SpaceX, led by Elon Musk, made a spectacular debut on the U.S. Nasdaq market on Dec. 12. The concepts Musk had championed — lunar exploration, Mars exploration, low-orbit communication satellites, and space data centers — are no longer far-fetched stories. Space development has now become a topic of public discourse, one that stock investors and ordinary people around the world are considering together. The explosive interest in space is evident in the fact that Musk became the world's first "trillionaire" right after the SpaceX listing.
While SpaceX draws attention because of its title as a space exploration company, the major big tech firms leading the artificial intelligence (AI) industry have also already entered the space development business, directly or indirectly. Notable examples include Nvidia, the world's largest company by market capitalization; AMD, which competes with Nvidia for dominance in graphics processing units (GPUs); Meta, which is transitioning from a social media company to a hyperscaler; and Google, which is building an Alphabet (Google's parent company) kingdom with a full-stack strategy spanning from chips to products.
At "GTC 2026" this past March, Nvidia announced its latest accelerated computing platform that supports the implementation of AI computing for orbital data centers (ODC) and autonomous space operations. It provides data center-grade performance even amid constraints such as size, weight, and power in space, serving to connect AI functions to operate between ground and space, and within space. Nvidia CEO Jensen Huang stressed in March, "The era of space computing, humanity's final frontier, has arrived," and "As we deploy satellite constellations and expand exploration deeper into space, intelligence must exist wherever data is generated."
The platform's flagship feature is the Nvidia Space-1 Vera Rubin module. Vera Rubin is Nvidia's latest AI accelerator that combines the Vera central processing unit (CPU) and the Rubin GPU. The Rubin GPU mounted in the module delivers AI computing performance up to 25 times improved in space-based inference compared to the Nvidia H100 GPU. This module provides large-scale data center-grade AI performance in space and supports the operation of large language models (LLMs) and advanced foundation models.
Starcloud can be cited as a space development startup that utilizes Nvidia's accelerated computing platform. The company launched a satellite equipped with the Nvidia H100 in November last year, fully launching its orbital data center business. The Starcloud-1 satellite is the size of a small refrigerator and weighs about 60 kg, and is assessed as being able to provide GPU computing performance 100 times more powerful than existing satellites. Philip Johnston, CEO of Starcloud, said on the Nvidia blog, "In space, you can obtain nearly unlimited and inexpensive renewable energy," and "Compared to supplying power from Earth during the data center's operating period, carbon dioxide emissions can be reduced by up to 10 times."
AMD, which is chasing Nvidia, is also focusing on the space infrastructure business by developing AI chips and software used in building space data centers and in Mars exploration. AMD operates the "AMD Versal AI Edge" series, which can be applied to aerospace as well as autonomous driving, factory operations, healthcare systems, and the defense industry. The Versal AI Edge series has full radiation tolerance so that satellites can function properly, and is capable of ultra-high-speed processing and reconfiguration. In November last year, AMD announced that it had advanced packaging technology that can support geostationary orbit satellites, lunar exploration, and deep space probe operations in space for up to 15 years.
According to AMD's announcement, Blue Origin, which competes with SpaceX for leadership in space exploration, is a key AMD customer. Blue Origin uses the AMD Versal AI Edge Gen 2 adaptive SoC in its flight computer. This computer is scheduled to be mounted on the Mark 2, which will carry astronauts to the Moon in early 2028. NEC (Nippon Electric Company) also announced in March this year that it would build Japan's first optical communication satellite constellation using AMD technology. The plan is to demonstrate high-speed network routing in space using the AMD Versal adaptive SoC, thereby providing high-performance signal processing for data transmission within the satellite constellation and improving Earth connectivity.
Mark Papermaster, AMD chief technology officer (CTO), explained, "In the long term, orbit-based computing will be key," and "As AI computing demand in data centers continues to increase, various attempts are being made to implement large-scale computing in space, where solar energy and low temperatures can be utilized." He stressed, "Large-scale orbital computing is constrained by power, heat dissipation, radiation tolerance, and communication constraints. How to remove the heat generated in large-scale computing systems is one of the most difficult challenges." He added, "Because space is a vacuum, heat must be transferred through heat sinks. This reality demands a shift to a modular structure rather than a single box-type data center. AMD provides a modular design philosophy and adaptive, scalable computing building blocks such as CPUs, GPUs, FPGAs (reconfigurable semiconductors), and accelerators that can be optimized for the mission."
In April, Meta signed a contract with Overview Energy to be supplied with up to 1 gigawatt (GW) of space-based solar energy. One GW is equivalent to the generating capacity of one nuclear power plant. Overview Energy is developing a satellite that collects solar energy in space. Meta, which is transforming from a social media company into a hyperscaler, plans to draw solar energy and use it as power for ground data centers.
In November last year, Google unveiled "Project Suncatcher," aimed at expanding machine learning computing into space. The project involves launching satellites equipped with Google's own AI chip, the "Tensor Processing Unit (TPU)," into space orbit and collecting and utilizing solar energy. The Wall Street Journal (WSJ) reported last month that Google and SpaceX are discussing signing a rocket launch contract to build a space data center.
