SpaceX (SPCX.US) advances first Starmind AI satellite launch to end of 2027, partners with Nvidia to accelerate development of orbital data centers

According to Zhitong Finance APP, Elon Musk is moving the AI race from Earth into space. On Monday, SpaceX (SPCX.US) CEO Musk announced on X that the company's first Starmind AI satellites equipped with NVIDIA (NVDA.US) chips will make their debut launch in Q4 2027 and are expected to reach a “significant scale” in 2028. This timeline is ahead of the previously disclosed 2028 plan in the prospectus.

On the same day, NVIDIA officially announced that SpaceXAI will deploy its next-generation Vera CPU to accelerate Agentic AI workloads and will send optimized Vera Rubin NVL72 rack-level systems into orbit as the compute core of the first-generation Starmind satellites. This collaboration marks a pivotal turning point as the world's first space AI data center project moves from concept to engineering implementation.

From "Ground AI Factory" to "Orbital Data Center": Vera Rubin’s Space Debut

NVIDIA Vera is the first CPU designed specifically for AI agents, featuring 88 NVIDIA-designed Olympus cores and LPDDR5X memory with bandwidth reaching 1.2TB/s. For agentic AI, reinforcement learning, and data processing workloads, task completion speeds can be up to 1.8 times faster than x86 CPUs.

NVIDIA’s Vice President of Hyperscale and HPC, Ian Buck, stated: “Agentic AI requires a new kind of computing system—not only to generate answers, but to take action. Vera gives AI agents the ability to act in real time—executing code, processing data, coordinating complex tasks. SpaceXAI is extending this architecture from ground-based AI factories to the next frontier of orbital computing.”

SpaceXAI President Mike Nicolls added: “Vera provides us with the CPU performance and memory bandwidth needed to handle large-scale coordination, scheduling, code, and data processing workloads, while letting the GPU focus on what it does best. This means higher-performing AI agents and more valuable work performed per watt of compute.”

Starmind: A "Space Supercomputer" of One Million Satellites

Starmind is a planned constellation project of orbital AI data centers by SpaceX. In January 2026, SpaceX applied to the US Federal Communications Commission (FCC) to launch up to one million computing satellites, aiming to expand AI compute capacity for the ground via low Earth orbit deployments.

Each Starmind satellite is essentially an optimized Vera Rubin NVL72 rack-level system equipped with NVIDIA Rubin GPU and Vera CPU. According to SpaceX’s prior disclosures, each satellite has a compute power of 120kW (peak 150kW), with hardware codename AI1. High-speed data transfer between satellites is realized via laser links, creating a distributed space AI supercomputer. The first batch of Starmind AI1 satellites will be based on an optimized rack-level Vera Rubin NVL72 system, marking NVIDIA’s accelerated computing platform’s debut in space.

Musk commented on X that these satellite-based data centers are “simpler, more cost-effective, denser, and lighter than traditional racks.” Compared to ground facilities, space deployments do not occupy large tracts of land and also have relatively less cooling pressure.

Compute Power Expansion: 2GW→10GW→20GW Aggressive Growth

Musk simultaneously outlined even more ambitious compute capacity goals: During an early August earnings call, Musk revealed that SpaceX plans to have over 2 gigawatts of AI compute by the end of 2026, and “closer to 10 gigawatts rather than 5 gigawatts” by the end of 2027. SpaceXAI will also expand ground AI infrastructure in Tennessee and Mississippi. The company is looking to accumulate compute capacity across multiple projects, with a target of reaching 20GW by the end of next year.

Musk admitted that some projects may not be delivered strictly on time, but expects about 15GW of online compute capacity by the end of 2027. He also revealed that SpaceX expects to secure a “very large share” of NVIDIA GPUs by 2027.

Trillion-Dollar CapEx and Debates on the "Non-Traditional" Path

The cost ledger for space data centers is under intense scrutiny from Wall Street. Wood Mackenzie estimates building a 1GW orbital data center would cost about $170 billion, more than three times that of conventional ground data centers. Morgan Stanley analyst Adam Jonas projects SpaceX’s AI-related capital expenditures will reach $53 billion in 2026, with another $130 billion added in 2027.

However, supporters argue that space data centers can avoid the regulatory barriers and community pushback faced by ground-based data centers. SpaceX is driving improved orbital computing economics by lowering launch costs through reusable Starship rockets. Musk previously explained that, prior to selecting NVIDIA systems, multiple chip options were assessed, and NVIDIA was deemed the best choice in the end.

From Exclusive Partnership to Deep Integration: SpaceX “All in” on NVIDIA

This collaboration is a comprehensive upgrade in the SpaceX-NVIDIA relationship. On August 4, during SpaceX’s first quarterly earnings call since going public, Musk announced the company had decided to exclusively use NVIDIA technology for its AI compute build-out. Musk explicitly stated at the time: “We have established close collaboration and partnership with NVIDIA at multiple levels, and we chose NVIDIA exclusively.”

SpaceX believes NVIDIA’s Vera Rubin architecture is “the best AI computer currently.” On August 4, the two companies announced a joint development of the Starmind AI1 satellite payload, with each satellite carrying NVIDIA’s Vera CPU and Rubin GPU. SpaceX has informed investors that its space server chips will be exclusively sourced from NVIDIA.

NVIDIA’s 13F filing on August 14 previously showed that the company holds about $21 billion worth of SpaceX shares. This deep integration is extending from the capital layer to the technical layer.

NVIDIA’s Wednesday Earnings Approaching: Space Narrative Becomes Latest Catalyst

The timing of this collaboration coincides with NVIDIA’s quarterly earnings announcement scheduled for Wednesday (August 26). The space AI data center story is adding a new dimension to NVIDIA's AI infrastructure narrative.

NVIDIA stated that the Vera Rubin platform integrates computing, networking, and software in a unified architecture, aiming to reduce per-token costs. With SpaceXAI advancing toward gigawatt-scale computing capacity, Vera Rubin provides a unified architectural foundation from ground AI factories to orbital computing.

In an orbital environment, constraints such as power consumption, cooling management, bandwidth, reliability, and physical integration are far more stringent than in traditional data centers. The collaboration between NVIDIA and SpaceX signals that its AI accelerated computing system is extending from Earth to low Earth orbit.