SpaceX announces a groundbreaking commitment to Nvidia's Vera Rubin, targeting an ambitious 10 GW in compute capacity by 2027.
In a dramatic move during its inaugural earnings call as a public entity, SpaceX unveiled an unprecedented commitment to Nvidia's Vera Rubin architecture. This strategy marks a significant shift, aiming to establish the company as a leading player in stablecoin-reserve-income-and-competition/">partnership-boosts-compute-capabilities/">AI infrastructure.
The aerospace behemoth has ambitious plans to build its compute capacity, targeting an installation of 2 gigawatts (GW) by the end of 2026. This figure is projected to scale rapidly to approximately 10 GW by the conclusion of 2027, with a tentative goal of achieving 20 GW, although CEO Elon Musk suggested that around 15 GW would be a more realistic expectation. This initiative represents the most substantial single compute investment ever launched within the realm of artificial intelligence.
SpaceX's robust financial strategy underpins this massive hardware deployment. For the second quarter of 2026, the company reported impressive revenues of $7.8 billion, a staggering 92% increase year-over-year, comfortably surpassing analysts' predictions of $6.93 billion. However, it's important to note that the firm’s AI division recorded an operating loss of $1.26 billion. Despite this, the revenue model is clear: a striking 95% of AI earnings stem from GPU rental agreements.
The company has landed significant contracts with major clients, including Google, which pays around $920 million monthly for the use of approximately 110,000 GPUs. Anthropic also stands out, contributing roughly $1.25 billion monthly via its partnership at the Colossus 1 facility in Memphis. As noted by CFO Bret Johnsen, SpaceX is successfully monetizing its available compute capacity, leading to considerable incremental EBITDA margins.
The strategic decision to adopt the Vera Rubin architecture reflects a calculated bet on both density and performance. Musk characterized the Vera Rubin system as “the best AI computer” available, signifying a forward-looking move to retain a competitive edge in the increasingly crowded AI landscape. Analysts like Ben Reitzes from Melius Research estimate that even the initial 2 GW deployment could potentially create an additional $100 billion in revenue for Nvidia, pushing the semiconductors giant closer to a $1 trillion market valuation.
This development is not mere speculation; SpaceX has already procured $6.7 billion in forthcoming cloud services contracts beginning in October 2026, indicating substantial demand for its compute capabilities.
SpaceX is not stopping at terrestrial data centers; the company is venturing into orbital computing as well. The upcoming Starmind AI1 satellite, outfitted with the Nvidia Space-1 Vera Rubin module, promises to deliver 25 times more AI compute power per GPU compared to its predecessors, such as the H100. An FCC filing reveals plans for up to 1 million satellites in sun-synchronous low Earth orbit, addressing latency and memory challenges typical of ground-based clusters. This innovative approach positions SpaceX to effectively compete with its own clientele, including its internal Grok 4.5 AI development.
The compute landlord thesis, previously scrutinized in relation to Nvidia's hefty $600 billion exposure to OpenAI, is now experiencing a significant transformation. By acquiring xAI in February 2026 and transitioning to a public company in June with an impressive $85 billion capital infusion and a staggering $1.75 trillion valuation, SpaceX is institutionalizing the landlord model within the AI sphere. The company is no longer merely constructing rockets; it is now establishing foundational infrastructure critical for the next generation of artificial intelligence.
Musk's assertion that achieving $100 billion in annual recurring revenue by December is a baseline expectation—even if no additional initiatives are undertaken—demonstrates substantial confidence in this capital expenditure. With capital expenditure (capex) reaching $28 billion for the first half of 2026, a stark rise from $7 billion a year earlier, SpaceX is investing significantly to reinforce its position as a preeminent host for powerful AI models.
The company's recent initial public offering was strategic, not just a liquidity event. It served as a vital mechanism to fund a colossal hardware pipeline worth hundreds of billions. The compute landlord thesis has undeniably transitioned from a mere $10 billion experiment into a substantial reality poised to shape the future of artificial intelligence.
The monumental pivot towards AI infrastructure signals an optimistic future for the development of AI models, frontier labs, and training methodologies. As competition among labs intensifies, and model capabilities continue to evolve, the landscape of AI research is set for substantial transformation.
SpaceX's gambit is not just a transition towards enhanced compute capabilities; it represents a larger vision for an AI-driven future. With ongoing investments, innovative strategies, and a commitment to expand compute infrastructure, the implications of this bet on Nvidia's Vera Rubin architecture will reshape the dynamics of the AI industry.
The developments at SpaceX and its exclusive partnership with Nvidia are expected to send ripples throughout the AI community and beyond. The move reinforces the idea of compute capacity as a pivotal factor driving advancements in artificial intelligence.
As SpaceX embarks on this ambitious journey, the expectation is that competitors will be compelled to rethink their own strategies regarding AI infrastructure. The convergence of aerospace and AI is taking center stage, and the implications could very well steer the future trajectory of both industries.