SpaceX signs a multiyear $30 billion computing deal with Google ahead of its IPO timetable

SpaceX has reportedly secured a massive multiyear cloud-computing agreement with Google, a deal described as roughly $30 billion in total value and disclosed shortly before SpaceX’s planned initial public offering. The arrangement underscores how quickly artificial-intelligence demand is pulling in infrastructure spending, from data centers to launch and satellite-adjacent supply chains.

According to reporting circulated by market outlets, Google will pay SpaceX $920 million per month for compute capacity beginning in October 2026 through June 2029. At full run-rate, the contract implies more than $30 billion in revenue over the period, though the exact total could vary depending on ramp-up schedules and any fees tied to performance or delivered capacity.

The contract also specifies the scale of hardware involved. Google will gain access to about 110,000 Nvidia GPUs, along with associated supporting components such as CPUs, memory, and related infrastructure, based on terms described in a post summarizing disclosures tied to a regulatory filing. This is the sort of “GPU lease” structure that matters to AI builders because it converts uncertain capital expenditure into a predictable service cost, while letting the supplier monetize large clusters without waiting to fill every server itself.

SpaceX’s obligation is not just about delivering equipment on paper. The reported terms describe a ramp of capacity through September at reduced fees, and a mechanism if the full committed GPU availability is not delivered by September 30. In that case, Google would receive a one-month grace period, after which the agreement could be terminated or adjusted, including a possibility of pro-rata monthly fee reduction tied to delivered capacity. That kind of clawback language can become central in cases where supply chain constraints or integration delays hit delivery timelines.

The reported agreement includes additional duration and exit details. After December 31, either party may terminate the contract with 90 days’ written notice, according to the same summary of the disclosed terms. The reporting also indicates Google retains ownership of its content, AI models, and related data, reflecting common practice in cloud and machine-learning contracting where the customer’s intellectual property remains with the buyer even when the compute comes from a third-party provider.

For Alphabet, Google’s parent, the deal represents yet another form of AI infrastructure monetization, even as Alphabet itself also invests heavily in its own data-center buildout. Instead of owning every component outright, Google can secure large, dedicated capacity while it continues training and serving AI models that require constant access to high-throughput hardware. For SpaceX, the revenue is notable because it comes from an industrial-grade compute service rather than aerospace launches alone, diversifying the company’s cash-flow profile as it seeks public-market valuation.

The timing is particularly striking. Multiple outlets characterized the contract as landing about a week before SpaceX’s IPO window, suggesting it could be used to demonstrate forward revenue visibility to potential investors. Investors often look for “de-risking” indicates ahead of listings, and a large, multi-year compute contract can function as a concrete demand indicator, even if it still carries execution risk related to hardware delivery and integration.

What remains unclear is how the company itself framed the deal publicly and exactly which regulatory document sections spelled out the economics and operational obligations. The available reporting focuses on the key headline terms, but does not fully detail whether the contract is enforceable at all points without carve-outs, how performance is measured, or what happens if demand from Google changes before or after the ramp period.