. Anthropic’s will be disclosed in the prospectus.
The real story is that public markets are running a real-time experiment in pricing AI infrastructure before the AI model companies arrive. SpaceX and Oracle are the reference points. Every institutional investor building a position in either company is, consciously or not, making a bet on how OpenAI and Anthropic will be valued when they finally open their books. Watch the Q4 SPCX earnings call – the first one under public reporting requirements, for the first hard data on whether the Google contract’s revenue recognition timeline matches what investors underwrote at $1.75 trillion.