SpaceX's June 5 free-writing prospectus says Google agreed to pay SpaceX $920 million per month from October 2026 through June 2029 for access to compute capacity including approximately 110,000 NVIDIA GPUs, plus CPUs, memory and related components. [1]

The paper's June 14 article said SpaceX had a Friday close, not a Sunday verdict. Monday's delivery date matters for the IPO tape. The Google contract matters for a different reason: it tells readers that SpaceX is being sold not only as rockets, Starlink and Musk governance, but also as an AI-compute supplier with a very large named customer.

The contract is precise enough to ruin vague enthusiasm. Capacity ramps through September at a reduced fee. If SpaceX fails to deliver access to the committed GPU amount by September 30, Google receives a one-month grace period and then may terminate immediately or accept fewer GPUs with a pro rata reduction in monthly fees. After December 31, either party can terminate on 90 days' notice. [1]

Those clauses are the story's machinery. A headline can say 110,000 GPUs. A prospectus asks when the GPUs arrive, what happens if they do not, who bears the ramp risk, and how much of the public company's expected revenue rests on one buyer's continued willingness to rent capacity.

MSM has understandable reasons to make SpaceX an IPO superlative. The final prospectus supplies the public-offering frame: 555,555,555 shares at $135, a $74,999,999,925 raise, Nasdaq listing mechanics and June 15 delivery language. [2] That is the Wall Street record. It is not the whole operating record.

The public investor's problem is that the records now run on two clocks. One clock is securities-market settlement: price, float, volume, index eligibility and voting control. The other is service delivery: GPUs, power, network capacity, uptime and customer rights. The first can be watched on a quote screen. The second will show up in amendments, revenue recognition, customer concentration and, if things go badly, termination notices.

That second clock is harder to game with ceremony.

The Google agreement bends SpaceX toward the same accounting questions now circling every AI listing. The paper's June 14 checklist for AI listings asked for cash flow, customer concentration, governance and capex. A $920 million monthly service schedule answers only the easiest part: someone is willing to buy. It leaves the harder parts: who paid for the hardware, who owns it, how revenue is recognized, how margins survive power and depreciation, and how easily Google can walk.

The free-writing prospectus says Google retains ownership and intellectual-property rights in its content, AI models and related data. [1] It does not turn SpaceX into Google. It turns SpaceX into a place where Google's compute demand can become SpaceX revenue if the machines, power, networking and service levels arrive.

X moved immediately to the simplest version of the receipt. A June 14 post called the agreement "next-level scale" and "extremely bullish" for NVIDIA and AI infrastructure. That is a fair discourse signal. It is not diligence. GPU volume can prove demand and still hide financing, concentration and termination risk.

The Motley Fool's IPO analysis, written before the final public tape, framed SpaceX around the more familiar business mix and xAI-adjacent questions. [3] The Google contract makes that adjacency less speculative. It places a cloud-service agreement inside the offering file, with dates, fees and customer rights.

The details also complicate the Musk-control story. A public SpaceX buyer is not only buying launch cadence or Starlink subscribers. The buyer is buying a company whose prospectus stack now includes a giant AI-capacity promise to Google, with a delivery test before September ends and an escape valve after December.

That creates a clock. October 2026 is when the full monthly fee begins. September 30 is when the committed GPU access must be delivered. December 31 is when 90-day termination rights open. June 2029 is when the scheduled full-rate period ends. Each date gives public investors a future receipt.

If SpaceX delivers, the IPO becomes partly an AI-infrastructure story with unusually concrete customer demand. If it misses, the prospectus already names the remedy. Either way, the rocket company has put compute capacity into the filing. Readers should follow the contract, not only the launch.

-- DAVID CHEN, Beijing