7 Reasons Data Center Inventory Falls Out of Sync
By Matt Mescall, Space and Power Product Manager
Data center inventory records fall out of sync for a few key reasons: ineffective inventory management solutions, the rate of change in data centers, and shifting priorities to remain competitive.
Some of the generic culprits for this include error-prone spreadsheets, siloed systems with conflicting records, and reliance on audits to track capacity. But in this blog, I’ll dive deeper into seven other reasons that your DCIM might not match the reality on the floor.
Why Does Data Center Inventory Accuracy Matter in 2026?
Data accuracy is critical for any organization focused on making data-driven decisions, launching AI agents, and preventing revenue leaks. Gartner research surfaces soaring costs from poor data quality, with organizations often treating data like an afterthought. And data centers aren’t exempt from that.
If anything, data centers are particularly under pressure to get their data right.
North American colocation vacancy is at a record low 2.3%. Meanwhile, capacity demand is at historic highs.
McKinsey currently projects $6.7 trillion in investment by 2030. But experts also warn that overinvestment may strand assets.
Whether you’re preventing stranded capacity, maximizing revenue, or optimizing your data center for performance, it all comes back to needing data you can trust.
7 Reasons Inventory Data Drifts
As-Built Records vs. Designs
Preleasing Gaps
Complex Power Capacity Tracking
Density Changes Mid-Lease
Untracked NetEx
Sales Holds That Never Release
The Decommission-to-Resale Lag
Reason 1: As-Built Records vs. Designs
When a new data hall gets built, the construction and commissioning teams produce “as-built” documentation to record what was actually installed. These records can differ from designs as breakers get upsized, a PDU gets relocated, or cabinet positions shift during construction. But that means the as-builts capture the final reality.
The problem is the format. That documentation is delivered as construction deliverables: CAD or PDF drawings, commissioning test reports, spreadsheets from the electrical contractor. None of it is structured data that loads cleanly into typical inventory or DCIM systems.
Someone has to translate those documents into digital inventory records. With the wrong systems, this can be tedious and error prone, especially since it’s often happening under tight deadlines because the data hall needs to start earning money. So, these digital records might end up being incomplete from the start, impacting every downstream capacity report, quote, and invoice.
Reason 2: Preleasing Gaps
Preleasing has become the industry standard. In North America alone, roughly 73% of capacity under construction is already preleased.
This business model makes sense, but here’s the problem: large deals don’t deliver all at once.
A 20 MW lease might hand over in 5 MW phases as each hall gets commissioned, spread across quarters or years.
That creates three distinct numbers for one customer, alive at the same time:
Committed: What the contract promises. 20 MW.
Delivered: What’s physically been handed over so far. 5 MW today, 10 MW next quarter, and so on.
Billable: What’s generating revenue right now, per the ramp schedule and billing start dates.
When you use multiple point systems, each number is owned by a different team in a different tool. Even if they start aligned when the contract is signed, they can drift with time if construction slips or phases complete early or late, or if ramp terms get renegotiated.
Reason 3: Complex Power Capacity Tracking
Power capacity gets recorded in multiple ways.
Contracted kW. This is the commercial number. For example: the lease says the customer gets 500 kW. It defines what they pay for and, in most colo contracts, what they’re allowed to draw. That number lives in the CRM.
Design capacity. This is the engineering number that states how power flows through the facility: utility feed, switchgear, UPS, panels, breakers, down to the circuits feeding the customer cabinets. When the facility team provisions that 500 kW customer, they physically build it as breakers and circuits, and breakers come in standard sizes. So, the customer might actually be wired for 600 kW of breaker capacity to provide headroom. That number lives in electrical drawings and the DCIM.
Metered draw. This is what the customer actually consumes. Maybe 340 kW on average. This number lives in the power monitoring or BMS system, owned by operations, and can change often.
If you only know one number, you don’t get the full picture. The trouble is that billing and capacity planning each need to combine them, and the systems don’t talk.
Billing depends on the contract model. If the customer pays a flat rate for their allocation, billing needs the contract number. If they pay for metered usage, billing needs the meter. Either way, billing also needs to catch the case where metered draw exceeds the contracted allocation, because that overage is either a chargeable event or a contract violation.
The same goes for sales. When sales asks, “How much power can we still sell in this data hall?” the honest answer depends on all three numbers. If the solution engineer only sees contracted numbers, the hall looks full while megawatts sit idle. If they only see metered draw, the hall looks like it has availability while ever kW is contractually spoken for.
Reason 4: Density Changes Mid-Lease
When a deal is signed, the operator engineers the space around a density assumption.
A customer running conventional enterprise servers draws maybe 8 to 12 kW per rack, so the operator sizes everything to that: circuits rated for that load, cooling designed for that much heat spread across that floor area, and raised floor rated for that weight.
Those assumptions become the power, cooling, and floor-loading records for that cage. They’re accurate on day one. But here’s the problem: a lease can run 10 to 15 years, but IT hardware refreshes every three to five, so a single lease contains multiple generations of customer equipment, and the records from signing day describe only the first iteration. Every refresh widens the gap between what was engineered and what’s on the floor today.
Reason 5: Untracked NetEx
Cross-connects are among the highest-margin recurring revenue in the building, and they change constantly.
Every change, new install, or disconnect is a work order. The gap is whether that ticket hits the billing system or not. Turning work orders into a new recurring charge, or stopping an existing one, is a separate step that depends on someone carrying it out. When the handoff is manual, it fails at some rate.
A single cross-connect billing error is a small monthly charge. But hundreds of missed connections add up to a sizable billing leak, increased customer disputes, and NetEx that you’ve lost track of.
Reason 6: Sales Holds That Never Release
When a rep is working a deal, they need to know the capacity will still be there when the customer signs. So, they hold it for their prospect. A good inventory system should mark that capacity as “unavailable” and prevent other reps from selling it.
But most systems aren’t connected from sales to billing. Releasing that hold and updating its status through the sales process might be manual or be forgotten altogether.
That means holds accumulate over time. Every hall carries a layer of held capacity attached to opportunities that are dead, shrunken, or dormant, and nothing in the system forces a review.
Reason 7: The Decommission-to-Resale Lag
When a customer churns, the clock starts on getting that space and power back to market. The cage itself may need teardown. Cross-connects have to be pulled from the meet-me room. Circuits have to be disconnected and breakers de-energized. The space may need cleaning or remediation before it can be shown. And somewhere in there, the inventory record has to change from occupied to available.
Each of those steps is a separate work order, owned by a different group, often with no shared deadline. Facilities handles the electrical. Network handles the cross-connects. Operations handles the physical teardown. Nobody owns the sequence end to end, and each individual ticket seems low-priority compared to work for paying customers.
That means the contract end date passes, but the record stays occupied because equipment is still on the floor. And every week this space sits off the market is revenue you can’t recover.
How Do You Keep Inventory, Power, and Billing in Sync?
The good news is that the root of many of these issues comes down to having the right system and processes to support data center management from quote to cash.
Fixing the pattern requires one inventory system that every downstream process relies on: quoting, design, provisioning, and billing. When a remote hands ticket closes or a phase gets handed over, the change should reach the capacity model and the billing record in the same motion.
That's the design principle behind platforms like Carma, which keep inventory, power, and revenue records on a single data model, so everything stays in lockstep as you operate.
Author Bio
Matt Mescall is Carma’s product manager responsible for space and power functionality. With an engineering background and over 15 years of experience in data centers, Matt brings first-hand expertise to Carma’s product development.