Google Just Put AI in Orbit and the Results Are Honestly Jaw Dropping

-

TL;DR: Google just launched a satellite called Suncatcher that runs Gemma 4 AI models directly in orbit using custom TPU chips. The satellite processed its first real-time weather prediction in space this week, cutting latency from 40 seconds to under 3. This is either the beginning of edge AI in orbit or the most expensive demo Google has ever built.

I've been following the Suncatcher project since Google first hinted at it in early 2026, and honestly I didn't think they'd pull it off this fast.

But they did. Last week Google confirmed that Suncatcher, its experimental AI satellite, ran its first successful inference workload in orbit. Not on the ground. Not in a data center simulating orbit. Actually in space, 550 kilometers up, processing real atmospheric data with Gemma 4 running on radiation-hardened TPUs.

What Is the Suncatcher Satellite

Suncatcher is Google's first purpose-built AI inference satellite. It launched earlier this year on a SpaceX rideshare mission and has spent the past several months in testing mode. The satellite carries a custom variant of Google's TPU chips that have been hardened against cosmic radiation, which is the main reason why most satellites have historically avoided putting serious compute in orbit. Radiation flips bits. Lots of bits. A chip that works perfectly on the ground can become unreliable very quickly once it's exposed to the particle streams you get outside Earth's atmosphere. Google's engineers claim they've solved this with a combination of error-correcting memory and a redundant chip architecture that can detect and route around failures in real time.

Gemma 4 Is Running in Space Right Now

The specific model they're running is a compressed version of Gemma 4 optimized for the satellite's power envelope, which is about 200 watts total for the compute module. That's not a lot of headroom for a serious language model, but Gemma 4's architecture lends itself to quantization and pruning in a way that Gemini's larger models don't. The on-orbit model handles weather pattern recognition, atmospheric CO2 analysis from spectroscopy data, and some early-warning wildfire smoke detection tasks. Google says the first real test last week involved predicting a storm system developing over the Pacific, and the satellite's inference completed in 2.8 seconds compared to the 40-plus seconds it takes to downlink the same raw data and process it on the ground.

Why This Actually Changes Things

The satellite internet connectivity problem isn't just about bandwidth. It's about latency and timing. When you're trying to monitor a fast-moving wildfire or track a tropical system that could hit land in 18 hours, every minute you spend waiting for data to hit the ground, get processed, and have results sent back up is a minute where things could go wrong. Suncatcher's approach of doing inference at the edge, in orbit, means you can get actionable outputs in near real-time without waiting for the full round-trip. Google is positioning this as the first step toward a constellation of AI satellites that could collectively monitor the entire planet continuously, not just when one of them happens to be flying over your area of interest.

The Hard Part Nobody Talks About

Here's what I think enuinely impressive and also genuinely hard: keeping the models updated. On the ground you can retrain and deploy a model update in hours. In orbit you're fighting physics. Suncatcher has a communication window of roughly 8 minutes per pass over a ground station, and uplink bandwidth is limited. Google apparently solved this by keeping most of the model frozen and only fine-tuning a small adapter layer that can be compressed and uplinked efficiently. But it's still an engineering challenge that no one had really solved before at this scale.

Who Comes Next

Amazon has a similar project in its Kr A initiative but it's at least 18 months behind by most estimates. Microsoft has been quiet about orbital compute but has made acquisitions in the space hardware space that suggest they're working on something. The real competition might not come from another US tech company. China's CASC has published papers on radiation-hardened AI chips for satellite applications and their timeline for a demonstration mission is reportedly Q2 2027. If this is a race, Google just lapped the field. But space is expensive and the economics of AI inference satellites are still genuinely uncert

What's your experience? Drop a comment below! 👇

Do you think satellite-based AI inference is the future of environmental monitoring, or is this more of a Google moonshot project that won't change much for most applications? What use case would you most want to see from an orbital AI system?

#Google #SpaceAI #Suncatcher #Satellite #TPU #Gemma #AIInSpace #SpaceTech #TechNews #ArtificialIntelligence

Comments

Post a Comment

Popular posts from this blog

Plank You App Support and Privacy

-
Plank You is a simple, ad-free interval timer for planks and core workouts. Set your work and rest times, hold your plank, and let the countdown beeps do the counting. Everything runs on your device. no ads, no accounts, no tracking. How to use Start / Pause / Resume: tap the main button to run or pause the timer. Reset returns to Rep 1 and the start of your work interval. Rep counter — counts up automatically after each full work + rest cycle. Beeps — the last 5 seconds of every phase beep down (5-4-3-2-1), with a longer tone when the phase changes. Changing intervals Tap the settings icon (top right) to choose an interval: 45 / 15 default (45s work, 15s rest) 30 / 30 50 / 10 Custom — set any work and rest time in 5-second steps You can also toggle Beep sounds and Haptics in Settings. FAQ I don't hear any sound. Sound plays even when your ringer is on silent, but make sure the media volume is up and that "Beep sounds" is turned on in Settings. If another app is ...
Read more

WWDC 2026 Preview: Apple Needs to Fix Siri or It's Game Over for Apple Intelligence

-
Apple's WWDC 2026 kicks off on June 8, and the stakes couldn't be higher. While Google announced 900 million Gemini users at I/O last week, Apple is heading into its biggest developer conference with one uncomfortable question hanging over the whole event: what happened to Siri? I've been using Apple devices daily since the original iPhone, and I'll say it plainly — Apple Intelligence has been the most underwhelming software launch in Apple's recent history. WWDC 2026 is Apple's chance to fix that. Here's what they need to announce, what we're expecting, and what a failure would mean. The Problem Apple Needs to Solve When Apple announced Apple Intelligence at WWDC 2024, the demos were impressive. When the features started rolling out through 2025, the reality was a different story. The "new Siri" was slower, less reliable, and often less capable than the old one for basic tasks. The "Personal Context" features that were supposed ...
Read more

This AI Startup Is Worth $26 Billion and Writes 90% of Its Own Code — Should Software Engineers Be Worried?

-
I've been watching the AI coding space very closely for the past year, and every few months something happens that forces me to recalibrate my sense of how fast things are moving. This week, that moment came from Cognition AI — the startup behind Devin, the AI software engineer — which just announced a $1 billion funding round at a $26 billion valuation . But the number that actually stopped me cold wasn't the valuation. It was this: 90% of Cognition's own code is now written by Devin. From $10B to $26B in Eight Months — The Numbers Are Staggering Let's put this in perspective. Just eight months ago, in September 2025, Cognition closed a $400 million round at a $10.2 billion valuation. Today, that valuation has more than doubled to $26 billion. The new round was co-led by Lux Capital, General Catalyst, and 8VC, with Ribbit Capital, Atreides Management, and Peter Thiel's Founders Fund also participating. But what's really driving this isn't hype — it...
Read more