Per-MW pricing, regional variance, and cost drivers for owners scoping hyperscale & AI builds.
Salary benchmarks across the 14 mission-critical disciplines.
If you need capacity fast or your demand may change, leasing usually makes more sense. If you have a stable multi-MW load, capital to deploy, and time to wait, building can cost less over the long run.
Here’s the short version:
Colocation Lease vs. Build Your Own Data Center: Cost & Key Metrics Compared
My takeaway: this choice is less about lease vs. own and more about certainty, capital, and timing. If your load is clear and large enough, build economics can work. If not, leasing can protect you from tying up millions too early.
Leasing trades large upfront CapEx for recurring lease payments. That can make a lot of sense when demand is hard to pin down, timelines are tight, or your team doesn't have the time to run a full build. The key question isn't just the monthly bill. It's how those charges shape total cost over time.
With colocation, you're paying for energized capacity, not just square footage.
A colocation contract usually bundles several cost items together. In most cases, you'll see rack or cage rent, or pricing based on $/kW-month, plus annual escalators, utility pass-throughs, taxes, cross-connect fees, and remote hands services.
If your setup needs site-specific changes, you may also see tenant fit-out or custom build-out charges. That matters because those extras can push some CapEx back to the tenant. So before you compare leasing with a self-build, model both:
That split is easy to miss, and it can change the math more than people expect.
A higher monthly spend doesn't automatically mean leasing costs more in the long run. In some cases, leasing still wins on total cost once you factor in timing and vacancy risk. Why? Because it cuts construction delay risk and keeps you from paying to build infrastructure you may not fully use.
Leasing tends to work best when you can add capacity in steps instead of building everything on day one. That's where the economics start to shift. The main advantage isn't just lower upfront spend. It's the way lease flexibility can offset capital risk and schedule risk.
These tradeoffs flow straight into break-even timing, financing cost, and staffing needs. That's why the next step is break-even analysis: the point where lease payments move past the cost of owning and running your own facility.
If leasing moves spend into OpEx, self-build does the opposite. It pushes spend into upfront CapEx, financing, and plain old execution risk. That model only works when you have enough scale, strong utilization, tight control over costs, and a build that stays on schedule.
A ground-up U.S. data center project comes with cost layers that normal commercial buildings just don’t. Owners usually track total development cost in two ways: cost per gross square foot and cost per megawatt (MW) of live IT load. Those numbers matter because each cost line changes the schedule, financing needs, and how long it takes to hit break-even.[3]
For a standard U.S. data center, total development cost lands at about $600–$1,100+ per sq ft or $8–$12M per MW of IT load. AI-tuned or liquid-cooled builds can climb to $15–$20M+ per MW.[1][3][4]
The main cost buckets are:
Ground-up delivery takes time, and delays aren’t just annoying - they’re expensive. Interest during construction (IDC) builds on drawn debt, and workloads that can’t leave colocation keep adding lease payments. So every extra month can mean two hits at once: more IDC and more lease spend.
Financing structure matters just as much as raw construction cost. Senior construction debt usually comes with terms of 12–36 months, interest-only on draws, at floating rates - often SOFR + 350–550 basis points, or about 9%–11% all-in - with 65%–75% loan-to-cost (LTC).[7][8] Mezzanine and preferred equity can push total financing to about 85% LTC, but the cost of capital rises to 11%–18%+.[8][10] After the facility stabilizes with contracted load, owners often refinance into longer-term permanent debt - CMBS or data center ABS - at around 5.75%–7.50% fixed.[8][9]
Power and staffing sit at the center of the ownership cost curve. A large owned facility usually costs $10M–$25M per year to operate.[11][12] Power makes up about 40%–60% of annual OpEx, with yearly electric bills of $1M–$5M per MW of IT load, depending on location and PUE.[11][14][16][17] Staffing often adds another 15%–25%.[13][14][15][16] The U.S. Chamber of Commerce puts annual OpEx for a typical large data center at $18.5M, including $7.4M for power, $2.8M for staffing, and the rest for taxes, insurance, and maintenance.[14]
Those inputs shape the break-even math in the next section.
The cost figures from the earlier sections don't tell the whole story. The real deciding factor is when one option pulls ahead of the other once you factor in utilization, financing, and schedule risk.
A simple way to pressure-test the choice is to look at three operating cases.
Scenario 1: Uncertain demand. When demand is unclear, leasing usually makes more sense because it keeps you from tying up capital too early. A greenfield build commits money before the load is proven, and that can leave you sitting on capacity you don't need yet.
Scenario 2: Stable multi-MW load. Ownership starts to look better when demand is steady and the power path is already in place. In large builds, phased turnover can bring specific data halls online while the rest of the site is still being finished. That helps utilization ramp sooner and gets the ROI clock started earlier. AI workloads add another layer here because density and cooling readiness can matter just as much as price.
Scenario 3: AI workloads. Leasing comes out ahead when existing colocation space already fits the workload's density, cooling, and power needs. If the setup is already there, you skip a lot of time, cost, and delivery pressure.
A break-even model should account for both discounting and delay risk. The main inputs are CapEx, lease rate, financing, PUE, ramp, and delay risk. The model is trying to answer one direct question: At what point does cumulative lease cost pass build cost?
Schedule risk is where many teams get tripped up. Delays almost never show up in isolation. They stack. One slip can ripple through financing, lease overlap, and lost revenue, which changes the math fast.
Here's the practical rule of thumb:
There's also an execution piece that matters. Self-build means handling permitting, procurement, commissioning, and owner oversight. Leasing pushes most of that work to the provider. And that choice shapes the team you'll need, the buying work involved, and the operating staff required to get the project over the line.
Once lease-versus-build economics are clear, the next question is simple: can your team actually deliver the plan? In practice, execution risk usually starts with talent. On self-build projects, missing people in mission-critical roles like construction leadership, MEP management, and commissioning is one of the main reasons schedules slip. And when the schedule moves, costs pile up fast through general conditions, liquidated damages, and delays to IT delivery.
A self-build program needs a deep bench before construction begins. The most important roles are project executives, MEP managers, superintendents, schedulers, commissioning leads, and QA/QC managers. One point matters more than most teams expect: commissioning authorities need to be involved during design, not brought in at the end. If that step comes late, the damage goes beyond schedule. It can also hit CapEx carry, lease overlap, and operational readiness.
A leased colocation program moves much of the construction load to the provider. But that doesn't mean the owner can step back. The owner still needs people focused on site diligence, SLA negotiation, fit-out coordination, and portfolio oversight. Without that layer of control, it's easy to end up with stranded capacity, higher power costs, or AI workloads that force retrofits later.
In short, the two models need very different teams.
Self-build teams also need operations roles that connect construction to live service. That includes SOPs, MOPs, EOPs, spare-parts planning, vendor onboarding, and acceptance testing. If handoff gets treated like a last-minute task, teams often run into commissioning failures and delayed readiness.
This staffing gap often decides whether a model works on paper only, or works in the field. Leasing gives you speed, flexibility, and lower execution risk. Building can produce stronger long-term ROI and more control at scale, but only if the owner has the right delivery and operations team in place.
Leasing often costs more once your long-term capacity needs get big enough to justify the steep upfront cost of building your own site. Across a 15-year period, operating costs account for about 62% to 68% of total ownership costs.
Building can lead to better ROI because the long-term cost per megawatt is lower. It also gives you more control over power pricing and infrastructure, helps you avoid AI colocation premiums, and makes it easier to standardize designs across sites.
Teams often miss a simple point: a “cheap” lease may not stay cheap once you factor in the costs they didn’t model.
That usually includes things like:
Soft costs and contingencies also get left out more often than they should. And benchmarked build $/MW figures can be misleading if they exclude land, early design work, and interconnection requests.
A self-build takes a lot of in-house bandwidth. To keep control and avoid losing know-how along the way, you need a permanent core team that covers development, finance, and operations. On top of that, you’ll likely need a few specialist roles to handle tough integration work.
A common setup is to keep permanent leaders in-house, then bring in embedded consultants or outside specialists during busy project stages. If your bench is thin, an owner’s representative or engineer can step in to handle technical oversight and program management.
