SpaceX began trading on June 12, 2026, one day after confirming a $135 share price for its IPO. The market headline was obvious: a historic debut that sent the stock sharply higher. The more important AI story was in the filings. SpaceX has already disclosed two enormous compute agreements — Anthropic at $1.25 billion per month through May 2029 and Google at $920 million per month from October 2026 through June 2029 — turning AI infrastructure from an internal xAI cost center into a revenue line public investors can model.
The IPO headline was big, but the filing details were bigger
TechCrunch reported on June 11 that SpaceX priced shares at $135 and raised $75 billion from the share sale to underwriters ahead of Friday trading. AP then reported that the stock jumped more than 19% in its June 12 debut. That matters because investors were not only buying launches, satellites, and Musk optionality. They were also buying into a company that now presents AI compute as part of its growth story.
SpaceX’s June 5 prospectus says the company can use its compute infrastructure both for its own AI applications, including Grok-5 training at COLOSSUS II, and for third-party customers. That framing matters. It puts SpaceX closer to an infrastructure supplier for frontier AI, not just an operator of its own models.
What SpaceX has already disclosed about AI compute revenue
The clearest signal is the size and structure of the contracts SpaceX has already put on the record. The Anthropic agreement covers compute capacity across COLOSSUS and COLOSSUS II, while a separate filing discloses a Google cloud services agreement signed on June 5.
Disclosed AI compute agreements tied to SpaceX
| Customer | Disclosed capacity | Monthly fee | Term |
|---|---|---|---|
| Anthropic | About 325,000 NVIDIA GPUs plus CPUs, storage, and networking across COLOSSUS and COLOSSUS II | $1.25 billion | Through May 2029, with ramping in May and June 2026 |
| About 110,000 NVIDIA GPUs, CPUs, memory, and related components | $920 million | October 2026 through June 2029, with ramp-up through September 2026 |
Google’s side agreement is especially revealing for enterprise readers. TechCrunch reported that Google described the deal as short-term bridge capacity to meet stronger-than-expected demand for Gemini Enterprise and its agent platform. That suggests the current bottleneck is not only model quality. It is also whether vendors can secure enough infrastructure to keep high-demand AI products available at commercial scale.
Why this changes the reading of Anthropic and OpenAI’s IPO plans
SpaceX’s debut gives the market a live benchmark for how investors may start valuing AI infrastructure cash flows, not just model revenues or application subscriptions. Axios reported on June 12 that SpaceX’s IPO process could become a model for Anthropic and OpenAI as both companies head toward public markets later in 2026.
That comparison matters because frontier AI companies are increasingly judged on three linked questions: who controls compute, who can finance expansion, and who can translate model demand into durable contracts. In that sense, SpaceX’s IPO is not just a space or Musk story. It is a public-market test of whether AI capacity itself can support premium valuations.
What businesses should watch next
The next signal is whether more large compute agreements become public, especially agreements tied to agent platforms, enterprise AI suites, and long-running inference workloads rather than one-off training runs. If more vendors start treating bridge capacity as a normal response to demand spikes, procurement risk may become a bigger part of AI rollout planning.
For businesses building AI agents and automation, the practical takeaway is straightforward: infrastructure access is becoming part of product risk, pricing risk, and deployment timing. The companies that look safest at the interface layer may still depend on fragile or expensive capacity underneath. SpaceX’s June 12 debut makes that reality much harder for the market to ignore.