On-premise building analytics and AI for agencies, utilities and operators whose data cannot leave site.
Aqueduct can be deployed on-premise, with its AI running on NVIDIA infrastructure on your site, for facilities and infrastructure whose operational data cannot leave. You keep the same platform as the cloud deployment, and the data stays where your policy says it must.
The same platform and the same AI, running where your policy says data stays.
Run Aqueduct inside your own network and policy
- Your infrastructure
aQ Core, the digital twins and analytics run on hardware at your site, inside your network rules.
- Same platform
On-prem uses the same platform as the cloud deployment, so features and workflows do not fork.
- Your access rules
Who can see and do what follows your own policies, set during scoping and security review.
AI on NVIDIA infrastructure, right next to the equipment
- Local answers
aQ Assistant answers questions on site, so operational data does not leave to produce an answer.
- NVIDIA Inception member
Milvian AI is a member of the NVIDIA Inception program and runs on-prem AI on NVIDIA infrastructure.
- Built for critical sites
Agencies, utilities and plants that cannot send data out still get plain-language analysis with evidence.
Monitor, measure, identify, act and verify without leaving site
- Baselines on premises
Weather-normalized baselines and savings are calculated locally from your own meter and bill history.
- Control with approval
Where control is enabled, with owner approval, commands run from site with an auditable command history.
- Verified results
Every fix is read back at the meter on site, and the verified metrics export from there.
Cloud-only tools versus Aqueduct on your premises
- The policy says data stays, so the analytics tool is ruled out.
- Critical infrastructure runs blind because cloud tools fail review.
- Procurement stalls for months on where data is hosted.
- An on-prem version lacks the features of the cloud product.
- AI answers require sending operational data off site.
The Aqueduct way- Aqueduct runs on site, inside the boundary your policy sets.
- Critical sites get the full flywheel without data leaving.
- Where data lives is answered at scoping, not at the end.
- On-prem runs the same platform as the cloud deployment.
- AI runs on NVIDIA infrastructure at your own site.
When cloud is not an option
Some operators cannot send meter, equipment or occupancy data off site, whatever the benefit.
Policy says the data stays
Public agencies, utilities and regulated operators often have rules on where operational data may be stored and processed.
Critical infrastructure
Water networks, plants and grids treat operational data as sensitive. Analytics has to come to the data, not the other way around.
Procurement asks first
Data residency is often the first question in a security review. An on-prem option answers it before it becomes a blocker.
What runs on site
The same platform, deployed on your premises.
aQ Core on your site
Sites, floors, zones and devices, ingest and storage, in your own environment.
AI on NVIDIA infrastructure
The AI behind aQ Assistant and the outcomes runs on NVIDIA infrastructure on site.
The full flywheel
Monitor, Measure, Identify, Act and Verify, with the same outcomes as a cloud deployment.
Your network, your rules
Deployment architecture, sizing and access are agreed with your IT and security teams during scoping.
Cloud or on-prem
Both are the same product. The choice is about where data lives and who runs the infrastructure.
| Question | Cloud | On-prem |
|---|---|---|
| Where data is stored and processed | In the cloud, delivered as a service | On your site |
| Outcomes available | Water, energy, air and operations | Water, energy, air and operations |
| Where the AI runs | In the cloud | On NVIDIA infrastructure on your site |
| Who runs the infrastructure | Managed for you as a service | Agreed during scoping with your IT team |
| Security and residency detail | Covered in our security questionnaire during procurement | Covered in our security questionnaire during procurement |
Who asks for on-prem
“Can I run analytics on my network without sending data to someone else's cloud?”
Yes. Aqueduct deploys on-prem AI on NVIDIA infrastructure, and it works alongside your SCADA. Control stays with SCADA.
“Do I lose features if I keep the data on site?”
No. The same outcomes and the same flywheel run on-prem.
“Can I specify it for a client with strict data rules?”
Yes. Specify on-prem deployment at design or procurement stage, and we size it with you.
Where your data lives should be your decision, not a product limitation.
Talk to our teamHow an on-prem deployment runs
Same delivery model as cloud: product from Milvian AI, installation through partners.
Scope
Agree which sites, systems and data stay on site, and what the infrastructure needs to be.
Answer the questionnaire
Your security and residency questions are answered in our security questionnaire during procurement.
Connect
Meters, sensors, BMS points and equipment connect to the on-site platform, through partners where hardware is needed.
Baseline and run
Upload 6 to 12 months of utility bills for day-one baselines, then run the flywheel on site.
About NVIDIA Inception
Milvian AI is a member of the NVIDIA Inception program. Aqueduct runs on-prem AI on NVIDIA infrastructure for sites that cannot send data out.
Hardware selection and sizing for an on-prem deployment are agreed per project. We don't sell hardware as our core business.
Where on-prem comes up
Glossary entries for sectors that often ask about residency.
Where teams put it to work
Analytics that clear a data policy
Agencies whose data cannot leave government premises run the same water, energy and air outcomes on site, inside their own network rules.
Learn more
Alongside SCADA, inside the fence
Utilities read network data alongside SCADA on their own infrastructure, while control of the network stays with SCADA where it belongs.
Learn more
Plant data stays at the plant
Manufacturing and process sites keep operational data on premises and still get baselines, fault findings and verified results from the same platform.
Learn moreRelated
Common questions
Yes. Aqueduct can be deployed on-prem, with its AI on NVIDIA infrastructure on your site, for data that cannot leave.
Milvian AI is a member of the NVIDIA Inception program. Aqueduct runs on-prem AI on NVIDIA infrastructure.
No. Water, energy, air and operations, and the full Monitor, Measure, Identify, Act, Verify flywheel, run the same way.
Sizing and supply are agreed per project during scoping. We don't sell hardware as our core business.
They are covered in our security questionnaire during procurement. Ask for it at the start of the process.
See Aqueduct on your own facility.
Bring a site, a utility bill or a question. We will show you where the loss is.