The on-site AI datacenter for robot fleets.
Robot brains are growing faster than robot bodies. We're the missing infrastructure.
Backed by
.webp)
CHALLENGE
The robot autonomy ceiling
Three forces are converging to break today’s robot infrastructure.
More robots, interventions rise
A 50-robot site is a mini datacenter.
New skills, throughput drops
Every new capability stacks more compute.
More automation, more overhead
Duplicating GPUs per robot doesn't scale.
There is no purpose-built compute layer between the cloud and the robot.
That's the gap Nectar fills.
SOLUTION
Box + Brain: The On-Site Intelligence Layer
One Box on-site runs the heavy AI for the whole fleet. Robots keep only safety-critical motion onboard.
THE BOX
27ft³ Immersion-Cooled GPU Node
Deployed on-site. Runs multi-camera perception, semantic understanding, and planning for every robot on the floor.
THE BRAIN
Fleet Orchestration Software
Manages workload scheduling, thermal monitoring, and predictive maintenance across every node in your fleet.
THE FLYWHEEL
Inference by Day, Training by Night
Your robots run inference during shifts. After hours, the same GPUs fine-tune models on the day’s data — every cycle makes the fleet smarter.
Before Nectar
After Nectar
Cloud-only, $3.50/hr per GPU
On-prem, $0.40/hr amortised
50-200 ms round-trip latency
<5 ms on-site inference
Data leaves the building
Data never leaves your site
We are a software company disguised as a box. Hardware is the toll booth. Brain is the recurring revenue.
Frequently Asked Questions
How does the Nectar Box differ from cloud GPU?
+
The Nectar Box sits on your factory floor, delivering sub-5 ms inference with zero data leaving your site. Cloud GPUs add 50-200 ms of round-trip latency and send proprietary production data off-premise.
What hardware powers the Nectar Box?
+
Each 27 ft³ immersion-cooled node packs NVIDIA H100 GPUs, high-bandwidth NVLink interconnects, and redundant power in a form factor that fits through a standard warehouse door.
Can I scale up after starting with one Box?
+
Absolutely. The Brain orchestration layer manages workload distribution across any number of Boxes. Add nodes as your fleet grows and the software handles scheduling, failover, and model versioning automatically.
Request a Demo
Train in the cloud. Decide at the edge.
Deploy in one click. No setup. No stress.
Frequently asked questions
How does the Nectar Box differ from cloud GPU?
What hardware powers the Nectar Box?
Can I scale up after starting with one Box?
Get in touch
Have questions? We'd love to hear from you.
By submitting, you agree to our Terms of Service and Privacy Policy.
By submitting, you agree to our Terms of Service and Privacy Policy.