- AI Valley
- Posts
- Nvidia is sending AI chips to space
Nvidia is sending AI chips to space
PLUS: AI is starting to move into the real world
Together with
howdy, it’s Barsee again.
happy tuesday, AI family, and welcome back to AI Valley.
here are the biggest things worth knowing today:
Nvidia is taking its AI chips to space
AI is starting to move into the real world
Plus trending AI tools, posts, and resources
Let’s dive into the Valley of AI…
VIKTOR

Courtesy: Viktor.com
Every model release makes AI smarter for free. The harness around the model decides how much of that turns into real work. A model that could do a great analyst's job is still stuck if it forgets everything between chats and only acts when someone types at it.
So we built Viktor for the models that are coming: a persistent workspace where it writes and runs code, memory that compounds across sessions, schedules it sets for itself, and 3,200+ tools it can call. The result is an AI employee in Slack and Microsoft Teams that runs work end to end. 60,000+ teams have hired one.
"I won't lie, when I first set it up... I was dubious. But wow... it's impressive." Simon Chapman, CTO
The full essay on our harness decisions is in the post.
*This is sponsored
THROUGH THE VALLEY

Google’s Project Genie | World Model prototype
1/ Nvidia is taking its AI chips to space - SpaceX is building its Starmind orbital data centers around Nvidia’s Vera Rubin platform, with a space version targeted for Q4 next year. SpaceXAI is also using Nvidia’s Vera CPU for Grok’s agents.
The space version is apparently lighter, denser and cheaper, and built to deal with radiation and heat. Orbital compute costs 4x+ what it costs on Earth today, but Musk thinks that eventually flips.
Nvidia is also going after the CPU side of AI. Vera is built to handle all the boring work around agents, like running code, calling tools, and moving data around, with Nvidia claiming up to 1.8x faster agentic task completion than traditional x86 chips.
So yeah, AI data centers might not all be on Earth forever.
2/ AI is starting to move into the real world - a bunch of new “world models” are being trained on videogames and simulations to control robots. Instead of learning what a cup is from text, they learn what happens when you actually move the thing.
General Intuition is one of the companies doing this. They’ve trained on millions of hours of videogame play, including the actions people take, and can apparently fine-tune the model to control a robot dog, drone or other machine.
Google DeepMind is already combining this with Gemini in SIMA 2. Gemini handles the reasoning and goals, while SIMA learns to actually carry them out inside 3D game worlds.
And there is a lot of money going into this. World Labs raised $1B at a $5.4B valuation, AMI Labs raised $1.03B at $3.5B, General Intuition raised $133.7M, Physical Intelligence raised $600M at $5.6B, and Wayve just raised $1.2B at $8.6B. Google DeepMind is also pouring resources into SIMA, Genie and Veo, and Demis has said he expects world models to become an important part of Gemini’s planning.
The fun question is what happens when these things get plugged into LLMs. LLMs are pretty bad at understanding a 3D world, and you probably don’t want your robot hallucinating when it’s moving around.
Feels like we’re at the GPT-2 stage of physical AI.
More on world models:
What’s a world model?
The Waymo world model - a new frontier for autonomous driving simulation
TRENDING TOOLS
Lucy 2.5 - live AI model for editing video in real time with more control
canIrun.ai - checks your hardware and tells you which AI models will actually run well before you download them
Energy - another desktop app to work with agents
Wan 3.0 - Alibaba launches its latest AI model
Spline - rebuilt its browser-based 3D editor on WebGPU and added an agent that can inspect scenes, use the editor’s tools, and take screenshots to check its work
Atlaso - gives Claude Code, Cursor, and Codex a shared memory, so what you tell one tool, the others already know
GPT-Image-2 - adds transparent backgrounds, making it easier to create reusable assets
Claude Academy - Anthropic’s collection of courses, tutorials, and job-specific guides for getting more out of Claude
WHAT I'M CONSUMING
The internet is filling with junk
AI adoption is a myth
The self-driving company - by Replit’s CEO
THE VALLEY GEMS
THAT’S ALL FOR TODAY
Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee
👍️ New reader? Subscribe here
Thanks for being here.
REACH 100K+ READERS
Acquire new customers and drive revenue by partnering with us
Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.
If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.

