Google Just Sent Its AI Chips Into Space — Now Comes The Hard Part

• Google’s TPUs reached orbit aboard SpaceX’s Transporter-18 mission, marking the chips’ first trip into space.
• Project Suncatcher is testing the chips against radiation, extreme heat and launch conditions.
• GOOGL stock declined the last two days, even as the retail sentiment on Stocktwits moved higher to ‘bullish.’

Alphabet, Inc. (GOOGL) has put its Tensor Processing Units (TPUs) into orbit for the first time, testing whether Google’s custom chips can survive the harsh space environment for a range of applications.

Google launched the first prototype for Project Suncatcher, its moonshot effort to determine whether large-scale AI computing could eventually be moved into low-Earth orbit.

The prototype launched aboard SpaceX’s Transporter-18 rideshare mission on Thursday, carrying the TPUs into orbit. SpaceX confirmed the mission carried 130 payloads to low-Earth orbit. 

The satellite was built in partnership with Planet Labs PBC (PL), which Google enlisted to develop the spacecraft platform. Planet Labs said when it announced the partnership last year that the project was designed to test Google’s TPUs in space and eventually demonstrate high-bandwidth communication between satellites.

The current mission is not an orbital data center. Google’s prototype is primarily designed to determine whether its AI chips can withstand launch forces, radiation and the unusual thermal environment of space.

Google Has Three Big Problems To Solve

Radiation, cooling and networking remain key challenges for Google’s space-based AI ambitions. 

The company said tests of its Trillium TPUs using a proton beam showed the chips could withstand radiation beyond what they would face during a five-year mission, although the attached high-bandwidth memory was more sensitive. 

With no airflow in space, Google is also testing heat pipes and radiators to move heat away from the TPUs, while its longer-term vision calls for satellite clusters linked by high-bandwidth optical connections.

Planet Lab said on Thursday that it successfully made initial contact with Tanager-2, the Project Suncatcher prototype, and two SuperDoves and began the commissioning process. 

Planet expects to make contact with the remaining 16 SuperDoves that will deploy on the D-Orbit ION orbital transfer vehicle on schedule. This marks Planet’s 40th successful launch and a total of 718 satellites built and delivered on orbit.

Why Put AI In Space?

Google said satellites in the right low-Earth orbits could receive up to eight times more solar power than solar panels on Earth, while also having access to sunlight for much longer periods. That could potentially address one of the biggest constraints facing AI data centers: power.

But the economics still depend heavily on launch costs. Google’s research previously estimated that if launch costs eventually fall to around $200 per kilogram, the economics of powering computing infrastructure in space could begin to approach those of terrestrial data centers.

“1kw of TPUs per sat. Godspeed to Google. TPU inference on orbit make total sense. Now they just need to scale that up 100x power,” said Aaron Burnett, CEO of space-tech research firm Mach33 Financial Group.

Google’s Gemini 4 Argon

Alphabet unveiled Gemini 4 Argon on Wednesday, a new frontier AI model that Morningstar said “leads or ties” competing models from OpenAI and Anthropic across a range of industry benchmarks, and analysts’ reactions are pouring in.

Morningstar said the launch is particularly significant because Alphabet appeared to have lost ground at the AI frontier. Its previous flagship, Gemini 3.1 Pro, had last been considered close to the frontier “almost 10 months ago,” raising questions among investors about Google’s ability to consistently produce top-tier models.

The biggest improvement appears to be agentic coding, with long-horizon software-engineering benchmarks showing a “clear upgrade” over Gemini 3.8 Flash and bringing it broadly in line with Anthropic and OpenAI models, Morningstar said.

Google is also pricing Gemini 4 “significantly lower” than rival frontier models, which Morningstar sees as a push to accelerate adoption.

GOOGL Stock Move, Retail View

GOOGL stock fell 4% on Wednesday and another 1% on Thursday, also pressured by a U.S. judge indicating that a class-action lawsuit alleging Google monopolized the digital advertising market could proceed to a jury trial. The plaintiffs are seeking $3.2 billion in damages.

On Stocktwits, retail sentiment for GOOGL was ‘bullish.’ The stock is up 8.3% year to date.

For updates and corrections, email newsroom[at]stocktwits[dot]com.