AquaWatt Pilot Program
A structured 6-month engagement to prove AI-driven resource optimization in a live data center environment.
6-Month Pilot Timeline
Month 1
Discovery + Integration
Site assessment, infrastructure audit, data source identification and initial AquaWatt Edge deployment.
Month 2
Data Validation + Baseline
Validate telemetry quality, establish operational baselines, configure digital twin models.
Month 3
Forecasting + Anomaly Intelligence
Activate AI forecasting for power, water, thermal and workload. Enable anomaly detection.
Month 4–5
Optimization Recommendations
Generate and evaluate optimization strategies. Begin human-approved actions for measurable improvement.
Month 6
Savings Verification + Expansion
Measure predicted vs actual savings. Document results and build the expansion business case.
Is your facility ready for a pilot?
Check the items your site already has. Nothing is submitted or stored — this is a quick self-assessment against the facility requirements we use when selecting pilot partners.
Facility requirements
Select the items your facility already has to see where you stand.
How we define a successful pilot
Success is agreed in writing before the engagement begins, so there is no ambiguity at month six.
- At least 95% of required telemetry is reliably ingested.
- Facility-level power forecasts meet the agreed accuracy threshold.
- Water forecasts meet the agreed accuracy threshold.
- At least three material inefficiencies are identified.
- Operators accept the recommendation workflow.
- At least one optimization measure produces verified savings.
- No recommendation causes an SLA or safety incident.
- Savings methodology is approved by the customer.
- A commercial expansion plan is established.
Pilot requirements and success criteria reflect the planned pilot design. Final scope, accuracy thresholds and savings methodology are agreed with each pilot partner during discovery.
The scoreboard, agreed before the pilot starts
A pilot that defines success at the end is a pilot that always succeeds. These are the measures AquaWatt holds itself to — including the ones that can go against us.
Useful computing output delivered per weighted unit of electricity, water, carbon and operating cost — while maintaining SLA compliance.
Every other measure exists to explain movement in this one. The SLA clause is not decoration: efficiency bought with degraded service does not count as progress.
Is the platform actually being used, and on how much infrastructure?
- Facilities onboarded
- Managed megawatts
- Active monthly users
- Connected assets
- Telemetry completeness
- Recommendations generated
- Recommendation acceptance rate
- Automated actions executed
- Copilot queries
- Reports generated
Did the facility measurably improve as a result?
- Energy saved
- Potable water saved
- Carbon avoided
- Cost saved
- Peak demand avoided
- Cooling efficiency improvement
- GPU utilization improvement
- Avoided thermal incidents
- Avoided downtime
- Deferred capacity expansion
Was the improvement earned safely, and can the platform be relied on?
- Unsafe recommendations
- Blocked policy violations
- Rollback frequency
- False alerts
- Missed critical anomalies
- Forecast accuracy
- Model drift
- Operator override rate
- Optimization-related SLA violations
This is the measurement framework proposed for pilot engagements. No values are shown because none have been recorded at a customer facility; targets are agreed individually with each pilot partner.