Dreamscale Labs
Y Combinator

Run robot brains
in the cloud.

Deploy frontier robot AI without expensive local GPUs.

Schedule a demo
Real-time execution of MolmoAct2 from a DC 1000km away.
40–50%

Lower tail latency than existing cloud providers.

320 ms

DreamZero 14B inference.
Beats NVIDIA's reported speed, with half the silicon

8+ robots

Running concurrently
on a single H100.

Choose a model.
Connect your robot.
We handle the compute.

Run models like DreamZero and MolmoAct2 in the cloud. We optimize the model, GPU kernels, and network to bring inference closer to your robot.

On-demand compute for the way robotics teams work: experiment, iterate, deploy.

Why we’re building this ↗

One platform.
The whole stack.

Performance is more than a fast model. We optimize the layers underneath it, so you don't have to.

Model

GPU kernels

Network

Real-Time Chunking (RTC)

Start with a frontier model.
Or bring your own.

Examples of providers whose models we can onboard.

  • NVIDIA
  • Ai2
  • Physical Intelligence
  • Robbyant
  • Perceptron
  • Hugging Face

Your checkpoint.

Fine-tuned for your robot.
Trained for your task.
Let’s get it running in the cloud.

Talk through your model ↗

Let’s get your robot
running in the cloud.

Schedule a demo