by Michael Hicklen - 12 hours ago - 3 min read
Google has agreed to pay Elon Musk’s SpaceX $920 million per month for access to massive computing capacity as demand for artificial intelligence infrastructure continues to surge, according to regulatory filings and multiple media reports. The multi‑year deal, disclosed in a Securities and Exchange Commission (SEC) filing, will run from October 2026 through June 2029, giving Google access to approximately 110,000 Nvidia GPUs, along with CPUs, memory and other infrastructure hosted by SpaceX’s data centers.
The arrangement comes just weeks before SpaceX’s long‑anticipated initial public offering (IPO) and mirrors another high‑value compute pact the company struck with AI firm Anthropic, wherein Anthropic agreed to pay SpaceX $1.25 billion per month for similar capacity. Combined, these deals underscore how SpaceX is positioning itself as a major provider of AI cloud infrastructure alongside legacy data‑center operators and specialized “neocloud” players.
Under the contract terms, Google will gradually scale up payment and capacity through September 2026, starting at a lower rate before reaching the full $920 million monthly commitment. The arrangement includes a provision that allows Google to terminate the agreement if SpaceX fails to deliver the agreed GPU capacity by September 30, 2026, while both companies can cancel the deal after December 31, 2026 with 90 days’ notice.
The NVIDIA GPUs involved in the deal are essential for training and running large language models and generative AI systems, including Google’s Gemini Enterprise platform and other cloud‑based AI services. Industry observers view the agreement as a way for Google to shore up critical compute capacity amid fierce competition with rivals such as Microsoft, Amazon Web Services, and bespoke AI infrastructure providers.
The timing of this compute agreement, alongside the similar Anthropic deal, has significant implications for SpaceX’s financial narrative as it heads toward its IPO. Analysts say long‑term, high‑value contracts like these could strengthen SpaceX’s revenue outlook and support investor confidence ahead of its anticipated public market debut. The combined value of the Google and Anthropic deals, if they run their full terms, could exceed $70 billion over multiple years and contribute upwards of $26 billion in annual contracted revenue.
SpaceX’s data‑center capability, anchored by facilities like Colossus 1 near Memphis, Tennessee, and underpinned by hundreds of thousands of Nvidia processors, is rapidly becoming a strategic asset in the AI cloud ecosystem. By monetizing this capacity through external compute leasing, the company is diversifying its business beyond aerospace and satellite internet into the lucrative and rapidly growing AI infrastructure market.
For Google, securing access to this scale of GPU capacity is a clear signal of how intensively legacy tech giants are investing in AI infrastructure to support next‑generation models and enterprise AI workloads. Even with its vast global data‑center footprint and custom accelerators, Alphabet’s sustained reliance on rented GPU infrastructure highlights both the acute demand for AI compute and the limits of existing deployments amid exploding usage.
The backdrop of these compute deals also highlights the widening scope of competition in cloud and AI infrastructure: alongside major hyperscalers, newer entrants such as SpaceX and specialized providers like CoreWeave and Nebius are securing high‑value contracts that were once the exclusive domain of traditional cloud platforms.