SpaceX has developed a unique AI business model, generating billions by renting its advanced computing power to competitors of its own AI chatbot, Grok, including Google and Anthropic. This ‘landlord’ strategy has propelled its AI unit to significant growth, though the company faces risks related to high capital expenditures and the potential for its key customers to eventually build their own infrastructure. The upcoming quarterly report will be crucial in validating the long-term viability of this high-stakes venture.
SpaceX, the aerospace giant, is navigating a fascinating duality in the Artificial Intelligence landscape. While its homegrown chatbot, Grok, struggles to gain significant traction against industry leaders, the company has quietly built a booming AI infrastructure business. This strategic pivot involves renting its substantial computing power to the very rivals of Grok, a move that is rapidly generating billions.
A recent Momentic Marketing report, leveraging May 2026 Similarweb data, placed Grok fifth out of seven major AI assistants by worldwide web visits, trailing powerhouses like ChatGPT, Gemini, Claude, and DeepSeek. The report also indicated a decline in Grok's global market share over the last quarter. Yet, even as Grok strives for competitive footing, SpaceX is profiting handsomely by serving as the 'landlord' for its competitors.
As Gwynne Shotwell, SpaceX's President and COO, stated in the August 4 earnings call, "We’re rapidly expanding our compute capacity to meet our own needs as Grok expands, as well as those for other leading companies such as Google and Anthropic." This highlights a pragmatic approach: monetize cutting-edge infrastructure regardless of whose AI ultimately dominates the market.
As of this morning, shares of SpaceX (NASDAQ:SPCX) were trading at $164.47, experiencing a slight dip of 1.87% on the day but boasting an impressive 11.17% gain over the past month.
SpaceX's AI Revenue Streams
SpaceX's AI unit primarily encompasses cloud compute services, Grok development, and the underlying X platform. This segment reported a staggering $2.56 billion in Q2 revenue, a 247% increase year-over-year. The lion's share, $2.194 billion, came from AI Solutions & Infrastructure, while advertising contributed $367 million. For context, Starlink’s Connectivity unit led with $4.29 billion, and Launch services generated $962 million.
The company has secured substantial Cloud Services Agreements totaling $14.1 billion in signed sales, with the initial phase already contributing $1.6 billion in incremental AI infrastructure revenue during the quarter.
The "Landlord" Advantage in AI Compute
The essence of SpaceX's strategy is simple yet powerful: much like a landlord, it collects rent regardless of which tenant ultimately thrives. Elon Musk himself indicated that Grok is projected to utilize only a minority of SpaceX’s compute resources, stating, “I would expect maybe 10% of our compute to be used for Grok training, something like that.”
Management has emphasized the economic efficiency of this model, claiming a payback period of less than one year on new capital deployments for compute infrastructure. This robust performance helped the AI segment achieve a positive adjusted EBITDA of $1.1 billion for the first time.
SpaceX isn't alone in adopting this "GPU landlord" model. Giants like Oracle (NYSE:ORCL) and CoreWeave (NASDAQ:CRWV) also operate similar businesses. Oracle’s significant backlog is largely attributed to OpenAI, while CoreWeave’s revenue has historically stemmed predominantly from Microsoft (NASDAQ:MSFT) and OpenAI, as noted by Hyperframe Research.
Risks: High Capital Costs and Client Concentration
Despite its impressive growth, this strategy is not without significant risks. SpaceX's capital expenditures soared in Q2, reaching $18.37 billion, with a substantial $15.83 billion allocated specifically to AI computing power. Management anticipates similar levels of CapEx for the next two quarters, indicating a massive ongoing investment.
The AI unit posted a $1.3 billion operating loss, contributing to the company's overall $541 million loss. A primary concern is the concentrated nature of its client list. Alphabet (NASDAQ:GOOGL), which owns Google, operates its own extensive data centers. The critical question remains: if Google and Anthropic build out sufficient internal capacity, will they cease leasing from SpaceX?
Elon Musk's comment that SpaceX expects a very small percentage of its chip supplier’s GPUs next year could signal a future shift in its own infrastructure procurement. With the stock trading at a forward P/E of 200, there is minimal room for any disappointment or faltering in this high-growth narrative.
Upcoming Report to Test the Thesis
The next quarterly report will be a crucial litmus test for the enduring strength of the "landlord" thesis. Key indicators to watch include whether management has signed an additional $6.7 billion in cloud services revenue over the six months commencing in October. Furthermore, observers will look for AI Solutions & Infrastructure revenue to exceed $2.194 billion, Cloud Services Agreements to surpass $14.1 billion, and compute capacity to rise above 1.4 GW, on track for over 2 gigawatts by year-end.
Should all these targets be met, the landlord strategy will be validated. However, if infrastructure revenue stalls while AI capital expenditures remain near $15.8 billion quarterly, the current model's viability could be significantly challenged.
