Customer stories · Independent / creative
Indie developer & design studio
The work that should stay home, One isn't 'send everything to the cloud'
A LoRA fine-tune and a 500-frame render run on-device for $0, fully private, because One bursts only when it actually helps.
The challenge
Not every job belongs in the cloud. Small fine-tunes and batch renders fit a Mac, paying to ship them out wastes money and privacy.
Workload understanding
One found both jobs fit comfortably under 80% of a Mac Studio's 192 GB unified memory (≈114 GB and ≈90 GB headroom), so it kept them on-device.
Best hardware for the job
On-device (Apple Silicon), no cloud hardware needed.
Completion
Ran locally; the data never left the machine and no instance was ever provisioned.
Outcome
$0 cloud
Cost
Data never left the device
Privacy
Full, no shrink-to-fit
Accuracy
“It just ran it on my Mac and told me why. No upload, no invoice.”
Run your workload in your own cloud.
Representative composite story. Placement, hardware matching, and completion are real system behavior; figures are transparent, editable model inputs.