“No Data Center required” — maybe the most important words uttered thus far in this new AI civilization. I won’t elaborate here, but some of you will know.
Historic: Nvidia CEO Jensen Huang just hand-delivered the world’s smallest AI supercomputer to Elon Musk at SpaceX 😳
Just hours before Starship Flight 11.
The device: NVIDIA DGX Spark.
1.2 kg. Desktop-sized.
1 petaflop of AI performance – that’s a trillion calculations per second on your desk.
Powered by the new Grace Blackwell GB10 Superchip, with 128 GB of unified memory and bandwidth 5× faster than PCIe Gen5.
It can fine-tune 70B-parameter models locally. No data center required 🤯
9 years ago, in 2016, Huang drove to OpenAI’s tiny San Francisco office
and hand-delivered the first DGX-1 to Elon.
That machine helped train the models that became ChatGPT.
It marked the start of the AI revolution.
Today, that same 1 petaflop fits in your backpack 🤖
Moore’s Law didn’t die, it evolved.
My takeaway is this:
DGX Spark democratizes supercomputing.
You can now run large models locally – with data privacy, no latency, and no cloud bill.
For researchers, startups, and creators, this is the MacBook moment of AI hardware.
P.S. check out 🔔linas.substack.com🔔, it’s the only newsletter you need for all things when Finance meets Technology. For founders, builders, and leaders.
