SpaceX orbital data centers to scale in 2028 using Nvidia chips

Moving high-density AI compute into orbit addresses the massive bandwidth demands of agentic AI, which Cisco estimates generates 450% more traffic than human users. For the streaming and broader tech ecosystem, this infrastructure could bypass terrestrial bottlenecks by providing low-latency processing directly at the satellite level. The success of this deployment depends heavily on SpaceX achieving high-frequency Starship launches and successful 'tower catches' of the upper stage to lower deployment costs. Watch for the next Starship test flight results to determine if the 2027 launch timeline for the Vera Rubin systems remains viable.

Nvidia's involvement in SpaceX's orbital data center plans arrives amid an unprecedented wave of AI infrastructure deals that have drawn scrutiny from investors and analysts. The chipmaker is working on AI deals worth more than $750 billion, including a partnership with SK Group to do more than $500 billion in business, and is in talks to backstop as much as $250 billion to help OpenAI lease computing power from a US data center project. These arrangements have raised concerns about circular financing and artificially inflated demand across the AI compute supply chain, a dynamic that could influence how SpaceX's space-based compute offering is perceived by the market.

The terrestrial data center buildout that SpaceX's orbital approach aims to complement (or bypass) continues to accelerate at extraordinary cost. Meta Platforms and BlackRock plan to build a 1-gigawatt data center complex in Texas that will cost about $14 billion to develop, with the facility going online in 2028 and BlackRock funds holding an 80% interest in the joint venture. That same 2028 timeline aligns with SpaceX's target for significant orbital scale, suggesting that space-based compute is being positioned not as a replacement but as a parallel track to terrestrial hyperscale expansion. Tech firms including Alphabet, Amazon, Meta, and Microsoft have been aggressively building out data centers financed through a mix of capital expenditure, private infrastructure funds, debt, and sovereign wealth.

On the competitive chip front, Nvidia faces growing pressure from alternative architectures that could also serve space-constrained workloads. Cerebras filed for an IPO with a reported $10 billion contract with OpenAI forming a cornerstone of its growth narrative, and the company's wafer-scale engine design offers massive parallelism with lower latency compared to Nvidia's GPU approach. Meanwhile, Nvidia's Vera Rubin NVL72 platform selected for SpaceX's orbital deployment represents the company's next-generation architecture, and the partnership signals that space-based AI compute is being treated as a distinct deployment category requiring purpose-built thermal and power solutions rather than repurposed terrestrial hardware. The convergence of these trends suggests that by 2028, AI infrastructure buyers will face an increasingly fragmented landscape spanning terrestrial hyperscale, orbital compute, and competing chip architectures.