Realtime AI News
Google's First Data Center Satellite Is Online, and the TPU in Orbit Writes Poetry
TechCrunch reports that Google's first data center satellite is now running, with the first TPU in space hosting a Google DeepMind Gemma 4 model that has been producing deliberately cheesy poetry. The demo matters less for its verse than for showing that inference can run end to end in orbit.

Google's first data center satellite is online, according to TechCrunch, and the payload includes the first TPU ever flown in space. It is running a Google DeepMind Gemma 4 model, which the report says has been writing poetry of the cheesy variety.
The task is deliberately modest. Nobody needs verse generation in orbit; the point is to show the whole chain working — model loaded, inference executed, output returned — on hardware that has left the planet.
The motive for putting data centers in the sky is not raw compute. Terrestrial facilities are increasingly constrained by power supply, cooling and planning approval, while orbit offers steady sunlight during daylight passes without competing for a grid connection.
Orbit imposes its own limits. Heat has to be shed by radiation in vacuum, maintenance is effectively impossible, hardware must survive launch vibration and radiation, and the link between satellite and ground is far thinner than the interconnect inside a machine room.
That makes this a validation milestone rather than a capacity story. One satellite, one comparatively small model, one short piece of text generation — still a long way from carrying production traffic.
Choosing an open-weight model like Gemma 4 fits the constraints: manageable size, deployable locally, sensible for a resource-limited platform. Proving feasibility with a frontier model would have multiplied the engineering variables.
For the industry, it is an aggressive answer to the question of where compute will find power. Cloud providers are competing for grid connections and land; orbit sidesteps some of that, at the price of launch costs and unrecoverable hardware.
Three things are worth watching: sustained stability of in-orbit inference, throughput on the space-to-ground link, and whether the economics work per unit of compute. The poetry is just the opening act.
Why it matters
This is the first operational check of moving AI inference off the terrestrial grid, aimed at the long-run tension between compute growth and the supply of power and land. It will not replace ground data centers soon, but it changes how orbital infrastructure gets discussed.
Nearby Updates
All10/31, 01:00
TestMu AI launches Kane CLI Exploration to turn a single URL into tests
TestMu AI has launched Kane CLI Exploration, an AI agent that according to 01net maps an entire web application from one URL and converts it into tests. The product folds site exploration and test generation into a single command-line workflow, targeting the setup work that consumes most of a QA team's time.
10/31, 01:47
Zest's AI Concierge Picks Restaurants and Books the Table Over Text
Zest has launched an AI concierge agent that works inside a text conversation, gathering a group's tastes, putting restaurant options to a vote and then making the reservation. TechCrunch frames it as a fix for the familiar deadlock over where to eat.
10/31, 01:50
US Justice Department Said to Question Nvidia's $12.9B Hugging Face Deal
The US Department of Justice has asked questions about Nvidia's $12.9 billion acquisition of Hugging Face, according to a report attributed to Pasquale Pillitteri. The report does not say what stage those questions have reached, but it puts the deal on antitrust authorities' radar.
10/30, 23:53
DeepBlue Technology Scraps Robot Projects, Dissolves VLA Team in Pivot to AI Agent Work Phones
DeepBlue Technology has scrapped its existing robot projects and dissolved its VLA team, redirecting resources toward AI agent work phones, according to a report on mrjjxw.com. The report also flags that becoming self-funding remains the open question for the company.