SpaceX's Most Consequential Pre-IPO Move: A $30 Billion Computing Deal That Transforms the AI Infrastructure Landscape

Just seven days before SpaceX (ticker: SPCX) is set to begin trading on the Nasdaq in what is expected to be the largest initial public offering in history, the Elon Musk-led aerospace and technology company dropped a bombshell that fundamentally reshapes the financial and strategic case for owning its stock: a $30 billion computing power agreement with Alphabet Inc.'s Google, disclosed in a regulatory filing on Friday, June 5, 2026. Google has agreed to pay Elon Musk's SpaceX $920 million a month for computing power as part of a cloud services deal that runs through mid-2029, its second such agreement with an AI competitor in a matter of weeks. Google will pay SpaceX the monthly fee from October this year through June 2029, SpaceX said in the filing Friday. The deal does not merely add $30 billion to SpaceX's revenue pipeline — it fundamentally repositions the company as one of the most important AI infrastructure providers on the planet, transforming its investment thesis from "rocket company with a satellite internet business" to "indispensable computing infrastructure platform for the AI supercycle."

The Deal in Full: $920 Million Per Month, 110,000 Nvidia Chips, and a 32-Month Run

The specific financial and operational terms of the SpaceX-Google computing agreement, disclosed in SpaceX's IPO-related regulatory filing, reveal the extraordinary scale and specificity of what has been arranged:

  • Monthly Payment: Google has agreed to pay SpaceX $920 million per month — or approximately $11 billion annually
  • Contract Duration: From October 2026 through June 2029 — a 32-month period
  • Total Contract Value: Approximately $29.44 billion over the 32-month period — widely described as "about $30 billion"
  • Computing Assets Provided: Access to approximately 110,000 AI chips from Nvidia — among the most powerful AI accelerators available
  • Contract Start Date: October 2026 — four months from now
  • GPU Delivery Condition: If SpaceX fails to deliver the committed number of GPUs by September 30, 2026, Google may either terminate the agreement immediately after a one-month grace period or accept the delivered GPU quantity and proportionally reduce the monthly payment
  • Termination Rights: Either party may terminate the agreement with 90 days' prior notice after December 31, 2026
  • Data and IP Rights: Google will retain ownership and intellectual property rights to its content, AI models, and related data
  • Google's Stated Rationale: A Google Cloud spokesperson stated that this agreement will help the company meet its AI service demands

Why This Is Google's Second AI Competitor Deal in Weeks — The Strategic Context

The SpaceX-Google computing agreement is not happening in a vacuum — it is the second major computing deal Google has struck with an AI-sector competitor in a matter of weeks, revealing a strategic pattern that is reshaping how the world's most powerful technology company manages its AI infrastructure capacity. SpaceX reached a similar deal last month to provide Anthropic, another leading AI lab, with computing power. Anthropic, which is also expected to go public this year, is paying SpaceX $1.25 billion a month. The Anthropic deal — at $1.25 billion per month versus Google's $920 million — is actually larger on a monthly basis, though both represent extraordinary commitments to SpaceX's computing infrastructure.

For Google, the logic of paying a rival's parent company for computing capacity may seem paradoxical at first — why would the world's most sophisticated cloud computing organization need to buy computing capacity from SpaceX? The answer reveals something important about the current state of AI infrastructure demand: Google said the deal would give it access to about 110,000 AI chips from Nvidia, which Google said would help it meet larger-than-expected customer demand for its AI models. The tech giant said in April that its cloud business had contracts totaling $460 billion that had yet to be fulfilled as revenue, indicating enormous demand for its services. That $460 billion of unfulfilled cloud contracts represents a demand backlog so large that Google cannot build data centers fast enough to address it through internal investment alone — making external computing capacity agreements not a sign of weakness but a pragmatic response to unprecedented scale of demand. For the complete financial and strategic analysis of the deal's implications for both Google and SpaceX, Bloomberg's exclusive reporting on the agreement provides comprehensive detail drawn directly from the SpaceX regulatory filing.

The xAI Connection: How SpaceX Became an AI Infrastructure Titan

To understand how SpaceX — a company founded to build rockets and colonize Mars — has become a provider of AI computing capacity to the world's biggest technology companies, you need to understand the xAI connection. The agreement helps establish SpaceX — which owns Elon Musk's artificial intelligence lab, xAI — as a major infrastructure provider as companies compete in a fierce global race to dominate AI. SpaceX's merger with xAI earlier in 2026 transformed the company's asset base: xAI had been aggressively acquiring Nvidia GPU clusters — building one of the largest privately held concentrations of AI computing infrastructure in the world at its Memphis, Tennessee data center facility. By merging xAI into SpaceX, Musk unified the computing assets, the space launch capability, the Starlink satellite internet infrastructure, and the AI model development under a single corporate umbrella — creating the precise multi-layered platform that makes the SpaceX IPO such a uniquely attractive and complex investment proposition.

The computing deals with Anthropic and Google represent the commercial monetization of xAI's GPU stockpile. Rather than using all 110,000+ Nvidia chips solely for xAI's own model training and inference, SpaceX is effectively running a private GPU rental business — leasing its AI computing capacity to the companies that need it most and cannot access it fast enough through their own procurement channels. At $920 million to $1.25 billion per month per customer, this is an extraordinarily high-margin revenue stream that significantly changes the financial modeling for SpaceX's public market valuation.

The IPO Timing: Why This Deal Was Disclosed Exactly One Week Before Trading Begins

The timing of the Google computing deal disclosure — exactly one week before SpaceX's June 12 Nasdaq trading debut — is clearly not accidental. SpaceX's IPO was priced at $135 per share on June 3, seeking to raise $75 billion at a $1.75 trillion valuation. At that valuation, investor concerns about the company's $4.94 billion net loss and $18.67 billion in 2025 revenue were significant — the revenue multiples implied by the valuation are extraordinary even by AI era standards.

The Google computing deal — disclosed in the IPO regulatory filing process — serves multiple functions simultaneously. It provides concrete, contracted, near-term revenue certainty that addresses the profitability concerns head-on: $920 million per month from Google alone, beginning in just four months, represents approximately $11 billion in annualized contracted revenue from a single customer. Combined with the Anthropic deal at $1.25 billion per month, SpaceX's annual recurring revenue from compute leasing alone approaches $26 billion — more than the company's entire 2025 revenue base. This contracted revenue pipeline transforms the SpaceX investment thesis from "bet on future growth" to "partially de-risked by contracted cash flows" — a much more comfortable investment proposition for institutional allocatees who needed revenue visibility before committing billions to the IPO.

The Nvidia Dimension: 110,000 Chips and the Scarcity Arithmetic

The specification that Google will gain access to approximately 110,000 Nvidia AI chips through the SpaceX deal provides a critical window into why this contract is worth $920 million per month. Nvidia's most advanced AI accelerators — including the H100, H200, and Blackwell GB200 series GPUs — are among the most scarce and sought-after industrial commodities on Earth. At the height of the AI infrastructure buildout, companies were routinely paying $30,000–$40,000 per GPU for purchase, with rental rates running $2–$8 per GPU per hour for cloud access through standard hyperscalers.

At 110,000 GPUs for $920 million per month, the implied rental rate works out to approximately $8,360 per GPU per month — or roughly $11.60 per GPU per hour. This premium pricing reflects both the scarcity of high-end GPU capacity at scale and the specific value proposition SpaceX is offering: a single, integrated, dedicated cluster of 110,000 chips rather than shared, on-demand cloud capacity. For a company running large AI model training workloads — which require massive, coordinated parallel processing across thousands of chips simultaneously — having a dedicated cluster under a long-term agreement provides reliability, latency, and coordination guarantees that shared cloud capacity simply cannot match. Google's willingness to pay this premium reflects exactly that calculus.

What This Means for the SPCX IPO Valuation and the Nasdaq-100 Thesis

The Google computing deal has immediate and significant implications for how investors should evaluate SpaceX's $1.75 trillion IPO valuation and the company's first-day trading dynamics when SPCX begins trading on June 12. The deal adds a contracted revenue stream of approximately $11 billion annually from Google alone, beginning in October 2026 — less than four months away. Combined with the Anthropic deal, SpaceX's compute leasing revenue pipeline represents a near-term cash flow contribution that substantially closes the gap between the company's 2025 loss position and the revenue scale that would justify its IPO valuation.

For the Nasdaq-100 inclusion thesis — where SpaceX's potential fast-track inclusion after 15 trading days could trigger enormous passive fund buying — the Google deal strengthens the case that SPCX will trade well above its $135 IPO price, making any Nasdaq-100 inclusion threshold based on market cap easier to achieve and sustain. A company with $30 billion in contracted Google revenue, $1.25 billion monthly from Anthropic, 6 million Starlink subscribers, dominant commercial launch market share, and a $100 million US government equity investment in its quantum computing subsidiary is a materially different investment proposition than a rocket company with a satellite business — and the Google deal is the single disclosure that most clearly communicates that transformation to the widest audience.

Google's Strategic Calculation: Why Pay a Competitor for Computing?

For Google — whose parent company Alphabet is simultaneously seeking fresh capital through its own $80 billion equity offering, whose cloud business is under intense competitive pressure from AWS and Microsoft Azure, and whose AI model capabilities are being challenged by OpenAI's GPT-series and Anthropic's Claude — the SpaceX computing deal represents a pragmatic resolution to a genuine infrastructure bottleneck. The $460 billion in unfulfilled cloud contracts represents demand that Google's existing data center footprint cannot service in the near term. Building additional proprietary data center capacity takes 18–36 months from land acquisition to operational readiness — too slow to capitalize on the AI demand wave that is happening now.

By paying SpaceX $920 million per month for access to 110,000 existing, operational Nvidia GPUs — hardware that SpaceX/xAI already owns and has already deployed — Google can immediately expand its AI computing capacity, fulfill more of its $460 billion backlog, and generate cloud revenue that more than covers the $920 million monthly cost. The deal is not a sign that Google lacks AI capabilities — it is a sign that AI demand has grown faster than anyone's ability to build physical infrastructure to service it, and that paying a premium for immediately available capacity is more value-creating than waiting 18 months to build it yourself.