The cloud is a real place. It has breakers, batteries, chillers, cables, sensors, suspiciously warm aisles and, somewhere nearby, a spreadsheet that everyone hopes is current. Modius lives in that gap between what a data center is supposed to be doing and what its machinery is doing right now. Its OpenData software collects the chatter from physical infrastructure, translates it into a common language and gives operators one view of power, cooling, environment, network and IT assets. This is data center infrastructure management, or DCIM - a category with the rare distinction of sounding even less glamorous than the equipment it protects.
That dullness is the point. A data center makes money by being available, sellable and boring. The expensive moments arrive when a circuit is fuller than the planning sheet says, cold air misses the rack that needs it, an alarm sits inside the wrong console, or a technician drives to a remote site only to discover the device was fine. Modius sells fewer of those moments.
The product is a translator before it is a dashboard
Founded in San Francisco in 2004 by systems engineer and entrepreneur Craig Compiano, Modius began with a stubborn premise: an operator cannot make an intelligent decision without real measurements. OpenData reaches into mixed fleets of old and new equipment - IP devices, serial gear, meters, UPS systems, generators, cooling units, environmental sensors, servers and switches - then normalizes their readings. “Vendor-neutral” is not decorative language here. Data centers are accumulated histories of purchasing decisions. Asking the customer to replace every device with one manufacturer’s hardware would erase the economic case.
How the machinery becomes a decision
On top of that common layer sit 13 modules. Asset and Network Connections track equipment, ports, cables and dependencies. Monitoring, Power, Cooling and EPMS watch health, energy and electrical systems. Analytics turns history into trends. Planning runs capacity and change scenarios. Workflows constrains who can alter what, records every step and leaves an audit trail. A Portal API moves data into other systems. AI/ML looks for deviations; 3D visualization gives the model a spatial face. Customers can start with a narrower set rather than buy the whole cupboard.
The most persuasive chart is a utility bill
Modius’s cleanest business case comes from an anonymized California technology company. An energy audit found that rapid growth had put the data center’s N+1 cooling reserve at risk. The first thing failing was not a chiller. It was the comfortable assumption that installed capacity and usable capacity were the same thing. Airflow was poor, some machines were over-cooled, and hot spots survived despite ample total cooling.
The customer undertook a $710,000 capital project. A utility incentive covered $130,000. OpenData supplied continuous measurements to establish the baseline, find inefficiencies and verify the result. The case study reports 2.3 million kilowatt-hours of annual energy savings, $262,000 lower yearly operating expense and a 2.2-year simple payback. It also reports $1.35 million in avoided capital expense because recovered cooling capacity delayed or displaced other spending.
One customer project / Reported economics
Important fine print: $710,000 was the customer’s broader efficiency project, not a published OpenData license price. Modius does not post list pricing.
The distinction matters. Modius sells enterprise licenses scoped by deployment and modules, through direct sales, resellers, managed-service partners and technology bundles. Its own buying material discusses subscription and one-time licensing, but buyers must request a quote. The economic question is therefore not “What does a dashboard cost?” It is “What are bad capacity estimates, unnecessary site visits, manual audits, energy waste and an avoidable outage already costing?”
The first useful act of “AI” in a data center is often less exotic: discover that the room has enough cooling, but the cold air is going to the wrong machines.YesPress analysis
T-Mobile’s broken view became Modius’s opening
The sharper competitive story arrived in 2025. T-Mobile’s third-party legacy infrastructure-management system struggled to integrate the measurements needed for capacity planning and operating efficiency. After an RFI/RFP process, the carrier selected OpenData for more than 80 data centers. The demonstration changed the conversation: Modius was able to model complex conditions from T-Mobile’s existing environments and connect to its advanced asset-management systems. Support for both AC and DC power mattered, as did a shared source of truth for multiple departments.
This pattern repeats across Modius’s named accounts. STACK Infrastructure wanted a global, round-the-clock view that combined customer service profiles with telemetry and equipment health. Prime Data Centers uses the platform across more than 20 facilities, including secure customer dashboards. 365 Data Centers deployed it across a multi-site colocation portfolio to monitor power and offer more flexible, usage-based pricing. Cologix built an integrated customer-facing capability on OpenData. The customers are not buying a prettier gauge. They are turning facility readings into capacity, service, billing and risk decisions.
Partnerships extend that route. Dell and Intel once put OpenData directly onto ruggedized edge gateways, removing hardware selection and installation work from industrial IoT deployments. A 2025 agreement with Mitsubishi Heavy Industries pairs Modius analytics with power, cooling and control engineering. X1Studio resells the platform in Japan. For a private company of roughly five dozen people, the strategy is sensible: let larger hardware, engineering and channel partners carry OpenData into places a direct software sales team cannot reach alone.
AI gave a 20-year-old category a fresh headache
Traditional enterprise computing was not perfectly predictable, but its rack loads were comparatively polite. GPU clusters are not. Modius notes that a conventional rack may draw 8 to 12 kilowatts steadily while a GPU cluster can jump through a much wider 40-to-80-kilowatt range. That volatility tightens the link between a workload in the white space and the breakers, cooling loops and backup systems in the gray space. A daily report can be accurate and still arrive too late.
Modius’s market position sits between building-management systems, narrow monitoring tools and IT-centered asset platforms. Rivals include Nlyte, Schneider Electric, Sunbird, Vertiv, FNT and Device42, plus the formidable incumbents known as spreadsheets and homegrown code. OpenData’s wedge is real-time, vendor-agnostic OT-plus-IT integration on a single data foundation. Its modular design makes a gradual deployment possible. Its ISO/IEC 27001 certification, completed in 2025, helps with the security review that comes before critical-facility software gets anywhere near production.
There are trade-offs. Modius itself says Nlyte can be the better fit when patented intelligent asset placement or deeply embedded ServiceNow and BMC Remedy workflows dominate the requirement. A small server room with a few modern devices may get enough value from simpler monitoring. A digital twin also becomes digital fiction when nobody maintains the asset inventory. Software cannot rescue an organization that ignores its alarms, refuses to standardize change procedures or will not give facilities and IT shared ownership of the data.
The playbook readers can steal
The transferable lesson is not “buy DCIM.” It is to measure before automating. Begin with the costly uncertainty: a cooling reserve nobody trusts, a remote alarm that triggers truck rolls, tenant power usage that cannot be defended, or a planned rack whose downstream dependencies live in three separate tools. Instrument that path and agree on the baseline before commissioning an AI model or a wall of charts.
Choose a capacity, energy, uptime or labor question with an owner and a dollar consequence.
Map device readings to assets, locations and dependencies before building predictions.
Measure the old process long enough to make the after-state credible to finance and operations.
Put alerts, approvals and accountability into daily work. A dashboard alone saves nothing.
Then run one closed-loop project. The California customer’s cooling program worked because readings did more than reveal a hot spot: they justified a capital request, supported a utility incentive, guided the physical changes and verified the savings afterward. That is a much higher bar than installing software. It is also why the result could survive a conversation with a CFO.
Where the model breaks
- The site is too small or simple for avoided risk and labor to exceed integration and license costs.
- Equipment cannot expose reliable telemetry, and the retrofit budget is missing.
- Asset records have no owner, so the operating model decays after launch.
- Teams want prediction without changing maintenance, approval or escalation behavior.
- A specialized competitor matches a non-negotiable ITSM, compliance or placement workflow more closely.
Modius is not the company making the AI model, the server or the chiller. It is making the layer that tells an operator whether those things can coexist without an expensive surprise. After two decades, that layer has become interesting for a simple reason: the machines are changing faster, the racks are drawing more power, and the old spreadsheet has finally run out of polite ways to pretend it is real time.