SpaceX is rapidly transforming AI computing services into its next major growth pillar.

At the Goldman Sachs Communacopia + Technology Conference 2026 on Thursday, September 10, SpaceX Chief Financial Officer Bret Johnsen revealed that the company signed another AI compute hosting agreement earlier this month, which is set to generate approximately $1.1 billion in monthly revenue starting December 1 — translating to an annualized figure of roughly $13.3 billion.

Johnsen did not disclose the client's name but expressed confidence that, based on December revenue annualized, SpaceX would hit its target of $100 billion in annualized revenue (ARR) by the end of this year.

According to Johnsen, SpaceX's core advantage does not lie in any single business segment but rather in its vertical integration spanning rockets, satellites, communication networks, and AI computing power and models — using Starship to reduce the cost of space access, generating cash flow through terrestrial AI compute, and eventually extending that compute capability into orbit.

Massive compute deals are landing in quick succession, and this latest contract is hardly SpaceX's first major AI computing agreement in recent months. In July, the company inked a six-month deal worth $6.7 billion for compute services, with the client remaining unnamed, though some analysts speculate it could be linked to the U.S. Department of Defense.

Beyond that, SpaceX has signed multi-year compute agreements with Anthropic, Alphabet Inc (NASDAQ: GOOGL), and Reflection AI. Under these deals, Anthropic pays SpaceX around $1.25 billion per month, running through May 2029, while Google will begin paying roughly $920 million monthly starting this October, continuing until June 2029. Google previously disclosed that the agreement involves approximately 110,000 Nvidia GPUs.

However, Johnsen also cautioned that most of SpaceX's compute contracts do not lock in revenue entirely, as nearly all of them allow both parties to terminate early after a few months' notice. The reason lies in SpaceX's own AI products requiring substantial compute capacity, and the company does not want to constrain its own business expansion by committing too much capacity to external clients over the long term.

Turning to orbital computing, Johnsen stated that SpaceX plans to launch its first batch of orbital computing satellites next year, with ambitions to scale up around 2028.

He believes that as Starship's reusability continues to improve, declining launch costs could bring about a crossover point around next year where orbital compute becomes economically comparable to terrestrial data centers. While ground-based facilities face mounting pressure from power, cooling, and land costs, SpaceX is betting that falling costs for rockets, satellites, and solar systems will drive the cost curve for space-based computing steadily downward.

In parallel, Starship itself is about to enter its commercial phase.

Johnsen revealed that the 14th test flight, expected later this month, will be the first Starship mission to generate revenue, carrying production-grade V3 Starlink satellites.

SpaceX aims to recover both the first and second stages of Starship by late this year; if all goes well, the company hopes to further enhance two-stage reusability next year. Through high-frequency, low-cost launches, SpaceX intends to build shared infrastructure for Starlink, AI computing, and future orbital compute operations.

On the ground-compute front, SpaceX expects to exceed 2 gigawatts (GW) of deployed capacity by year-end, expanding further to between 5 GW and 10 GW in the coming year. Johnsen said the company is even more confident in its $100 billion ARR objective, with AI computing, Starlink, and future orbital compute collectively powering SpaceX's next growth trajectory.

Johnsen also noted that the next-generation direct-to-cell satellites are scheduled to begin launching in 2027, targeting service initiation in the first half of 2028, and aiming to upgrade from the current SMS and lightweight voice capabilities to "full 5G quality" services.