SpaceX has released its first post-IPO financial report, revealing that while rockets made the company famous, its AI business has quietly stolen the spotlight. Total quarterly revenue reached $7.814 billion, up 92% YoY and surpassing market expectations of approximately $6.9 billion. Notably, AI business revenue soared to $2.561 billion, representing a massive 247% YoY and over 213% QoQ increase, now accounting for nearly one-third of #SpaceX's total revenue.
However, this exponential growth requires heavy financial backing. Of the total $18.369 billion in capital expenditures in Q2, AI capital expenditures accounted for $15.828 billion—a staggering 86%. This aggressive upfront spending and ongoing losses caused SpaceX's stock to experience a "post-earnings drop," falling over 6% after hours. Meanwhile, short positions surged, with S3 Partners estimating short interest at around 32% of publicly tradeable shares.
A closer look at the financials reveals that the growth driver is not AI advertising (which actually decreased YoY) but "AI solutions and #infrastructure," which generated $2.194 billion this quarter. Newly signed cloud service agreements contributed $1.6 billion in fresh infrastructure revenue, with total non-cancelable backlog reaching $141 billion. SpaceX is rapidly transforming from an internal infrastructure built for training Grok into an AI cloud service provider renting compute to major clients like Anthropic and Google.
To sustain this massive supply capacity, SpaceX’s capital expenditure on infrastructure has been aggressive. Q2 Capex focused heavily on expanding the Colossus II giant AI supercomputer data center, mass-purchasing GPU servers, and building dedicated power and cooling facilities. As of June, SpaceX’s total operational power capacity reached 1.4GW—up from 1GW in Q1 and a threefold increase YoY. Additionally, the company spent $2.178 billion on R&D, primarily for training #Grok models, collaborating with coding tool Cursor, and developing proprietary scheduling and cloud orchestration software.
Although the AI division recorded an operating loss of $1.26 billion, this was better than worst-case analyst projections and proved that SpaceX's AI arm has substantial commercial monetization capabilities. During the earnings call, Elon Musk emphasized that SpaceX is committed to using NVIDIA GPUs, as they remain the best on the market, with a tentative target of reaching 20GW of power and cooling capacity by the end of next year.
[AgentUpdate Depth Analysis] SpaceX's scaling of gigawatt-class AI supercomputing infrastructure signals a profound transformation for the AI Agent ecosystem. As modern AI Agents transition from simple API call wrappers to complex, high-frequency, multi-agent orchestration systems, their demands for ultra-low latency and scalable runtime environments are soaring. SpaceX’s collaboration with Cursor and its proprietary workload scheduling software indicate that Musk is building a highly optimized hardware-software vertical stack. Compared to hyperscalers like AWS or Azure, SpaceX’s unique capability to secure massive self-sustained power grids could drastically lower the cost of continuous, long-running agentic workflows. Ultimately, this compute-heavy infrastructure will not only power Grok-driven agents but also serve as the foundational backbone for physical AI agents in robotics, aerospace, and spatial computing.
