What will the combined annualized (run-rate) revenue (USD billions) of SpaceX's AI segment — xAI/Grok plus Cursor/Anysphere — be as of December 31, 2026?

The median estimate of $61.75B is anchored by a bottom-up aggregation of known compute-leasing contracts, the recently acquired Cursor, and legacy X/Grok revenue. The central 50% range runs from $52B to $70B, while the outer range extends from $40B at the 10th percentile to $78B at the 90th percentile. This baseline is balanced against execution risks in data center rollouts and the structural ambiguity of how the annualized run-rate will be calculated—specifically, whether it relies on a December revenue multiple or a full Q4 multiple.

Four major cloud services agreements provide the bulk of the segment's revenue. If fully billing by December, these include Anthropic at $1.25B per month finance.yahoo.com, Google at $920M per month finance.yahoo.com, an unnamed six-month deal announced in August at roughly $1.12B per month s21.q4cdn.com, and a new hosting deal starting December 1 at $1.11B per month finance.yahoo.com. Together with smaller deals like Reflection AI, contracted compute capacity represents roughly $53B to $55B in annualized revenue.

The remainder of the AI segment consists of Cursor and legacy operations. Cursor reportedly generated roughly $4B in annualized revenue prior to its acquisition and is projected to reach $6–8B by year-end 2 sources. Legacy X advertising, Grok, and subscriptions are expected to contribute an additional $3–5B. Summing these non-compute components with the fully realized cloud service agreements lands the bottom-up run-rate in the low-to-mid $60B range.

Management's public targets drive the upside estimates, while the company's massive compute-infrastructure scaleout supports a wider upper tail if capacity comes online faster than the baseline assumes. The CFO recently reiterated that SpaceX is on track to hit a $100B total annualized run-rate across the entire company by year-end finance.yahoo.com, a trajectory Elon Musk characterized as what would happen "if we basically did nothing" s21.q4cdn.com. Assuming Starlink and the Space segment contribute $25–27B, the AI segment would need to generate $73–75B to close the gap. Reaching this upper bound likely requires additional unannounced deals or highly accelerated ramps, supported by the company's efforts to raise $40B in debt for Nvidia chips finance.yahoo.com. Broader expectations for the company's full-year consolidated growth and the potential for upside scaling justify extending the 90th-percentile outcome to $78B, above the $70B 75th-percentile estimate.

Conversely, the downside is defined by execution hurdles and accounting mechanics, with the scale of the infrastructure expansion creating potential bottlenecks severe enough to widen the lower tail. Reports of data-center operational issues and potential build-out slowdowns 2 sources suggest that newly contracted capacity may not be fully live or billing without ramp credits by December. Additionally, if the run-rate is resolved by multiplying total Q4 revenue by four rather than annualizing December's exit rate, the back-weighted timeline of these new contracts will drag the annualized figure down into the low-to-mid $50B range 2 sources. That supports a 25th-percentile estimate of $52B, while a combination of more substantial execution bottlenecks, delayed billing, and unfavorable annualization mechanics produces a 10th-percentile outcome of $40B.