Argus Research upgraded SpaceX to Buy from Hold and set a $160 price target after the company's earnings call, saying its heavy AI infrastructure investments position it as a winner in the sector.
What Happened
SpaceX spent nearly $16 billion on AI infrastructure in Q2 alone. CFO Bret Johnsen said during the earnings call that SpaceX could pay back anything on new deployments within a year by monetizing high-demand GPU compute capacity faster than traditional data center economics would suggest. The company has built Colossus and Colossus II primarily for its own model training, but much of that capacity is already rented out to third parties including Anthropic, Google, and Reflection AI. These partnerships are adding billions per month to SpaceX's spreadsheet, according to Argus Research. Compute capacity scaled from roughly 0.4 GW annually at the start of Q2 to 1.4 GW by quarter's end, with targets exceeding 2 GW by year-end.
Why It Matters
The $160 price target and upgrade reflect Wall Street's growing recognition that SpaceX is converting its massive AI capital outlays into recurring revenue through a hyperscale compute rental model serving other leading AI companies. The firm noted that new compute capital behaves more like cost-of-goods-sold than traditional multi-year capex given rapid monetization patterns, with many contracts generating revenue within months of capacity coming online. Vertical integration advantages in power and cooling infrastructure combined with high incremental margins on rental once capacity is deployed could accelerate cash recovery compared to conventional data center projects.
The Bottom Line
AI-segment revenue rose to roughly $2.6 billion in Q2, driven primarily by infrastructure contracts but also software streams including Grok subscriptions, X ads and enterprise APIs. SpaceX's strategy of leasing excess compute capacity while building its own AI applications could shorten payback timelines on the heavy infrastructure buildout.