SpaceX is transforming its vast computing infrastructure into a high-speed cash-generating engine. On the eve of its anticipated U.S. stock market debut, Elon Musk's SpaceX has secured massive compute leasing agreements with Anthropic and Alphabet (GOOGL), with the combined contracts valued at over $70 billion and generating an annualized revenue run-rate of $26 billion, providing a powerful boost to its IPO narrative.
On June 6th, Reuters reported that SpaceX announced a multi-year cloud services agreement with Google, a subsidiary of Alphabet (GOOGL), on June 5th. According to regulatory filings, Google will pay SpaceX $920 million per month from October of this year until June 2029 for access to approximately 110,000 Nvidia GPUs and related computing resources. During a preceding ramp-up phase, Google will pay at a lower rate.
Previously, in May, Anthropic announced it would lease the majority of the compute capacity at SpaceX's Colossus 1 data center in Memphis, Tennessee, for $1.25 billion per month, with an initial commitment period of six months. Combined, these two deals will generate approximately $2.17 billion in monthly revenue for SpaceX from compute leasing, translating to an annualized figure of around $26 billion. If both contracts are fulfilled to their full terms, the total combined contract value exceeds $70 billion.
Boosting the IPO Narrative
These transactions directly strengthen the AI growth story for SpaceX's upcoming IPO. SpaceX plans to complete its listing next week, targeting a fundraising scale of up to $75 billion, which could set a record for the largest IPO in U.S. stock market history. The explosive growth of its compute leasing business provides quantifiable, high-visibility revenue. However, this development coincides with rising external doubts about the in-house research and development capabilities of SpaceX's AI lab, xAI.
Details of the Google Agreement
According to SpaceX's filings with regulators, the Google Cloud services agreement was formally signed on June 5th, 2026. The agreement stipulates that Google will use the compute capacity at a lower rate before September of this year, during a ramp-up phase. From October through June 2029, Google will make fixed monthly payments of $920 million.
The provided computing resources include approximately 110,000 Nvidia GPUs, along with CPUs, memory, and other related components.
The agreement also includes protective clauses: if SpaceX fails to deliver the agreed-upon number of GPUs by September 30th, Google may terminate the agreement immediately after a one-month grace period, or choose to accept the actual number of GPUs provided with a proportional reduction in the monthly fee.
Furthermore, after December 31st, 2026, either party may terminate the agreement with 90 days' notice. Google will retain full ownership and all intellectual property rights to its content, AI models, and related data.
Analysis suggests this means the disclosed total contract value of $70 billion is, to some extent, contingent on both parties choosing to fulfill the agreements until their expiration dates.
Anthropic's Compute Deal
Anthropic's compute agreement with SpaceX was announced in May. Anthropic will lease the majority of the compute capacity at SpaceX's Colossus 1 facility for $1.25 billion per month.
This facility, located in Memphis, Tennessee, houses over 220,000 Nvidia processors. It will provide 300 megawatts of additional computing power for the Claude chatbot and is scheduled to be fully delivered within a month.
Underlying Context of the Anthropic Deal
According to a report by The Information, this deal is linked to a lesser-known competitive entanglement. xAI had long tracked and attempted to leverage Anthropic's technology to develop its own products—including continuing to use its model outputs through various channels after Anthropic revoked access.
The report, citing six informed sources, states that xAI conducted a distillation project for several months, directly using Claude's responses to train its own programming model. In January, xAI co-founder Tony Wu informed employees that Anthropic had cut off access. However, some xAI engineers subsequently continued to access Claude using personal accounts until Anthropic gradually shut these accounts down. Two sources revealed that xAI also accessed Anthropic models through the intermediary Blackbox AI, with the most recent usage traceable to mid-May.
Elon Musk himself publicly acknowledged in a court deposition in May that xAI had "partially" used OpenAI models to train Grok, stating such practices were common in the industry. However, neither he nor the company disclosed the extent of xAI's use of Anthropic technology or the various workarounds employees employed to gain direct access.
Challenges at xAI
The Information's report reveals that xAI is facing a series of significant challenges in AI R&D, creating a tension with the AI growth story SpaceX is presenting to Wall Street.
On the personnel front, xAI has experienced major turmoil: a large number of core personnel, including eight co-founders, have departed, and the engineering team has been reduced.
According to the report, one source indicated that as of mid-May, the xAI pre-training team had fewer than five members, down from over twenty a year ago. The Grok Code project has had at least four different leads in a matter of months, with the most recent, Beibin Li, departing in May. Additionally, an employee accidentally deleted core training data for an AI programming product during a data migration, resulting in the loss of work equivalent to two to three weeks of effort.
Strategic Shifts
On a strategic level, Musk reportedly stated in an internal meeting earlier this year that he planned to bring in Cursor and Mistral AI to fill key technology gaps—with Mistral handling pre-training and Cursor handling post-training. However, early engagement with Mistral did not lead to a formal partnership: Mistral executives visited xAI headquarters in mid-April, but high-level talks between the two sides broke off after a few weeks. Mistral co-founder and engineer Devendra Chaplot joined xAI but left after only about a month in May.
The collaboration with Cursor, however, has deepened. In April, SpaceX announced it would open its infrastructure to Cursor to train its own model, Composer. SpaceX also secured an option to acquire Cursor later this year for $60 billion, with a $10 billion break-up fee if it chooses not to proceed. Cursor CEO Michael Truell is now a frequent presence at xAI offices.
Questions for the Future
According to The Information, the recent series of setbacks at xAI has led to questions about whether SpaceX will focus its efforts on in-house AI development or on its compute leasing business. However, Anthropic's payments to SpaceX are helping to offset the financial shortfall of the deeply loss-making xAI division, which reportedly caused significant losses for the overall company in the first quarter.
SpaceX's imminent IPO is highlighting compute leasing revenue as a core growth driver. Yet, the rapid expansion of this business model has also sparked external questioning about the company's strategic positioning in AI.
Musk has characterized the Anthropic agreement on his X platform as a "short-term arrangement," stating, "I think we may need the compute back at some point." SpaceX's IPO filing also states that the company is continuously training the next major version of the Grok model but has not disclosed any release timeline.
Comments