Data Center Solar Siting: Land Screening for EPCs

Data centers are one of the fastest-growing loads on the US grid. A 2024 report from Lawrence Berkeley National Laboratory estimates data centers used about 4.4% of US electricity in 2023 and could use 6.7% to 12% by 2028, according to the US Department of Energy summary.

Many large buyers want that power to be carbon-free. Lazard’s 2026 LCOE+ report says renewables remain the most cost-competitive form of new-build generation on an unsubsidized basis, so utility-scale solar is a natural fit for this demand.

Last updated: November 2026

SurgePV publishes this post and sells solar design software.

For EPCs and developers, this is a different kind of siting problem. The load is flat, the timeline is aggressive, and the interconnection queue often decides who wins. This guide is for teams moving into data center renewable energy work from residential, C&I or general utility-scale practice.

In this guide:

• Why data center demand is reshaping utility-scale solar development
• The six-stage data center solar siting process, from load definition to PPA
• GIS land screening criteria, with threshold values and data sources
• How to read interconnection queues and pick a point of interconnection
• Load matching math: annual energy, hourly coincidence, and CFE scores
• Battery storage sizing and the real cost of chasing 24/7 carbon-free energy
• Financial modeling inputs, sensitivities, and deal structures
• A worked, hypothetical 200 MW example with stated assumptions
• Seven siting mistakes that kill data center solar projects

Why Data Centers Are Driving Utility-Scale Solar Demand

US data center electricity use reached roughly 176 TWh in 2023, about 4.4% of national consumption. The LBNL 2024 United States Data Center Energy Usage Report projects 325 to 580 TWh by 2028, which is 6.7% to 12% of all US electricity. The DOE summary of the report adds that usage climbed from 58 TWh in 2014.

AI training and inference drive much of the growth, and large AI-focused campuses are being planned at hundreds of megawatts. A load that size can exceed what a local utility can serve without new generation and transmission.

Corporate climate commitments turn that load into renewable demand. RE100 reports over 400 member businesses in more than 150 markets, each committed to 100% renewable electricity by 2050 at the latest. Some buyers, including Google, have also published 24/7 carbon-free energy goals, which are stricter than annual matching.

Solar fits this demand on three grounds: cost, scalability and flexibility of siting. It can be built in phases, and the output profile is predictable enough to model hour by hour. There is also a structural reason it suits data centers: cooling load tends to rise with afternoon heat, which overlaps with solar output. The match is not perfect, and the gaps are covered below.

For EPCs, the implication is direct. The residential credit under Section 25D ended for expenditures after December 31, 2025, which pushes some teams toward C&I and utility-scale work, where data center contracts are pulling engineering talent and capital.

Teams that already run disciplined design workflows (layout, shading, production modeling and bankable proposals) are positioned to move up-market. The skills transfer, but the scale and the contract scrutiny do not. This guide assumes you bring the design discipline and focuses on what changes at data center scale.

Scale is the other shift. Utility-scale solar, meaning ground-mount projects above roughly 5 to 10 MW, now anchors the growth pipeline in many markets. A single data center procurement can be hundreds of megawatts, which is the size of a mid-size developer’s entire annual book. EPCs that win these contracts lock in multi-year backlogs and earn the reference projects that win the next one.

The Data Center Solar Siting Process

Siting solar for a data center is not the same as siting solar for a utility PPA. A utility optimizes for the cheapest delivered megawatt-hour. A data center optimizes for hourly supply to a specific meter, under a fixed construction deadline.

That changes the order of operations: standard prospecting starts with land and checks the grid later. Data center solar siting starts with the load and works backward to the land. This guide uses a six-stage process.

Stage 1: Define the load profile. Get the data center’s expected load in megawatts, its annual energy in megawatt-hours, and its hourly shape. Ask for the power usage effectiveness (PUE) assumption, because meter load is IT load multiplied by PUE. A 50 MW IT load at a PUE of 1.3 is a 65 MW meter load.

Stage 2: Pick the market. Choose the ISO/RTO or bilateral market based on the data center’s location, queue depth, and renewable procurement rules. ERCOT, PJM and MISO are active US markets. Each has different queue mechanics, deliverability rules, and congestion patterns.

Stage 3: Screen land. Run GIS constraints over the candidate region: slope, land cover, wetlands, flood zones, protected habitat, and setbacks. Overlay parcel data and transmission. The output is a ranked shortlist of sites, not a single winner.

Stage 4: Analyze interconnection. For each shortlisted site, evaluate the nearest substations and lines: available injection capacity, existing queue congestion, and likely network upgrade costs. Kill sites with fatal flaws here, before lease money goes out.

Stage 5: Design and load-match. Model hourly solar production against the data center’s hourly load. Size the array, the storage, and the grid contract to hit the carbon-free energy target at the lowest cost.

Stage 6: Model the finances. Build the pro forma: capital cost, PPA revenue, tax treatment, curtailment risk, and sensitivity cases. This is where the deal structure, physical PPA, virtual PPA, or self-build, gets decided.

Stages 3 and 4 run in parallel in practice. A site with perfect land and a dead interconnection path is worthless, and the reverse is rare. Budget several months for stages 1 to 4 on a serious prospect.

One note on division of labor: the data center owner usually handles grid power procurement at the facility meter. The developer or EPC handles the solar side siting. The two teams must share load data early, or the solar design will be sized against the wrong number.

Assemble the siting team before stage 3. You need a GIS analyst, a transmission engineer, a land agent, an environmental consultant, and a financial modeler.

Smaller EPCs cover these roles with 2 or 3 people plus outside consultants, but the role you cannot skip is the transmission engineer. Every other workstream depends on the interconnection answer.

Land Screening: GIS, Zoning, and Environmental Constraints

Start with the acreage math. Land need depends on module efficiency, tracking versus fixed tilt, row spacing and setbacks. Plan on about 5 to 6 acres per MW as a screening default, so a 100 MW farm needs roughly 500 to 600 acres, and replace the default with a real layout.

Scale that to data center load. A facility drawing 100 MW on average can imply several hundred MW of solar for high hourly coverage, because solar runs at a fraction of nameplate. At the screening default, 300 to 500 MW is 1,500 to 3,000 acres, often split across two or three sites.

GIS screening converts a region into a ranked site list. The standard workflow applies exclusion layers first, then scores what remains. Exclusion layers remove land that cannot be permitted or built; scoring layers rank the survivors by cost and risk.

Transmission proximity deserves its own pass. The gen-tie, the line from the project to the point of interconnection, is the project’s responsibility. Every mile of gen-tie adds cost, easement negotiations, and schedule.

As a screening heuristic, treat a few miles as a ceiling for most projects, and keep the comfort zone to a mile or two. Beyond that, the easement count and the per-mile cost usually overwhelm whatever the site saved on land. Measure corridor distance along realistic routing, not as the crow flies.

Zoning is the least technical criterion and the most political. County solar ordinances now exist in most active markets, and several states have seen moratoria.

Check the county comprehensive plan and any pending ordinance amendments before spending on surveys. A site that passes every GIS layer can still die in a planning commission meeting. Local opposition most often forms around visual impact, drainage, and the loss of farm income.

Weight the scoring layers by deal impact, not by data availability. One example weighting gives transmission distance and queue congestion 40% of the score, land assembly and zoning 35%, and environmental and physical factors 25%. Adjust the weights per market, in ERCOT, congestion dominates; in the Southeast, zoning and wetlands dominate.

Community engagement belongs in this stage too, not after permitting starts. Counties that feel surprised by a solar application respond with moratoria. A 30-minute briefing with the county planner and letters to adjacent landowners cost a week and can save quarters.

Land control follows screening. Developers typically secure option agreements, which are multi-year rights to lease or purchase, rather than buying land outright. Options keep capital light during the interconnection wait. Typical option terms run for years, so match the term to the expected study timeline.

Expect to negotiate with multiple landowners per site. Per-acre option payments can climb in counties with good grid access, where developers bid against each other. Walk away from parcels where 1 holdout owner controls the only viable corridor to the substation.

Environmental due diligence runs in parallel with land control. Wetland delineation, habitat surveys and cultural resource studies can take several seasons. Start them as soon as the option is signed, not after the interconnection study returns.

Here is the contrarian point on land: the cheapest acre is often the most expensive megawatt. Remote land trades at a discount precisely because it lacks transmission access. Teams that screen by land price first end up with cheap options, long gen-ties, and network upgrade bills that erase the savings.

Interconnection and Transmission Capacity Analysis

Interconnection is where data center solar projects live or die. LBNL’s Queued Up interconnection queue analysis has tracked more than 2 TW of generation and storage capacity in US queues in recent editions. Check the latest edition for the current total and for completion rates.

Time from queue entry to commercial operation has stretched to several years in many markets, and only a fraction of queued projects reach construction. The rest withdraw, often after a study assigns them upgrade costs they cannot carry.

The point of interconnection (POI) decision drives everything downstream. The POI is the substation or line tap where the project injects power. Its available capacity, voltage class, and existing queue congestion determine the study outcome and the upgrade bill.

Read the queue like a credit report. Every ISO publishes queue data: requested megawatts per substation, study phase, and withdrawal history.

A substation with 800 MW of active requests ahead of you on a 200 MW transformer is a warning, not an invitation. High withdrawal rates in a cluster usually mean the upgrade costs came back ugly. Healthy substations show a mix of early-phase requests and a few projects actually reaching operation.

Queue reform is slowly improving the picture. FERC Order 2023 moved the large ISOs from serial first-come studies to cluster studies, with readiness deposits and withdrawal penalties. Transition clusters are still working through backlogs, so treat announced study timelines as optimistic and ask for the actual history of recent clusters.

Deliverability deserves a definition. Energy resource interconnection service (ERIS) lets the project inject energy when the grid has room. Network resource interconnection service (NRIS) adds deliverability, the project counts toward capacity obligations and can sell capacity.

Data center PPAs increasingly expect capacity value. NRIS or its market equivalent is usually worth the higher study cost. Confirm the deliverability status of the POI before assuming any capacity revenue in the model.

Network upgrade costs are the silent killer. A cluster study can assign a project its share of a new 345 kV line or a substation rebuild.

Large upgrade allocations can break a pro forma. Screen recent study results in the target ISO before choosing a POI. Public study reports show what comparable projects were actually charged.

Three strategies shorten the path. First, target substations with announced transmission expansions and reserve queue position early.

Second, consider co-location with an existing or retiring generator, where injection capacity already exists. Third, evaluate behind-the-meter (BTM) configurations at the data center itself, where the solar farm connects on the load side and avoids the generation queue entirely.

BTM deserves a caveat. On-site or adjacent solar avoids queue delay and delivers power with no transmission charge.

But land near data center campuses is expensive, and parcel sizes limit capacity. The data center’s construction schedule rarely waits for solar either. Most large deals end up as remote generation with a financial PPA, not a physical wire.

Submit the interconnection request as early as site control allows. Queue position is a real asset in every ISO. A defensible, early request at a well-chosen POI is worth more than a perfect site with a late queue number.

Load Matching: Solar Production vs. Data Center Consumption

Data center load is nearly flat. A well-utilized facility runs at a high share of capacity around the clock, every day of the year. AI training loads add short-duration ramps, but the daily profile has no evening dip and no weekend trough.

Solar production is a bell curve that exists only during daylight hours and shifts with season and weather. Solar’s capacity factor, meaning actual annual energy divided by maximum possible energy, is lower for fixed tilt than for single-axis tracking. Use the capacity factor from a site simulation and not a rule of thumb.

The arithmetic follows from those two shapes. As an illustration, a 100 MW array at an assumed 28% capacity factor produces about 245 GWh a year (100 MW x 8,760 hours x 0.28). A 100 MW data center at an assumed 90% utilization uses about 788 GWh a year, so annual energy matching alone needs roughly three times the solar capacity.

Hourly matching is stricter. The industry measures it with a carbon-free energy (CFE) score: the share of consumption hours in a year where carbon-free supply meets or exceeds load.

A solar-only portfolio is limited by daylight, so its hourly CFE score stays well below its annual energy coverage. Higher scores require storage and oversizing, and the final stretch to 100% is where costs escalate.

Work the problem in this order. First, pull 8,760 hours of load data, actual interval data if the facility exists, or a modeled profile from IT capacity and PUE if it does not.

Second, generate an 8,760-hour solar production profile for each candidate site using satellite irradiance data. Third, test array sizes and storage sizes against the CFE target and the budget. Run at least a P50 and a P90 weather year so the buyer sees the bad-year picture.

Seasonality cuts against solar in most US markets. Summer days deliver much more energy than winter days, while data center load stays flat. Check the monthly breakdown, not just the annual average, winter hours dominate the uncovered remainder.

Oversizing is the cheapest lever. Moving from 1.5x to 2.5x annual energy coverage raises the CFE score materially, because the array covers shoulder hours and cloudy days more often.

The cost is curtailment, midday hours where production exceeds what the grid or the battery can absorb. Model the trade explicitly. High curtailment usually means the array is too big or the battery is too small.

Geographic diversification helps more than most teams expect. Two sites 100 miles apart have meaningfully different cloud patterns. A portfolio of 2-3 solar farms in one ISO delivers a smoother combined profile than 1 large farm, at the cost of more development work.

One modeling discipline matters above all: use time-synchronous data. Solar production and load must be modeled on the same timestamps, in the same time zone, with daylight saving handled identically. A one-hour offset between the two series is enough to inflate coincidence numbers and undermine the study in front of the buyer.

SurgePV’s generation and financial tool covers energy-yield and financial modeling for the production side of this analysis. Pair that output with the facility’s interval data in a spreadsheet or portfolio model before committing to an array size, and check the result against your own hourly calculation.

Battery Storage and Firming Resources

Solar alone cannot power a data center 24/7. That is physics, not a temporary technology gap.

Nighttime load must come from storage, the grid, or another generator. The design question is how much of each, and at what cost. The answer changes every year as storage prices fall.

Four-hour lithium-ion storage is the default firming resource. It absorbs the midday solar peak and discharges through the evening.

Pairing solar with 4-hour batteries lifts hourly CFE materially, by an amount that depends on oversizing, load shape and weather, so quantify it with an hourly model. Our solar storage design guide covers the sizing math in detail, and the same logic scales from C&I to utility projects.

Plan augmentation from day one. A battery that cycles daily loses usable capacity over time, so a long PPA needs added battery modules (augmentation) to hold contracted output. Contracts increasingly specify augmentation obligations explicitly, price them in the pro forma, or the battery runs short in later years.

Now the contrarian part: chasing 100% hourly CFE from day one is usually a mistake. The last stretch of hours, such as multi-day cloudy periods and winter lows, costs disproportionately more to cover than the first 85% or so.

Long-duration storage or firm clean generation can close the gap, but prices and availability remain immature. Many buyers accept a high but not perfect CFE target with contractual commitments to improve. Paying for 100% on day one usually means overbuilding solar and storage that sits idle most of the year.

Annual matching is a legitimate alternative. Under annual matching, the buyer procures enough renewable energy to cover 100% of annual consumption, sells back the midday excess, and buys grid power at night. It is cheaper, it uses proven contract structures, and it still funds new carbon-free capacity.

The criticism, that night power may come from fossil generation, is real. But the marginal abatement cost of the last CFE percent often exceeds what the same money would save elsewhere in the portfolio. Buyers should pick the metric deliberately, not by default.

The practical firming stack for a data center solar project has four layers. Solar covers daytime hours. Batteries shift 3-5 hours of solar into the evening peak.

A grid supply contract or a retail tariff covers overnight load. Renewable energy certificates or a settlement true-up handle the annual accounting. Each layer has a price, and the pro forma should show the marginal cost of each CFE point gained.

Emerging options sit on the horizon, including longer-duration lithium systems, multi-day chemistries, geothermal PPAs and small modular nuclear. Watch them, but be careful about sizing a 2026 financial model around them until lenders accept them.

Size storage against the load shape, not against a rule of thumb. A data center with flat 24/7 load uses the battery every night, high cycle counts, fast degradation, strong utilization. That is a different storage business case than a peaking application, and it deserves its own degradation and augmentation model.

Financial Modeling for Data Center Solar

The pro forma for a data center solar project has four layers: capital cost, revenue, tax and risk. Each layer has inputs specific to this deal type, and each deserves a sensitivity case.

Capital cost comes first. Utility-scale solar capital cost varies by market, year and scope, so take current figures from recent benchmark reports and EPC quotes and not from an article. Add the cost of storage, plus gen-tie and network upgrades as project-specific line items. Land appears as lease escalators or purchase cost, and rates vary widely by county.

Revenue comes from the power purchase agreement (PPA). Data center and corporate PPAs use three common structures: fixed price with an escalator, hub-settled virtual PPA, and physical sleeve through a utility. Merchant exposure after the PPA term, the tail, belongs in the model at conservative hub prices. Lenders will haircut the tail regardless of your forecast.

Tax treatment is the commercial side of the story. The technology-neutral clean electricity investment credit under Section 48E applies to commercial solar and storage, subject to the phase-out and commencement schedules in current law.

The residential credit under Section 25D ended for expenditures after December 31, 2025. Do not confuse the two in front of a buyer. Commercial projects also use MACRS depreciation, and credits are transferable. Transferability simplifies financing for developers without tax appetite.

Three sensitivities tend to break deals. Network upgrade overruns top the list, model the pessimistic study outcome, not the median.

Curtailment comes second, because data center PPAs with strict delivery shapes penalize both over- and under-delivery. Third is schedule. A long delay against a data center’s go-live date can trigger termination clauses.

Do not skip the probability-weighted production cases. Lenders size debt on P90 production, and buyers price PPAs on P50. A model that shows only P50 flatters the project and sets up a painful renegotiation later.

Run the model hourly if the PPA has shape requirements. Annual-average models hide the hours where the portfolio under-delivers and buys back at peak prices. The hourly production profile from your solar design software feeds directly into revenue certainty, the same 8,760 data the buyer’s analysts will demand.

Deal structure is the last modeling decision. The table above uses assumed values for illustration. Self-build gives the data center owner full control and full risk.

A developer-build with a PPA shifts construction risk but adds margin. Community or portfolio PPAs spread load across several projects. Each structure changes who owns the siting mistakes, price accordingly.

Worked Example: A 200 MW Data Center Solar Project (Hypothetical)

This example is hypothetical. It was built to show the workflow and the arithmetic, not to describe a real client. Every input is an assumption, and the numbers are rounded for readability.

The buyer: a colocation operator building a campus with 60 MW of IT load. At a PUE of 1.25, meter load is 75 MW average, which is about 657 GWh a year at 24/7 operation (75 MW x 8,760 hours). The buyer wants a high hourly carbon-free score and a 15-year fixed-price PPA.

Stage 3 screening covers a four-county region. Suppose GIS exclusions remove most of the land area for slope, playas and floodplain. A transmission overlay then cuts the shortlist to six parcels within about two miles of high-voltage substations. Queue analysis rejects four of the six: two sit behind congested clusters and two have recent upgrade allocations that are too large.

The winning site in this example: 1,400 acres of former cropland, 1.2 miles from a substation with visible headroom and one late-phase project ahead in the queue. Assumed option terms are $650 per acre per year on a four-year option that converts to a lease at notice to proceed. Wetland and habitat surveys come back clean within two seasons.

The design settles on 200 MW DC with single-axis tracking at an assumed 27% capacity factor, which gives about 473 GWh a year (200 MW x 8,760 hours x 0.27). That is about 72% of the buyer’s annual energy (473 / 657). The project alone cannot reach a high hourly CFE score at that coverage, so a 100 MW / 400 MWh battery shifts midday output into the evening, and the buyer closes the remaining gap with grid supply contracts, a second solar phase or other carbon-free sources.

A quick financial check on these assumptions. Year-one PPA revenue on delivered energy is about $25.5M (473,000 MWh x $54). Against about $343M of capital ($205M + $120M + $18M), that is a simple revenue-to-capital ratio of about 7.4% before operating costs, tax credits and any storage revenue. That ratio is why the tax credit, the network upgrade bill and the storage revenue decide whether a project like this clears a developer’s hurdle rate.

A sensitivity run should push upgrades 40% over estimate, curtailment to double its base value and the PPA price 10% lower, and then check the result against the hurdle rate.

Here is the lesson that repeats in deals of this kind. Everyone models the array, the battery and the PPA. Fewer teams model the gen-tie corridor properly. A corridor can cross a second landowner, a pipeline easement and a drainage district, which creates three title problems after the interconnection study has started. Run title and easement screening on the gen-tie path at the same time as the main parcel, and treat corridor risk as a site-selection criterion.

The second lesson is that the buyer’s load forecast can move during development. Design for the contracted capacity and not the aspiration, and price expansion options separately.

Common Siting Mistakes That Kill Data Center Solar Projects

The same failures repeat across teams entering this vertical. Most are process errors, not engineering errors. Here are the seven that cost the most.

1. Screening by land price instead of delivered cost. Cheap remote land carries long gen-ties and heavy network upgrades. Rank sites by levelized delivered cost to the data center’s market node, not by option price per acre.

2. Ignoring the queue until after lease money. Options signed before queue analysis are lottery tickets. ISO queue data is public, so pull it first. A site behind several congested projects at the same substation is not a site; it is a waiting room.

3. Trusting the headline load number. The buyer says 100 MW; the contracted phase one is 40 MW; the meter load at PUE 1.3 is different again. Model against the contracted, metered number, in writing.

4. Sizing for annual matching while promising hourly matching. Annual energy coverage of 100% delivers a much lower hourly CFE score. If the buyer’s marketing says 24/7 carbon-free, an annual-matched design creates a disclosure problem, so align the metric before sizing the array.

5. Underestimating the schedule stack. Interconnection study, environmental surveys, county permitting, and equipment lead times run sequentially more often than teams plan. A realistic timeline pitched as a shorter one destroys trust when the first milestone slips.

6. Skipping gen-tie and easement diligence. The corridor to the POI crosses land you do not control. Title work on the corridor belongs in stage 3, alongside the main parcel, not after the interconnection agreement.

7. Treating curtailment as free. Oversizing raises CFE scores, but unsold excess energy at negative or zero prices drags the IRR. Model curtailment revenue hour by hour, and do not assume the market absorbs everything at the hub price.

A pattern connects these mistakes: each one optimizes a single variable in isolation. Data center solar siting is a multi-constraint problem where interconnection, land, load shape, and contract structure interact. The teams that win run the stages as one workflow, with the same model feeding engineering and finance.

Conclusion

Data center solar siting is land screening, grid analysis, load math, and deal structure, run as one process. The demand side is not speculative. DOE’s summary of the LBNL report projects data center use rising to 6.7% to 12% of US electricity by 2028, and corporate buyers need carbon-free supply on fixed deadlines.

The teams that capture this work share three habits. They start from the load and work backward to the land. They treat interconnection queue position as the scarcest resource in the project, and they model hourly because the buyer’s analysts will.

Solar covers the daytime; storage and grid contracts close the rest. Projects that pencil often accept a high but imperfect hourly carbon-free score now and contract a path upward. Chasing perfection on day one prices you out of deals that good engineering would win.

For EPCs moving up from residential and commercial solar work, the workflow is familiar at a larger scale. Site assessment, production modeling, financial analysis, and a committee-ready proposal, the discipline is the same, and the stakes are bigger. SurgePV’s scope covers shadow analysis, energy-yield and financial modeling, and proposal generation for that workflow.

The same logic applies beyond the US. Data center growth in Ireland, Singapore, India, and the Gulf runs into the same constraints: land, transmission, and hourly carbon-free supply. Some European data center hubs have already seen what happens when load growth outruns the grid: connection limits and long delays.

The screening criteria in this guide travel well; only the data sources change. Developers who bring disciplined siting to those markets early will hold the queue positions everyone else wants later.

Pick a market, pull the queue data, and screen your first region this quarter. Start with the buyer’s contracted load, not with a land listing. Run the load-matching math before you commit to a parcel. The interconnection clock rewards whoever files first. Want to see a candidate site modeled end to end? Book a SurgePV demo.

Frequently Asked Questions

What is data center solar siting?

Data center solar siting is the process of finding land and grid capacity to build solar that serves a data center. It combines GIS land screening, interconnection queue analysis, hourly load matching and environmental due diligence, and it starts from the data center’s load and works back to the land.

Why do data centers need dedicated solar farms?

Data centers run around the clock, and dedicated solar paired with storage or grid contracts can supply carbon-free energy and help meet goals such as RE100. RE100 members commit to 100% renewable electricity by 2050 at the latest, and dedicated projects add new generation instead of only buying certificates.

How much land does a data center solar farm need?

As a screening default, plan on about 5 to 6 acres per MW, so a 100 MW farm needs roughly 500 to 600 acres. Module efficiency, tracking, row spacing and setbacks change the figure, so replace the default with a real layout. A large data center load can imply several hundred MW of solar across more than one site.

What is the biggest challenge in data center solar siting?

Interconnection is usually the biggest challenge. US queues hold more than 2 TW of generation and storage in recent LBNL editions, and studies can assign large network upgrade costs. Screen for substation headroom and recent upgrade costs, and file the interconnection request as early as site control allows.

Can solar power a data center 24/7?

Solar alone cannot power a data center 24/7 because it produces only during daylight hours. It needs battery storage, a grid supply contract or other firm resources for the rest. Solar covers daytime load, and the design question is how much storage and grid supply to add.

What software is used for data center solar siting?

Teams combine GIS tools such as ArcGIS, solar prospecting platforms, interconnection data from ISO and RTO queues, and financial models. Hourly production and load data should sit in one model. Check any tool’s own documentation for what it supports.