Google Launches First AI Chips Into Orbit In Bid To Build Data Centers In Space

by
Alisa Davidson

Printed: September 25, 2026 at 9:17 am Up to date: September 25, 2026 at 9:17 am

by Anastasiia O

Edited and fact-checked:
September 25, 2026 at 9:17 am

To enhance your local-language expertise, generally we make use of an auto-translation plugin. Please observe auto-translation might not be correct, so learn original article for exact data.

Expertise firm Google will launch its first synthetic intelligence {hardware} into house subsequent week, marking the opening step in an bold effort to find out whether or not low Earth orbit may someday host large-scale AI computing infrastructure. The experimental satellite tv for pc, named MVP, is scheduled to fly aboard SpaceX’s Transporter-18 rideshare mission from Vandenberg Area Drive Base in California, developed in partnership with satellite tv for pc operator Planet Labs.

The MVP satellite tv for pc carries 4 of Google’s Trillium tensor processing models, the specialised AI chips that energy its Gemini fashions, along with roughly one kilowatt of photo voltaic panel capability — corresponding to the output wanted to run a single knowledge heart server. Its objective is strictly experimental: to collect knowledge on how the chips stand up to launch forces, radiation publicity, and the thermal extremes of the house surroundings.

The engineering challenges are substantial. Throughout ascent, particular person parts can expertise forces of fifty to 100 occasions Earth’s gravity, which the workforce replicated by three-axis vibration testing on the bottom. Radiation poses a unique risk — photo voltaic occasions and cosmic rays could cause bit flips that corrupt knowledge — so Google subjected its chips to proton beam assessments at UC Davis’s Crocker Nuclear Laboratory whereas operating stay AI workloads. Preliminary outcomes had been promising: the {hardware} tolerated a complete ionizing dose exceeding what a five-year house mission would ship, and most radiation-induced errors could possibly be cleared by restarting the chips.

Cooling stays essentially the most troublesome downside. Within the vacuum of house, there is no such thing as a airflow, and warmth can solely be dissipated by radiators. Google’s answer combines warmth pipes, thermal interface supplies, and an aluminum-copper radiator stack. In orbit, the chips can function for roughly quarter-hour earlier than requiring a cooldown cycle, throughout which they’ll course of brief queries from house.

Why Orbital Knowledge Facilities Matter for the AI Trade

The strategic logic behind Challenge Suncatcher displays mounting stress on terrestrial knowledge facilities. AI-driven compute demand has triggered neighborhood opposition over noise, grid pressure, and rising electrical energy prices in a number of areas. A satellite tv for pc in the proper orbit receives as much as eight occasions extra photo voltaic vitality than equal capability on Earth, probably providing near-constant, clear energy with out competing for terrestrial grid capability.

Google shouldn’t be alone in pursuing this imaginative and prescient. SpaceX, Blue Origin, and startup Starcloud are all engaged on orbital compute ideas, with some proposals envisioning constellations of hundreds and even hundreds of thousands of interconnected satellites. Skeptics, together with OpenAI chief government Sam Altman, doubt that orbital computing can obtain business scale this decade, citing launch prices, satellite tv for pc manufacturing bottlenecks, and unresolved engineering issues.

Google’s roadmap is intentionally incremental. A two-satellite check in 2027 will consider the high-bandwidth laser hyperlinks required to synchronize computing throughout satellite tv for pc clusters — a precision problem in comparison with hitting a coin-sized goal from miles away whereas each factors are in movement. Longer-term designs envision fleets of greater than 80 satellites flying in formation, every carrying dozens of TPUs. The corporate estimates that, as launch prices decline, orbital knowledge facilities may attain price parity with terrestrial services someday within the mid-2030s.

For the AI trade, the launch alerts a shift towards treating vitality and geography as solvable constraints quite than mounted ones. Whereas a working orbital knowledge heart stays years away, a profitable mission would validate the foundational premise that AI {hardware} can function reliably outdoors Earth’s ambiance — opening a genuinely new frontier for compute infrastructure at a second when terrestrial growth is assembly its limits.

Alisa, a devoted journalist on the MPost, focuses on crypto, AI, investments, and the expansive realm of Web3. With a eager eye for rising traits and applied sciences, she delivers complete protection to tell and interact readers within the ever-evolving panorama of digital finance.

• 
  
  
  

More articles