Lynnway Park: Assumptions & Sources

Rates are Boston-area market pricing, not vendor or tenant quotes. New Build defaults to a blend (60% Inference / 30% Enterprise / 10% Media ≈ $237/kW/mo), used only to report the landlord ratio below; the rent itself is an input. 626 Lynnway 2F defaults to a dedicated MEDC archetype.

Tenant archetypes. At 6–12MW a diversified Standard Colo mix is hard to fill, and a single specialist tenant (a "neocloud" GPU operator) is more realistic. Revenue Model, Advanced Analysis and MC Simulation toggle between three archetypes:

Construction is not what sets the timeline at this size. Uptime Institute puts ground-breaking to operation at 8–10 months for a 5–19.9MW build using best practice, excluding permitting and power (Uptime Institute, Best-in-class data center provisioning, 2020). The power connection does: National Grid's Massachusetts timetable gives 15–35 weeks for commercial service of 200kW or more, and 1.5–3.5 years for loads that need new substation or transmission work (National Grid, Connections Durations, September 2025). The 0 Circle Avenue feasibility study puts procurement, construction and commissioning at 15–24 months, 18–30 months in the downside; the model's 14-month construction sits below that range and should be read as the optimistic case. NeoCloud stabilises faster because large AI/GPU deals are increasingly pre-leased before completion. The MEDC conversion is faster still: a fit-out of an existing floor with power already on site, phased 1MW → 2MW → 4MW (see §3). Both Shell-Only archetypes default to the same $28.37/kW/mo rent, so NeoCloud's higher operator rate only lowers the landlord ratio, and its higher liquid-cooling CapEx lowers yield on cost. A higher-paying tenant type does not by itself raise landlord income.

The 2nd floor of 626 Lynnway (74,353 SF) is 12% occupied (≈65,431 SF vacant), of industrial construction, with an existing elevator. The site has two 3,750 kVA utility transformers, one of them idled.

Conversion runs 10–15% below new-build cost. Electrical and mechanical work is new either way; the saving is the avoided shell and civil work. The industrial construction and existing elevator also reduce structural reinforcement and vertical logistics costs.

Flood: the data hall is on the 2nd floor, but the 2021 as-builts put the 4,160V switchgear and transformers on the 1st floor, so 626 Lynnway carries its own $0.40M estimate. Power (2021 electrical as-builts): two 3,750 kVA utility transformers (~3.4MW each), one idled. The live one leaves room for about 2MW of IT after existing tenants, so Phase 3 (2 → 4MW) depends on National Grid re-energising the second; 626 Lynnway carries its own power estimates (480V fit-out $0.15M/MW, reinstatement $0.20M). Fibre: a second diverse route is available but not installed (estimate $0.25M). Landlord CapEx is floor preparation only, with structural, fire suppression and security costed per SF. Site Scenario Comparison has the full landlord breakdown.

Three 626 Lynnway items sit outside the CapEx table. Landlord rent: the data-hall rent is an input, default $11.43/kW/mo all-in. It is the figure the earlier method produced at the 4MW / 15,000 SF default: a 12% yield on the $2.72M CapEx (≈$7.99/kW/mo) plus the floor space at the Lynn industrial average asking rent ($11/SF/yr, range $5–25/SF; ≈$3.44/kW/mo at 3.75 SF/kW), the two lines the 626 Lynnway feasibility study's §7 table sums to its "owner basis". Nothing is added on top, and the operator's heads of terms will replace it. The remaining 25,000 SF of the ~40,000 SF lease is billed separately at $11/SF/yr, about $0.28M a year. Yield on cost is the result: ≈26.5% at the default, above DataCenterHawk's 7–14% shell deal band because the building is already owned (CapEx is fit-out only) and NOI includes the balance-space rent. 24/7 guard post: an annual landlord OpEx line ($0.1835M/yr, the midpoint of the feasibility study's Massachusetts wage-only benchmark, $145–222K/yr), recovered like insurance and property tax under the 98% NNN structure: mostly passed through to the tenant-operator at stabilised occupancy, with the landlord bearing a larger share during the early occupancy ramp.

The cage and floor preparation (structural, fire suppression, CCTV) is a single build-out sized to the full 15,000 SF data hall and spent once in Phase 1. Only the racks and equipment (operator CapEx, not landlord CapEx) scale with each later phase's added MW, so these three lines do not recur in Phases 2 and 3; the table shows the full 4MW build-out. The conversion is phased 1MW → 2MW → 4MW: grid work and logistics scale per phase, while flood, structural, fire suppression and security are Phase 1 only, sized to the full data hall.

The conversion's small landlord CapEx and rent reflect its scope. Standard "powered shell" market benchmarks (construction cost or rent, both in $/SF) price a landlord building a new powered shell from scratch: foundations, envelope and a new utility service. At 626 Lynnway the building already stands, power is on site (two 3,750 kVA utility transformers per the 2021 as-builts, one of them idled) and the operator brings its own equipment. The landlord confirms the floor can take the load, provides a security cage and mantrap, and covers flood protection and grid work, so a new-shell comparable would price in scope Lynnway Park does not fund. With no clean data-center-specific rent comparable for this scope, the default rent was set from a 12% yield on the fit-out CapEx plus the floor space at Lynn industrial asking rents, and the resulting yield is checked against DataCenterHawk's shell deal band. The operator's heads of terms will replace it with a real figure.

New Build has two fundable structures, a toggle on Basic Analysis, Advanced Analysis and Revenue Model: Shell-Only (above: Lynnway Park builds only the shell; an operator funds and installs the MEP/IT) or Full-Turnkey Landlord-Funded (Lynnway Park funds the complete facility and leases directly to the tenant, with no operator in between), sourced from a dedicated 0 Circle Avenue feasibility study. Full MEP cost already includes shell/core (Turner & Townsend's cost allocation puts shell/core at 9–14% of the full-MEP total), so it is not a separate line on top of Shell-Only's.

The rent is an input and stays editable; yield on cost is the result (first stabilised year's NOI ÷ CapEx, at 6MW with land owned). Each default is what the earlier method produced at 6MW from a per-profile yield target (≈9%/8%/6%, from the 0 Circle Avenue feasibility study's implied gross rent/cost screens at each profile's midpoint pricing), so it is an estimate. The result sits below those targets because NOI is net of the landlord's maintenance and unrecovered OpEx. Civil and site work ($2M/MW) and a 20% soft-cost markup are Full-Turnkey-only lines; grid, flood and fibre reuse the Shell-Only site-level fields. Here the landlord bears facility maintenance (unlike Shell-Only and the conversion) but not operating staff, which stays with the tenant (§9). Recovery is 85%, not Shell-Only's 98% NNN, reflecting the modified-gross lease the 0 Circle Avenue feasibility study describes. A micro-colo operator can sublease at retail rates well above this landlord screen ($155–250/kW/mo per CBRE); that spread is the operator's margin, not landlord rent.

The behind-the-meter lead is the WIN Waste Saugus waste-to-energy plant, 1.1 miles away, 57.8MW nameplate and about 33MW firm. It is a commercial question (uncommitted capacity, power purchase position, and a private-wire easement across urban Lynn), not a modelling one, and is not priced here. See Environment §3.

A standard 24/7 Tier III colocation crew at 2,080 hours per FTE. A size-scaled industry benchmark suggests about 2 FTE for a 6MW site (see Benchmark Comparison §7), and security and monitoring are often outsourced or shared, so a leaner operator would likely run 6 to 8 FTE (about $0.6–0.8M a year). None of this is in the model.

No public benchmark splits data center maintenance by trade; industry puts total maintenance at ~2–5% of asset value a year, or ~40% of OpEx at scale. The model uses the CapEx sub-trade proportions (§4) as a proxy, since maintenance intensity broadly tracks the capital value maintained.

1. Data Center Staffing Levels, broadstaffglobal.com
2. Proper Data Center Staffing is Key to Reliable Operations, Uptime Institute Journal
3. Data Center Staffing Ratios 2026, irecruit.co
4. How Much Does it Cost to Build a Data Center?, dgtlinfra.com
5. Data Center Cost Per MW: 2026 Benchmarks, irecruit.co
6. Utility Interconnection in Massachusetts, Mass.gov
7. Data Center Retrofit and Conversion Costs in 2026: Turning Existing Buildings Into Critical Infrastructure, Terrapin Construction Group
8. Liquid Cooling vs Air Cooling for AI Data Centers: 2025 Analysis, introl.com
9. Ignore Data Center Water Consumption at Your Own Peril, Uptime Institute Journal
10. Hunter JO 05: CapEx and OpEx, hunterjo.nsw.gov.au
11. How Much Does It Cost to Build a Data Center? (O&M / maintenance OpEx share), encoradvisors.com
12. Data Center Maintenance Budgeting: Cost Benchmarks & Modeling Tools, caeled.com
13. Utility Interconnection / Basic Service (National Grid MA), nationalgridus.com
14. Data Center Retrofits vs. New Builds: A Contractor's Perspective, cadencenow.com
15. Data Center Retrofit & Conversion Costs 2026 (Warehouse-to-DC), terrapincg.com
16. Data Center Construction Costs Explained (structural loading, PSF/steel cost), truelook.com
17. Data Center CAPEX Calculator (contingency, design/engineering & PM soft-cost benchmarks), resistancezero.com
18. Data Center Cap Rates 2026: Complete Investor Reference (powered-shell development Yield-on-Cost, 7.00–8.50%), maxlifedevelopment.com
19. Cap Rates, Lease Structures, and Pricing Data Centers (MaxLife Academy course, cross-references the same 7.00–8.50% powered-shell YoC figure), maxlifedevelopment.com
20. Powered Shell Data Centers: Everything You Need to Know (scope definition: building envelope, raw utility power, fibre access; 10–20% of full turnkey cost), landgate.com