In short
AI data center construction is booming. Projects now run into the tens of billions of dollars. Standard construction insurance policies were not designed for the power density, battery hazards, and scale of these facilities. New multi-billion-dollar products from Zurich, Aon, and Marsh now offer tailor-made coverage. Yet coverage gaps remain around cyber physical damage, pollution, and phased occupancy disputes. Insurance can consume up to 10 percent of a project budget, so getting it right is a hard-dollar decision for developers, general contractors, and their legal teams.
The average data center project value reached $633 million in 2025, a 70 percent increase over the year before. Gallagher, Apr. 2026 Builder’s risk premiums on a $500 million build typically range from $750,000 to $2.5 million a year. Hotaling Insurance, 2026 Delay in startup coverage, which pays lost revenue when a physical loss pushes back the go-live date, can cost two to three times the base builder’s risk rate. Hotaling Insurance, 2026 Owner-controlled insurance programs, which bundle coverage for all contractors on site, can trim total insurance costs by 10 to 20 percent. Hotaling Insurance, 2026
We cover the coverages, the risks that set AI data centers apart, the new specialty products, and the gaps that still keep insurers and borrowers up at night.
What types of insurance cover an AI data center during construction?
Several lines of insurance work together during the building phase. For a project of any size, the core pieces are a builder’s risk policy, delay in startup coverage, general liability, excess casualty, and some form of surety. Large campuses often add a wrap-up program and sometimes subcontractor default insurance.
Builder’s risk insurance
Builder’s risk, also called course of construction insurance, is a property policy that covers direct physical loss or damage to the project while it is being built. It protects the structure, building materials, equipment, and machinery on site. The coverage is temporary. It starts on the project start date and ends when the building is finished or a set date passes. AGC 2026 White Paper at 6
Most builder’s risk policies are written on an all risk basis. They cover every cause of loss except those listed in the exclusions. The policy can also be written as a named peril form, but all-risk is far more common for AI data centers.
A standard builder’s risk policy has a defective workmanship exclusion. That exclusion bars coverage for the cost of fixing faulty or inadequate work, including a subcontractor’s mistakes. Courts treat negligent or poor judgment work the same as truly defective work. Most policies include an ensuing loss provision, which restores coverage for damage that results from the defective work. The insurer still does not pay to redo the bad work itself, but it pays for the harm that the bad work caused. AGC 2026 White Paper at 6-7
The scale of AI data centers makes builder’s risk a major budget line. Premiums in the market range from $0.15 to $0.50 for every $100 of project value. A $500 million building would pay $750,000 to $2.5 million in annual premium. Hotaling Insurance, 2026 Total insurance costs can reach 10 percent of the entire project budget, or $200 million on a $2 billion campus. Bisnow, cited in Hotaling
Delay in startup and advance loss of profits
Delay in startup coverage, often called DSU or advance loss of profits, sits inside the builder’s risk program. It pays for lost revenue or rental income when a covered physical loss causes a construction delay that pushes back the date the facility can start earning money.
DSU is not a trivial add on. DSU limits for major AI data center projects typically run from $500 million to $1.5 billion, and the premium rate is often two to three times the base builder’s risk rate. Hotaling Insurance, 2026 Lenders and equity investors look hard at DSU because their return depends on the facility going live on schedule.
A concrete example is the Stargate Phase 2 project in Abilene, Texas. JPMorgan structured a $7.1 billion construction loan around specific delivery milestones. The DSU coverage on that project is tied directly to meeting those milestones. Hotaling Insurance, 2026
General liability and excess casualty
A commercial general liability policy, or CGL, covers third party claims for bodily injury, property damage, or personal and advertising injury that arise out of construction operations. Every contractor on site carries a CGL policy either individually or through a wrap-up.
Several standard CGL exclusions matter more for AI data centers than for a typical office build. The contractual liability exclusion (Exclusion b) can limit coverage for liability the GC assumes under the prime contract. The care, custody, or control exclusion (Exclusion j(4)) applies to personal property in the contractor’s care. Whether servers and IT equipment already inside a powered shell fall under that exclusion is an open debate. GL policies typically are subject to exclusions for pollution related claims, which can extend to diesel emissions from power equipment or PFAs from two phase cooling systems. Data Center Knowledge
Excess casualty insurance sits above the CGL and provides additional limits. For a large AI data center project, standard contracts commonly require excess or umbrella limits of $10 million to $25 million. Hotaling Insurance, 2026
An important endorsement for AI data centers is the completed operations extension. It extends CGL coverage beyond the policy period for construction defect claims that arise after the project is finished. In an AI data center, cooling, electrical, and technical systems may not fail until the facility operates under full load. For AI data center construction, the insured should ensure the completed operations extension fully addresses CGL exposure for the entire statute of repose, which in some states is as long as 10 years. AGC 2026 White Paper at 9, FindLaw 50-state survey, AGC 2026 White Paper at 9, FindLaw 50-state survey
Wrap-up programs, OCIP and CCIP
A wrap-up is a single insurance program that covers all the contractors and subcontractors on a project for general liability and workers’ compensation. It is purchased by either the owner (an Owner Controlled Insurance Program, OCIP) or the general contractor (a Contractor Controlled Insurance Program, CCIP).
For an AI data center, an OCIP gives the owner important advantages. The owner sets the safety standards and manages claims centrally. If the GC leaves the job, the OCIP stays with the owner and the coverage does not break. Tax alert A CCIP, by contrast, is sponsored by the GC. If the GC is terminated, the CCIP goes with them, forcing the owner to scramble for new coverage in the middle of construction. Reed Smith podcast
The cost benefit of an OCIP is real. Campus scale program premiums range from $10 million to over $50 million, but the consolidated program typically saves 10 to 20 percent compared with having every contractor buy a separate policy. Hotaling Insurance, 2026 One real example is a $315 million OCIP placed by Shepherd Insurance for an AI data center developer in Virginia. The program provided a $2 million primary general liability limit plus $10 million in excess coverage over a five year term. The developer saved 15 percent on premium, a $76,000 reduction, because the insurer gave a credit for the developer’s use of Procore technology for safety and quality control. Shepherd Insurance case study
Subcontractor default insurance
Subcontractor default insurance, or SDI, is a two-party agreement between the contractor and the insurer. It pays out if a subcontractor defaults. Unlike a performance bond, which is a three party instrument involving a surety, the GC itself prequalifies the subcontractor rather than relying on a bonding company. SDI coverage can go beyond the subcontract value up to the policy limit, which gives more flexibility. The catch is that SDI market capacity is limited for the mega-scale subcontracts found on AI data center projects. Placing enough SDI often requires several carriers. AGC 2026 White Paper at 15
Performance and payment bonds
A surety bond is a three-party guarantee. The surety promises the owner that the contractor will perform the contract and pay its subcontractors and suppliers. Under the federal Miller Act, construction projects on federal land must carry performance and payment bonds. The Miller Act requires a performance bond in an amount the contracting officer considers adequate and a payment bond equal to the total amount payable by the contract terms. 40 U.S.C. § 3131 Under the implementing FAR clause, both bonds must be at 100 percent of the original contract price for contracts above the threshold. FAR 52.228-15 Private lenders on AI data center projects commonly impose the same requirement by contract.
The surety market for AI data centers has grown dramatically. Bonds no longer cover only the prime contract. Interconnection bonds guarantee payment for the substations and transformers needed to connect an AI data center to the grid. PPA bonds secure obligations under long-term power purchase agreements. A surety bond also works as off balance sheet credit, which helps the developer keep its bank lines open. WTW
One broker, WTW, supports surety syndications on AI data center developments ranging from $5 million up to $500 million. WTW, June 9, 2025 That scale reflects the size of today’s projects.
What makes AI data center construction riskier than a standard commercial building?
Five features distinguish an AI data center from a run of the mill office or warehouse project. Each feature drives a bigger insurance bill and more complex underwriting.
Power density and electrical configuration
An AI server rack can demand more than 100 kilowatts, compared with 5 to 15 kilowatts for a traditional server rack. That is a 7 to 20 fold increase in power density. In some configurations, the increase can reach 50 times. CRC Group, Apr. 2026
Power supply is also the largest single cause of business interruption. The Uptime Institute found that power issues caused 45 percent of impactful AI data center outages. Uptime Institute Global Data Center Survey 2025 A construction error in the electrical system can create a claim that is both a property damage loss and a massive delay in startup loss.
Lithium-ion battery fire hazard
Lithium-ion batteries are the standard backup power source for AI data center server halls. They are also a severe fire risk. An FM Global 15-year study found that fire accounts for only 10.9 percent of AI data center loss events but drives 42.3 percent of loss costs. SWISS Re sigma 07/2026
Two recent fires illustrate the danger. In September 2025, a lithium-ion battery backup unit failed during maintenance at the South Korea NIRS data center. The resulting explosion and fire crippled 647 government services, including emergency response and tax systems. SWISS Re sigma 07/2026 In September 2024, a lithium-ion battery thermal runaway at a Digital Realty data center in Singapore burned for over 36 hours and heavily disrupted Alibaba Cloud services. SWISS Re sigma 07/2026
Insurers are reacting. In its January 2026 loss prevention guidance, FM Global added a two hour fire resistance wall rating for compartmentation between multiple data processing equipment rooms and introduced more stringent sprinkler expectations for AI data rooms. FM Global DS 5-32, Jan. 2026, SWISS Re sigma 07/2026 Many insurers are now requiring compliance with NFPA 855, the standard for stationary energy storage systems, as a condition of coverage. NFPA 855 (2026 ed.), Insurance brokerage analysis
Liquid cooling and water risk
Liquid cooling systems move heat away from dense server racks. They also introduce new property damage and business interruption exposures. FM Global’s data review found that liquid-related losses accounted for nearly 24 percent of total AI data center loss costs. Fire-related sprinkler leakage made up 9.3 percent, and escaped liquid from cooling systems accounted for 10 percent. FM Global DS 5-32, Jan. 2026, SWISS Re sigma 07/2026
Water supply is a second, separate risk. A large AI data center can consume up to 5 million gallons of water per day for cooling. Verisk Maplecroft projects that 52 percent of global AI data center hubs may face high or very high water stress by 2030. CRC Group, Apr. 2026 A facility that cannot secure enough water may be forced to pause or relocate, events that a standard builder’s risk policy would not cover.
Long lead time equipment and supply chain
Transformers now carry a procurement lead time of more than one year. GPUs can take months or even years to arrive. AGC 2026 White Paper at 4 A delay in any single critical piece can push out the completion date by months. That makes expanded transit and offsite storage limits essential. Standard builder’s risk forms cover materials once they reach the site. An AI data center policy often needs to cover equipment while it is still at the manufacturer or crossing the ocean.
The value shift from structure to contents compounds the problem. In many AI data centers, the GPUs, servers, and networking hardware cost 1.5 to 2 times as much as the building itself. Aon, Building Bankable Data Centers A larger share of the total insurable value sits inside the equipment, not the walls.
Phased occupation and powered shell construction
Many AI data centers are built in phases. Tenants move servers into one portion while construction continues in another. That arrangement, called powered-shell construction, creates dust, vibration, temperature swings, and equipment contamination risks that are heightened compared to traditional construction because sensitive AI data center equipment may be operating in one area while construction continues in another. AGC 2026 White Paper at 2-4
Phased occupation also creates a coverage question of when does the builder’s risk policy terminate? Courts have split. An Arkansas court found that partial occupancy of more than 40 percent of a fertilizer plant triggered completion and ended the builder’s risk policy. A Michigan court held that partial occupancy of a few apartments did not trigger completion for the larger project. AGC 2026 White Paper at 7, 15 An AI data center that is partly occupied could find itself in a coverage fight with no clear answer.
Beyond the physical risks, AI data center contracts impose performance criteria including continuous data processing, storage capacity, internet speed, power consumption, cooling output, ambient temperature, and humidity limits. Liquidated damages attach if the facility fails to meet these standards. A structurally sound building can still generate large GC liability if the systems do not perform as specified. AGC 2026 White Paper at 4
Several other risk factors appear in today’s pipeline. Roughly 30 percent of planned U.S. AI data center capacity may include on site power generation, which adds new property and environmental exposures. SWISS Re sigma 07/2026 Over 25 percent of U.S. AI data center capacity may be in areas with three or more large hail days per year, and roughly 40 percent could sit in zones with three or more EF1+ tornado days annually, according to Swiss Re modelling. SWISS Re sigma 07/2026 Large AI data center sites can also draw community objections over natural resources and wildlife impacts, as a proposed $30 billion campus in Mooresville, North Carolina demonstrated. Charlotte Observer, Charlotte Observer
What specialized insurance products exist for AI data centers?
Recognizing the gaps in standard cover, major brokers and carriers have built purpose built facilities for AI data center construction. Three products dominate the market in 2026.
Zurich Data Center Project Guard. Zurich launched this first of its kind builder’s risk product in December 2025, with coverage available January 1, 2026. It is written on a non-admitted basis, meaning it is sold through surplus lines brokers rather than standard admitted carriers. The program requires a project value above $250 million and Zurich acting as lead insurer. Zurich has insured more than 250 data center projects across more than 20 states and over $350 billion in project value. Zurich NA
Project Guard adds several features not found in a standard builder’s risk form. It includes up to 12 months of operational property coverage with business interruption protection once construction ends. It embeds parametric weather coverage for the first year of construction, with rain and wind triggers included and cold, heat, snow, heat index, and air quality available as options. It expands transit and off site limits for high value equipment. It covers failure of climate control systems, temporary monitoring equipment, crisis management expense, and radioactive contamination. Zurich NA, Dec. 10, 2025, Zurich NA data center construction insurance
Aon Data Center Lifecycle Insurance Program (DCLP). Aon launched the DCLP in July 2025 as a multi-line facility. It has expanded twice. It reached $2.5 billion of capacity in January 2026 and $3.5 billion in April 2026. That $3.5 billion breaks out across several lines, including construction all risks and DSU up to $2.5 billion on a maximum foreseeable loss basis, first year property damage and business interruption up to $2.5 billion, cyber and technology E&O up to $400 million (including non damage DSU and service level agreement violations), third-party liability up to $100 million (excluding U.S. exposures), and project cargo up to $500 million. As of April 2026, the DCLP also extends coverage to existing operational AI data centers beyond the first year. Aon PRNewswire, Apr. 15, 2026
Marsh Nimbus. Marsh launched Nimbus in June 2025 for the United Kingdom and Europe. It expanded to $2.7 billion of capacity by January 2026 and now covers major AI data center construction projects in the U.S., Canada, Europe, Australia, and New Zealand. Nimbus combines construction all risks, DSU, property damage, and business interruption into a single facility. Insurance Journal, June 16, 2025
Other specialty tools. Aon’s reinsurance team has designed and placed the first ever data center specific reinsurance treaty, providing up to $5 billion of capital behind a single leading insurer. Aon CEO Greg Case, cited in Insurance Journal, Mar. 9, 2026 CRC Specialty offers an Insurisk Data Center product with up to $500 million of combined property and casualty capacity. CRC Group, Apr. 2026 Lockton has an SLA specific insurance product that covers penalties when an AI data center fails to meet contractual performance standards. DCD
The table below compares the three main facilities.
| Facility | Lead | Capacity | Key additions beyond standard coverage | Minimum project size |
|---|---|---|---|---|
| Data Center Project Guard | Zurich | Project-specific | 12 month ops tail, parametric weather, HVAC failure, crisis management | $250 million |
| Aon DCLP | Aon (facility) | $3.5 billion total | Cyber/Tech E&O, cargo, liability, operational DC extension | Not publicly stated |
| Marsh Nimbus | Marsh (facility) | $2.7 billion total | Single tower for CAR, DSU, PD, BI globally | Not publicly stated |
What coverage gaps and legal disputes keep coming up?
Even with tailored facilities, several coverage questions remain open. Most of them stem from the mismatch between policy language written for conventional buildings and the reality of a high tech AI data center.
Physical loss or damage. Property policies require a direct physical loss or damage to trigger coverage. Courts have split on whether loss of data or microscopic heat damage counts. Ingram found coverage for loss of programming information. Ashland Hospital found coverage for microscopic heat damage. CPU found no coverage for mere loss of use without physical damage. Tax alert When an AI data center suffers a server malfunction that corrupts data without leaving visible damage, the outcome of a claim may depend on which jurisdiction’s law applies.
Completion and phased occupancy. As noted above, when builder’s risk coverage ends on a partially completed and partially occupied AI data center is a live dispute. A policy that terminates on substantial completion could leave an entire phase without coverage if the insurer takes the position that partial occupancy equals completion.
Utility services exclusion. Many property policies exclude loss caused by off premises power failure. An AI data center that goes down because the grid fails may not have a covered property claim, even though the resulting business interruption is enormous. Law firm analysis
Cyber physical gap. A standard property policy carries broad cyber exclusions. A standalone cyber policy may not cover physical damage. An attacker who gains access to a building management system and causes servers to overheat could find the loss falling between the two policies. This silent cyber gap is a known problem that has not yet been solved by standard market forms. Munich Re, 2025, Lockton, 2021
AI exclusions in cyber policies. Some cyber insurers have begun adding specific AI exclusions. As generative AI tools become embedded in building control systems, a future loss caused by an AI-driven malfunction could face an explicit policy bar. Tax alert
Pollution exclusions. The CGL pollution exclusion may be invoked to deny coverage for diesel exhaust from backup generators or for PFAS releases from two-phase cooling systems. Insurers have also argued that noise and vibration are pollutants, which could bar coverage for nuisance claims from nearby communities. Law firm analysis
Period of restoration mismatch. Traditional business interruption coverage pays until the damaged property is repaired, or could have been repaired, for a reasonable period. For an AI data center, the economic harm can continue long after the physical repair is done because of customer churn, SLA penalties, and reputational damage. Standard BI language was not built for that dynamic. Law firm analysis
Contingent business interruption. AI data centers depend heavily on third-party power providers, cooling infrastructure, and network connectivity. Contingent BI coverage is often sublimited or subject to restrictive triggers. A disruption at a utility may fall outside the contingent BI grant for that reason. Law firm analysis, 2025
Equipment breakdown interplay. Equipment breakdown insurance and the property policy must be carefully aligned. If the definition of covered equipment does not extend to the specific generators, switchgear, cooling units, and servers in an AI data center, a loss can fall into an unintended gap. Law firm analysis, 2025
These gaps do not mean coverage is unavailable. They mean that for a large AI data center project, off the shelf policies are not enough. Every one of these points is something that must be addressed through manuscripted endorsements, dedicated sub-limits, or the specialty products described above.
How is the insurance market coping with projects that exceed $10 billion?
The investment figures are historically high. U.S. AI data center construction investment was $74 billion in 2024. Construction starts reached a historic $25.2 billion in January 2026 alone. Gallagher, Apr. 2026 Average project values climbed 70 percent year over year to $633 million in 2025. Gallagher, Apr. 2026 Capital spending by the five largest cloud service providers is forecast to exceed $600 billion in 2026, with roughly 75 percent tied directly to physical AI infrastructure. SWISS Re sigma 07/2026
No single insurer can absorb the risk of a $30 billion campus. A layered insurance program with multiple carriers is a practical necessity. That structure introduces an operational challenge. A single physical loss can trigger claims across several policies, each with its own wording and claims protocols. Tax alert
Global data center insurance premiums are expected to rise to $24.2 billion by 2030, up from $10.6 billion. SWISS Re sigma 07/2026 The narrower AI data center insurance market, covering specifically AI grade facilities, is estimated at $2.26 billion in 2025 and growing at a 16.3 percent compound annual rate toward $4.84 billion by 2030. Research and Markets, Feb. 2026
Capacity is growing, but retained risk remains large. Catastrophe deductibles in tornado and hurricane zones are 2 to 5 percent of project value. On a $2 billion facility, the owner retains $40 million to $100 million in risk before insurance pays a dollar. Hotaling Insurance, 2026 DSU limits, while reaching $1.5 billion on a major project, still leave a tail if the actual delay cost exceeds the limit.
On the surety side, capacity has kept pace. WTW supports bond syndications ranging from $5 million to $500 million for data center developments. WTW, June 9, 2025 The insurance and surety markets now treat AI data centers as a distinct asset class and are building the capital and policy structures to match.
Key takeaways
- Builder’s risk is the foundation. For an AI data center, a manuscripted builder’s risk policy with expanded transit, off-site, and equipment testing coverage is essential. Standard forms are designed for commercial buildings, not for high voltage, liquid cooled server dens.
- DSU limits must match the project’s revenue profile. A DSU shortfall can be just as damaging as a property damage shortfall. Lenders will require DSU to align with debt service and revenue milestones.
- Use an OCIP for large campuses. An OCIP centralizes control and can cut 10 to 20 percent off total insurance costs. The program stays with the owner no matter what happens to the GC.
- Address phased occupancy in the builder’s risk policy. For a powered-shell delivery or phased move-in, negotiate the completion trigger explicitly. Do not leave it up to a common law substantial use test.
- Review the pollution exclusion. Diesel generators and two-phase cooling chemicals can trigger a CGL pollution exclusion. Secure a manuscripted exception or a separate environmental policy.
- Close the cyber-physical gap. Property and cyber policies must be coordinated so that a cyberattack that causes physical equipment damage does not fall into a black hole of competing exclusions.
- Parametric weather and SLA insurance are now real products. They can fill specific coverage gaps around construction delay and performance failure.
- Surety bonds extend beyond the GC contract. Interconnection and PPA bonds are becoming standard on large AI data center deals. Budget for them early.
- Catastrophe deductibles can be enormous. In hail and tornado zones, the retained risk may be $40 million or more. Have a plan for it, whether through captive insurance, reserves, or parametric hedges.
- Specialty facilities are worth evaluating. Data Center Project Guard, Aon DCLP, and Marsh Nimbus each bring features that a traditional placement cannot match. A broker with AI data center experience should compare them against the project’s specific risk profile.
Frequently asked questions
Q:What is builder’s risk insurance?
A:Builder’s risk, also called course of construction insurance, is a temporary property policy that covers direct physical loss or damage to a project while it is being built. It protects the structure, materials, and equipment on site. AGC 2026 White Paper at 6
Q:How much does construction insurance cost for an AI data center?
A:Builder’s risk premiums range from $0.15 to $0.50 for each $100 of project value. A $500 million building would pay $750,000 to $2.5 million in annual premium. DSU coverage can cost two to three times the base builder’s risk rate. Total insurance costs can reach 10 percent of the project budget. Hotaling Insurance, 2026
Q:What is an OCIP and why is it used?
A:An Owner Controlled Insurance Program is a single policy bought by the owner that covers general liability and workers’ compensation for all contractors on the project. It saves 10 to 20 percent on total insurance costs, keeps the owner in control of claims, and stays with the owner if the GC leaves. Reed Smith podcast, Hotaling Insurance, 2026
Q:What is delay in startup insurance?
A:Delay in startup, or DSU, is coverage inside the builder’s risk program that pays for lost revenue when a covered physical loss causes a construction delay that pushes out the operational start date. IRMI
Q:What are the biggest fire risks in an AI data center?
A:Lithium-ion backup battery units pose the largest fire risk. Fire accounts for 10.9 percent of data center loss events but 42.3 percent of loss costs. Recent battery fires in South Korea and Singapore crippled government services and cloud operations. FM Global’s DS 5-32 (January 2026) recommends one hour fire rated interior walls for data processing equipment rooms and two-hour walls for compartmentation between multiple such rooms. Swiss Re sigma 07/2026, FM DS 5-32, Jan. 2026
Q:What is the Aon Data Center Lifecycle Insurance Program?
A:The Aon DCLP is a multi-line facility that combines construction all risks, DSU, property, business interruption, cyber and technology E&O, liability, and cargo coverage. As of April 2026, it provides up to $3.5 billion in total capacity and can cover an AI data center from construction through operations. Aon PRNewswire, Apr. 15, 2026
Q:Are there exclusions for cyber damage in construction policies?
A:Many property policies have broad cyber exclusions. Standalone cyber policies often exclude physical damage. This creates a gap where a cyberattack that causes equipment damage might not be covered by either policy. Specialty facilities like Aon DCLP include cyber property damage coverage to help close that gap. Bracewell, 2025, Aon, 2026
Q:What is parametric weather coverage?
A:Parametric insurance pays a pre-agreed amount when a specified weather trigger, such as a certain amount of rainfall or wind speed over a defined period, is met. It does not require proof of physical damage. For AI data center construction, it can provide quick cash to cover delay costs after a storm. Data Center Project Guard embeds parametric weather coverage for the first year of construction. Zurich NA
Q:Does the Miller Act apply to private AI data center projects?
A:The Miller Act requires performance and payment bonds on federal construction projects. Under the FAR, the default amount for each bond is 100 percent of the original contract price, subject to exceptions for both bonds. 48 C.F.R. § 28.102-2 It does not apply to private projects by its own terms, but private lenders on large AI data centers may require performance and payment bond coverage as a condition of the construction loan. SFAA, Protect your construction lending capital with surety bonds, Grit Insurance, Performance bonds for contractors, Law firm analysis
Q:What happens if an AI data center building is occupied before construction is finished?
A:Phased move-in creates a dispute risk over when the builder’s risk policy ends. Courts have reached different results on when partial occupancy triggers completion. Some insurers may offer endorsements or coverage extensions to maintain builder’s risk protection during phased occupancy, and contracts and policies must define when coverage terminates to avoid gaps for the parts still under construction. AGC 2026 White Paper at 7, 15
Subscribe to The Compute Law Brief
The Compute Law Brief is a free weekday newsletter on the law of AI infrastructure across tax, real estate, construction, power, and deals. The big US build markets and federal law. Three minutes a morning. No paywall, and no email gate to read the blog. Subscribe if you want it in your inbox.
Junde Liu, JD, LL.M. (Taxation) candidate at UF Law. Originally published on Compute Law Blog. This article is general information and does not constitute legal advice. Reading it does not create an attorney client relationship. The reader should not act on the basis of any content here without first consulting a licensed attorney in the relevant state. Last reviewed for accuracy May 23, 2026.