The Contract as IPO Argument
SpaceX's decision to amend its prospectus days after signing the Google deal is the clearest sign that the compute contracts are serving a narrative function as much as a commercial one. The filing added Google's $920 million monthly commitment — 110,000 NVIDIA GPUs, CPUs, and memory, running from October 2026 through June 2029 — directly to the investor-facing document . The sequencing tells the story: SpaceX needed evidence that its data center infrastructure had marquee customers before the IPO window closed, and Google's signature provided it.
The problem is that the evidence is forward-dated. None of the Google revenue has begun; the contract starts in the fourth quarter of 2026 . The Anthropic deal, which predates the Google contract and was promoted as a three-year commitment, is contractually cancellable in 90 days . SpaceX's S-1 presents these as durable infrastructure relationships. The fine print makes them options with long headlines attached.
What the S&P Rejection Actually Prices In
The S&P 500's decision to reject SpaceX — declining to waive the requirement that a company show profitability across the most recent quarter and the prior four — is the institutional market's answer to the IPO narrative. Index committees are not hostile to AI companies on principle; they are applying a standard that SpaceX, alongside OpenAI and Anthropic, cannot currently meet. The rejection means that the largest passive investment vehicles in the world will not be forced buyers of SpaceX equity at IPO — a significant constraint on the automatic demand that index inclusion would have generated.
The profitability question is not incidental to the AI infrastructure story — it is central to it. SpaceX is arguing that compute-contract revenue from Google and Anthropic justifies a valuation built on AI infrastructure margins. The S&P committee is arguing that contracted future revenue is not the same as demonstrated earnings. Both positions are internally consistent. The committee's position has more institutional weight.
The Closed-Loop Problem
The skepticism that cuts deepest in the public conversation is not about SpaceX specifically — it is about the architecture of AI spending that the Google deal makes visible. When Google pays SpaceX for compute that runs on infrastructure connected to xAI, and xAI competes with Google's Gemini, the capital is moving in a circle rather than generating new demand . The deals validate each other's balance sheets without necessarily validating that an external market is paying for AI services at the implied scale.
The counter to this is the argument that SpaceX and OpenAI are signal companies — that their success, if it comes, will flow outward to the supply chains and infrastructure vendors that depend on them . That argument requires accepting that the present-day compute deals are investment in future demand rather than evidence of current demand. It is a reasonable bet. It is not the same as the bet the IPO multiple asks investors to make.
Retail Exposure Without Retail Choice
The least-discussed dimension of the SpaceX IPO is the one that affects the most people: passive investors with 401(k) exposure to NASDAQ 100-linked funds will hold SpaceX once it clears the listing requirements, regardless of whether they have evaluated the AI infrastructure thesis . This is not unique to SpaceX — it is how index mechanics work — but the specific combination of a future-dated revenue profile, a contested profitability record, and an AI narrative that the S&P committee has already declined to credit makes the exposure consequential.
The investors who will price the SpaceX IPO in the secondary market are not primarily analysts who have read the 90-day cancellation clause in the Anthropic contract. They are funds that buy what the index includes. SpaceX's IPO team knows this, which is why the compute contracts appear in the prospectus amendment and the cancellation terms do not appear in the press releases.
The Category Migration and Its Limits
SpaceX's attempt to migrate from aerospace company to AI infrastructure provider is the most ambitious rebranding in the current IPO cohort. The Google deal is the clearest proof point — a near-$1 billion monthly commitment from one of the largest AI spenders in the world is not a minor data center contract . It establishes that SpaceX can attract hyperscaler customers and that its compute capacity is genuinely in demand.
What the migration cannot resolve is the timeline mismatch. The AI infrastructure identity is built on contracts that begin in late 2026 and carry exit options that could be exercised in 90 days. The IPO is now. Investors are being asked to price a company as AI infrastructure before that infrastructure has generated a single dollar of the contracted Google revenue. The companies that benefit from SpaceX's success — the supply-chain validators — are not yet visible on any balance sheet. The data centers exist. The durable infrastructure business they are supposed to anchor is still a contractual claim.