Building the intelligence layer for the physical infrastructure behind AI.
The AI era runs on data centers, and those data centers run on power, cooling, and water. AquaWatt exists to make that physical layer intelligent, efficient, and accountable — without ever compromising reliability.
“Every watt and every drop of water behind AI should be measured, understood, and optimized — with humans firmly in control.”
The AquaWatt mission
The principles we build by.
These commitments shape every product decision we make.
Reliability before efficiency
Uptime and equipment protection always come first. Efficiency gains never compromise reliability.
Optimize the whole system
Power, cooling, water, and compute are optimized together, not as isolated silos.
Explain every recommendation
Operators see the reasoning, expected impact, and trade-offs behind every action.
Simulate before acting
Changes are tested against a digital twin before they are ever proposed for the real facility.
Humans control autonomy
Operators decide how much authority to delegate, and can step in or roll back at any time.
Verify actual outcomes
We measure the real-world result of every action, not just the modeled prediction.
Keep safety-critical capabilities local
Safety logic runs at the edge and keeps protecting equipment even without a cloud connection.
Respect local resource constraints
Optimization reflects the real energy, water, and carbon conditions of each site.
Our team
Leadership and team information coming soon. We’re assembling a team that spans data center operations, control systems, AI, and sustainability.
Partner with us on the pilot program.
We’re working with a select group of forward-looking data center operators.
Where AquaWatt sits in the stack
For data-center operators facing increasing electricity, cooling and water constraints, AquaWatt DC OS is an AI-powered resource operating system that continuously coordinates facilities, workloads and energy systems to reduce cost and environmental impact without compromising uptime.
Unlike conventional DCIM, BMS or sustainability-reporting tools, AquaWatt predicts future conditions, simulates potential actions, optimizes multiple resources simultaneously and safely progresses from recommendations to autonomous operations.
DCIM tells you what happened
AquaWatt predicts what is about to happen
BMS controls one system at a time
AquaWatt optimizes power, cooling, water and carbon together
Sustainability tools report after the quarter
AquaWatt makes sustainability an operating variable
Point tools recommend and stop there
AquaWatt simulates, verifies and progresses safely toward autonomy
Where this could go
Once a facility’s resources are modelled, forecast and verified, a much larger set of problems becomes addressable. These are directions under consideration — explicitly not commitments, and not part of any current release.
Comparisons describe general categories of tooling rather than any named product or vendor. Future opportunities are exploratory and carry no delivery timeline.
From visibility to resource autonomy.
AquaWatt DC OS earns trust before it takes control. The platform begins in observation and recommendation modes and progresses toward controlled automation only after models, integrations and safety controls are validated.
Discovery and design
0–3 monthsVisibility MVP
4–9 monthsRecommendation intelligence
10–15 monthsGuarded automation
16–24 monthsPortfolio orchestration
25–36 monthsResource-autonomous data centers
Beyond 36 monthsForward-looking product roadmap. Phases, durations and capabilities are planning targets and are subject to change based on pilot validation, customer requirements and engineering findings.
The things that could go wrong
Every one of these has sunk an optimization programme somewhere. We would rather name them here than discover we disagree about them during a deployment.
- Poor-quality telemetryData-quality scoring, sensor validation, missing-data handling and onboarding diagnostics.
- Customer reluctance to permit automated controlBegin with observation, shadow mode and recommendations; provide complete explainability and rollback.
- Incorrect optimization compromises uptimeIndependent safety-policy engine, digital-twin testing, approval workflows and hard limits.
- Savings are difficult to proveEstablish approved baselines before recommendations and retain complete measurement lineage.
- Different facilities require extensive customizationStandardized equipment models, protocol adapters and configurable facility templates.
- Cybersecurity exposureZero-trust architecture, outbound-only edge communication where possible, private connectivity and isolated control channels.
- Electricity optimization increases water consumptionMulti-objective optimization with local water-stress weighting.
- Water optimization increases carbon or costDisplay trade-offs and enforce customer-defined objective limits.
- AI model driftContinuous monitoring, scheduled validation, shadow deployments and automatic fallback.
- Equipment vendors restrict external controlSupport recommendation-only mode and use approved vendor APIs and integration partners.
- Regulatory requirements differ across regionsConfigurable reporting boundaries, policies, data residency and regional deployment.
Mitigations describe the approach AquaWatt DC OS is being designed around. They are engineering commitments under development, not guarantees of outcome at any facility.
How an engagement is likely to be structured
We are not publishing a price list before we have earned one. What we can share is the shape of a commercial relationship we think is fair for infrastructure of this consequence — and the terms we believe belong in every contract.
- Annual platform subscription
- Price per managed facility
- Price per managed megawatt
- Price per optimization module
- Implementation and integration fees
- Digital-twin configuration fees
- Premium support
- Managed optimization services
- Percentage of independently verified savings
Four things every contract should settle up front
Ambiguity in these four areas is where optimisation programmes usually go wrong. We would rather negotiate them at the start than argue about them at the first quarterly review.
- Baseline methodology
- How the starting point is calculated and normalised, agreed before any optimisation begins.
- Data ownership
- Your operational data remains yours. Rights, retention and portability are written down, not assumed.
- Control responsibility
- Exactly which decisions are advisory and which, if any, are automated — and who signs off.
- Savings verification
- The method, confidence threshold and review process used to declare a saving verified.
Commercial components listed here are working assumptions under active development. They are not an offer, a quotation or a published price list, and pricing will be agreed individually with each early partner.