Remember when "cloud computing" was just a metaphor?
Well, Google is taking it rather literally.
On October 30, 2026, Google announced that its AI model Gemma 4 had successfully generated a response from orbit and transmitted it back to Earth.
And the first response?
A haiku.
Yes. We've reached the point where artificial intelligence is writing poetry in space.
Meet Project Suncatcher, Google's ambitious experiment to explore whether future AI data centers could operate in orbit, powered by solar energy.
Here's what happened:
October 1, 2026: Google launched its prototype satellite aboard a SpaceX rocket.
Onboard AI hardware: The satellite carries Google's data-center-class Tensor Processing Unit (TPU).
Onboard model: Google's open-weight Gemma 4.
First successful response: Engineers sent a prompt from Earth asking the model to write a haiku about being in space.
The result: Gemma processed the request in orbit and transmitted its poem back as the satellite passed over Svalbard, Norway.
The poem:
Blue marble below
Data streams in silent grace
Starlight feeds my core.
Google CEO Sundar Pichai called it a "Hello World" moment for Project Suncatcher.
But Google isn't alone in this race.
In November 2025, Starcloud launched an NVIDIA H100 GPU into orbit, later demonstrating AI training and inference in space.
Now Google has achieved its own milestone using a TPU rather than an NVIDIA GPU.
And here's where things get really interesting.
Google is exploring a future in which networks of solar-powered satellites could provide large-scale AI computing infrastructure.
Imagine orbital data centers connected by laser links, operating above Earth's atmosphere.
Of course, there are enormous engineering challenges:
Cooling powerful processors in a vacuum
Protecting chips from radiation
Maintaining high-bandwidth satellite connections
Launching, maintaining, and eventually replacing orbital hardware
Making the entire system economically viable
So no, Google hasn't built a fully operational space data center yet.
But it has demonstrated that an advanced AI model can respond to a prompt using data-center-class hardware in orbit.
Next milestone: two interconnected experimental satellites in 2027.
And if Starship eventually makes launching large quantities of computing hardware economically practical, things could get very interesting.
I suppose one day I'll tell my grandchildren:
"You know, we called it cloud computing long before the cloud was actually above our heads."
And they'll probably think I'm making it up.
Google's official announcement:
https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-satellite-response/
Project Suncatcher:
https://blog.google/innovation-and-ai/technology/research/google-project-suncatcher/
Starcloud's orbital AI experiment:
https://blacknightspacelabs.com/blog/starcloud-nvidia-h100-orbit-first-ai-training-space
#GoogleAI #ProjectSuncatcher #Gemma4 #SpaceComputing #SpaceX #AIDataCenters
YouTube / ABVX