Google has deepened its partnership with SpaceX through a significant infrastructure deal that will see the search giant pay the rocket company $920 million each month for access to computing power spanning nearly three years. The agreement represents one of the most substantial commitments to date by a major technology firm for SpaceX’s growing data centre operations, underscoring how aggressively Elon Musk’s enterprise is pivoting toward artificial intelligence infrastructure as it prepares for an expected initial public offering.
Google backs SpaceX infrastructure to power enterprise AI platform
A regulatory filing submitted to the on Friday revealed the scope of the arrangement, which commences in October 2026 and extends through June 2029.
According to the disclosure, Google will acquire computing resources encompassing “approximately 110,000 NVIDIA GPUs, CPUs, memory, and other related components” to be made available from SpaceX’s data centre portfolio.
The contract includes provisions allowing either party to exit the arrangement after a 90-day notice period beginning in 2027, providing flexibility for both organisations as their infrastructure needs evolve.
Bridging capacity for Gemini Enterprise deployment
Google indicated that the infrastructure procurement was essential for meeting unexpected growth in customer adoption of its enterprise-focused AI offering. In a statement to Business Insider, a Google Cloud representative elaborated on the decision-making process behind the deal:
“are long-time partners. This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected.”
The characterisation of the deal as temporary suggests Google views the arrangement as a means of addressing immediate demand whilst it expands its own data centre infrastructure. The emphasis on “bridge capacity” indicates the company expects to reduce or eliminate its dependence on SpaceX’s systems once internal capacity is scaled.
SpaceX monetises spare compute ahead of IPO
The secured with Anthropic, details of which surfaced in the company’s S-1 filing last month. Anthropic has committed to paying SpaceX $1.25 billion monthly through May 2029 in exchange for access to processing capacity from the firm’s Colossus data centres. On an annual basis, the Anthropic deal represents approximately $15 billion in committed revenue per year, making it a cornerstone of SpaceX’s infrastructure monetisation strategy.
Both agreements reflect what SpaceX characterised in its regulatory filing as a deliberate business model: “This structure allows us to monetise unused compute capacity in our infrastructure, while still permitting reallocation of the capacity for our own internal initiatives if needed in the future.”
Key compute partnerships and timeline comparison
Tom Brown, the AI laboratory’s compute chief, disclosed in May that the Colossus infrastructure would be deployed for inference operations—the computational work required for AI models to generate outputs based on user inputs.
Mounting infrastructure costs strain SpaceX’s capital plans
The capital outlay required to acquire sufficient graphics processing units for these commitments has begun to exert significant pressure on SpaceX’s finances.
According to the company’s S-1 filing, artificial intelligence-related operational losses increased fourfold in the previous fiscal year to exceed $6 billion, primarily driven by escalating cloud computing expenses and accelerated depreciation of GPU hardware. During the first three months of the current year, these losses more than doubled to nearly $2.5 billion.
The magnitude of GPU acquisition costs has prompted SpaceX to consider manufacturing its own processors—a move that would position the company as a direct competitor to Nvidia, which maintains overwhelming dominance in the market for high-end graphics processing units.
within its disclosure of “substantial capital expenditures” the organisation intends to undertake.
Partnership complexities and emerging competition
The arrangement with Google introduces competitive tensions into what has long been a partnership between the two companies. SpaceX operates Starlink, its satellite internet division, which collaborates with Google Cloud on infrastructure initiatives.
The two organisations announced a partnership in 2021 under which SpaceX installed Starlink ground stations within Google data centres to facilitate connectivity.
By selling hundreds of megawatts of computing capacity to Anthropic and now Google, SpaceX has begun to function as a competitor to Google Cloud itself. The dynamics grow more complex given that Google is engaged in discussions with SpaceX to assist with the development of orbital data centres—a venture that could position Google as both customer and collaborator.
SpaceX positions itself within trillion-dollar AI infrastructure market
as a significant emerging market opportunity. The company’s S-1 filing estimates that the total addressable market for artificial intelligence compute, based on present demand patterns and existing GPU rental pricing, could reach approximately $2.4 trillion (roughly ₹20 trillion).
The Google and Anthropic agreements represent the opening salvo in what SpaceX clearly intends as a much broader assault on the compute infrastructure sector. As the company prepares for what is anticipated to be a record-breaking public listing, these deals provide both immediate revenue and proof of concept that major technology firms are willing to pay premium rates for computing capacity sourced from outside the traditional cloud infrastructure providers.