Rising Data Center Demand in the AI Era: Challenges Arising from Building Use under Japan’s Building Standards Act and Divergent Decisions by Designated Administrative Agencies

The growing adoption of generative AI and cloud services is rapidly increasing the importance of data centers capable of supporting computational workloads at far higher densities than conventional facilities. Although further development is expected across Japan, no unified interpretation has yet been established as to how data centers should be treated as a building use for the purposes of Article 48 of the Building Standards Act.

A data center generally comprises multiple functions, including server rooms, monitoring rooms, offices, power supply facilities, and cooling systems. Accordingly, when an application for confirmation is examined or conformity with the restrictions applicable to a use district is assessed, a central question arises as to whether the facility should be treated as an office, warehouse, factory, or another building use. Decisions may differ among designated administrative agencies, and in some areas, disputes over the legality of data center development—including the classification of the building use—have developed into litigation.

Drawing on international research, this article first outlines the basic structure of data centers and the factors driving future demand. It then examines their treatment under the Building Standards Act, the problems arising from divergent decisions among designated administrative agencies, and the regulatory and institutional reforms currently being considered by the national government. Finally, based on the findings of this analysis, it discusses the structural challenges that should be addressed from a medium- to long-term perspective.

What Is a Data Center?

A data center is a facility in which information and communications technology equipment, including servers and network devices, is installed on a concentrated basis to store, process, and transmit data. Many digital services used in everyday life—including email, websites, cloud services, and video-streaming platforms—are supported by data centers.

Although the siting of data centers has recently become the subject of legal disputes in Japan, data centers themselves are by no means new facilities. They originally developed as facilities for the secure operation of information systems owned by private companies and public authorities. Their primary functions were to store and distribute data used for email, business systems, websites, and other digital services. Accordingly, the conventional market was centered on enterprise data centers and regional colocation facilities operating with relatively stable electricity demand.

The rapid expansion of generative artificial intelligence, or AI, has, however, begun to transform the role of data centers. Training and running inference for generative AI require enormous computational capacity supported by large numbers of high-performance graphics processing units, or GPUs. As a result, data centers are evolving from facilities that primarily store and distribute information into essential infrastructure responsible for large-scale computational processing.

Recent studies have identified AI data centers as a new category of electricity demand with characteristics that differ from those of conventional data centers, including substantially higher power densities and rapid fluctuations in electrical load. The Japanese government has also begun to distinguish these facilities from conventional data centers by using the term “next-generation AI data centers.”

There is currently no uniform, cross-sector statutory definition of either a data center or a next-generation AI data center. Certain facilities may be characterized as buildings that accommodate and operate telecommunications facilities under the Telecommunications Business Act. In practice, however, high-performance GPUs, electrical receiving and transformation equipment, large-capacity storage batteries, and cooling systems operate as an integrated and inseparable system.

The electrolyte contained in lithium-ion batteries may also constitute a hazardous material under the Fire Service Act. These facilities therefore extend beyond a conventional telecommunications function and increasingly serve as social infrastructure supporting both electricity-intensive computation and information processing. For this reason, next-generation AI data centers need to be distinguished from conventional data centers and ordinary server rooms installed within other buildings.

This transformation also has major implications for architectural and facility planning. In conventional data centers, the server room was generally the central component of the building. In next-generation AI data centers, however, electrical receiving and transformation equipment, uninterruptible power supply systems, emergency generators, cooling systems, and telecommunications facilities are becoming substantially larger.

Planning such facilities therefore requires an integrated approach that extends beyond the architectural design of the building itself to include electricity supply capacity, cooling capacity, and the overall configuration of the infrastructure supporting the facility.

No internationally or domestically standardized system has yet been established for classifying data centers by scale. Existing overseas studies and public-sector materials instead use multiple indicators, including server-room area, total floor area, number of server racks, IT capacity, and electrical receiving capacity.

Based on previous studies and examples of data center developments in Japan and other countries, this article provisionally classifies data center scale as shown in Table 1.

Table 1. Indicative Classification of Data Centers by Scale

Category	Indicative server-room area	Indicative total floor area	Indicative power capacity	Principal uses and facility types
Micro / edge	Several m² to 100 m²	Several tens of m² to 500 m²	Several tens of kW to less than 1 MW	Telecommunications base stations, retail premises, factories, and geographically distributed edge processing
Small-scale	100–500 m²	500–2,000 m²	Up to several MW	Enterprise data centers and privately operated server facilities
Medium-scale	500–3,000 m²	2,000–10,000 m²	Several MW to 10 MW	Regional colocation facilities
Large-scale	3,000–10,000 m²	10,000–30,000 m²	10–50 MW	Large-scale colocation facilities
Hyperscale	More than 10,000 m²	More than 30,000 m², potentially exceeding 100,000 m²	Several tens of MW to 100 MW or more	Large-scale cloud computing, internet search, video streaming, generative AI, and related services
Campus-type reference category	Multiple buildings	More than 100,000 m² in total	More than 100 MW	Integrated AI and cloud-computing hubs

As Table 1 shows, conventional data centers and next-generation AI data centers differ substantially in both their functions and their physical scale. Hyperscale data centers currently being developed in Japan and overseas are no longer merely individual buildings. They are increasingly taking on the characteristics of urban infrastructure integrated with large-scale electricity supply facilities and telecommunications networks.

Depending on their location, however, these facilities may affect surrounding urban areas and regional infrastructure through their substantial electricity demand, noise generated by electrical receiving and transformation equipment and emergency generators, heat discharged by cooling systems, and the increasing physical scale of the buildings themselves.

From this perspective, although data centers are not currently subject to Article 51 of the Building Standards Act, some large-scale facilities may be approaching the characteristics of facilities that require urban-planning-based coordination of their location, rather than being assessed solely through individual building confirmation procedures.

Article 51 requires locational coordination for facilities such as wholesale markets, crematoria, and waste-treatment facilities. In a similar way, the siting of a large-scale AI data center may need to be considered not only as a question of whether an individual building complies with building regulations, but also as a broader matter of land-use planning and urban infrastructure.

Why Data Center Demand Is Rising Rapidly in the AI Era—and How Future Demand May Develop

Demand for data centers has steadily expanded alongside the growth of the internet and cloud services. Recent increases in demand, however, cannot be explained solely by larger volumes of stored data or network traffic. The rapid adoption of generative AI is creating a need for enormous computational capacity, fundamentally changing both the physical scale of data center facilities and their electricity requirements.

AI Requires Computation, Not Merely Data Storage

As noted above, conventional data centers primarily stored and distributed information used for email, websites, corporate information systems, and other digital services. Generative AI, by contrast, involves both training, in which an AI model is developed using large datasets, and inference, in which a trained model is used to generate text, images, and other outputs.

Some recent foundation models are reported to contain hundreds of billions of parameters, with certain models reaching a total parameter count on the order of one trillion.

In models based on a Mixture of Experts, or MoE, architecture, however, only a subset of the model’s parameters is activated during inference. Total parameter count therefore does not directly represent the amount of computation required for each operation.

MoE refers to an architecture in which only the components of a very large AI model that are needed for a particular task are activated, rather than operating the entire model each time.

Advanced AI training systems may operate thousands or tens of thousands of GPUs or other specialized accelerators in parallel. Some planned systems are expected to exceed even this scale. Consequently, AI data centers require substantially higher power densities and cooling capacities than conventional data centers.

Their electricity-demand profiles also differ. AI training often requires large numbers of GPUs to operate synchronously for extended periods, creating sustained and concentrated electrical loads. Inference workloads may involve shorter individual processing times, but as the number of users increases, they can create a continuous baseline load that fluctuates according to the time of day and patterns of use. These characteristics differ significantly from the load profiles traditionally associated with conventional data centers.

Global Data Center Electricity Consumption Is Also Expected to Rise

The International Energy Agency, or IEA, estimates that global annual electricity consumption by data centers, which was approximately 415 TWh in 2024, could rise to around 945 TWh by 2030. This would be broadly comparable to Japan’s current total annual electricity consumption.

Since 2017, electricity consumption by data centers worldwide has increased at an average annual rate of approximately 12 percent. The expansion of generative AI is identified as one of the principal drivers of this growth.

These projections nevertheless involve substantial uncertainty. Future demand will depend on several factors, including the pace of AI adoption, improvements in semiconductor performance, the efficiency of AI models, and the development of electricity infrastructure.

As shown in Figure 1, the IEA presents four scenarios for data center electricity demand through 2035:

• Lift-Off: AI adoption expands rapidly, while electricity infrastructure is developed without major delays.
• Base: The reference scenario considered the most plausible under current conditions.
• High Efficiency: Semiconductor performance and AI-model efficiency improve more rapidly than currently expected.
• Headwinds: AI adoption slows because of electricity-supply constraints, weaker economic conditions, or other limiting factors.

The projected level of global data center electricity consumption in 2035 ranges widely, from approximately 700 TWh to 1,700 TWh. Even so, all four scenarios anticipate higher demand than at present. The direction of change is therefore consistent: data center demand is expected to continue growing in the AI era.

In Japan, additional data center development is likewise expected to increase electricity consumption. This demand is also likely to become geographically concentrated in particular locations where large-scale facilities are developed.

Demand Growth Is Reflected in Facility Performance, Not Simply in the Number of Buildings

The expansion of next-generation AI data centers is taking two forms. On the one hand, new campus-type facilities are being developed in which multiple buildings are operated as a single integrated complex. On the other hand, existing facilities are being upgraded through the installation of high-performance GPUs and high-density server racks, increasing IT equipment capacity, electrical receiving capacity, and cooling capacity without any substantial change in building size.

For this reason, conventional building statistics alone cannot fully capture demand for AI data centers. Even where two facilities each have a total floor area of 10,000 square metres, a facility accommodating conventional servers and one accommodating high-density GPU systems may require substantially different levels of electrical receiving capacity, cooling equipment, emergency power supply, and heat-rejection capacity. In other words, buildings of the same size do not necessarily impose the same demands on the regional power grid, water resources, or surrounding environment.

The increasing scale and the increasing density of next-generation AI data centers should also be treated as distinct phenomena. Expansion in scale refers to the enlargement of sites and buildings, including the integrated operation of multiple buildings as a single campus. Higher density, by contrast, refers to an increase in IT equipment capacity or electricity consumption per unit of floor area. The former expands the spatial effects on land use and urban infrastructure, while the latter may intensify localized pressures associated with electricity supply, cooling, noise, and heat discharge.

Data center demand in the AI era should therefore be assessed using multiple indicators, including:

• total floor area;
• IT equipment capacity;
• electrical receiving capacity;
• power density per unit of floor area;
• cooling method and cooling capacity;
• utilization rate; and
• the overall scale of a campus comprising multiple buildings.

Unless these indicators are distinguished, demand for buildings may be confused with demand for infrastructure. Electrical receiving capacity and cooling capacity may increase substantially even where total floor area changes only slightly. Conversely, a facility occupying a large site may initially operate at a relatively low utilization rate, meaning that its actual electricity demand remains below its planned capacity.

Future demand should therefore be evaluated not only in terms of the physical size of a building, but also in operational terms: how much computing capacity is provided, and how much electricity and cooling are required to support it. Greater disclosure of information concerning the equipment, capacity, and operation of data centers in Japan would be necessary to obtain a more accurate understanding of actual conditions.

The growth of next-generation AI data centers may therefore be better understood not simply as an increase in the number of buildings, but as an expansion of a new form of infrastructure demand integrating land, electricity, cooling, and telecommunications. A multidimensional assessment is required, based on indicators such as IT equipment capacity, electrical receiving capacity, power density, utilization rate, and the overall scale of the data center campus.

Future Demand Should Be Assessed through Multiple Scenarios

The spread of AI will not necessarily result in a proportional increase in electricity demand.

In recent years, improvements in the performance of GPUs and AI-specific semiconductors, together with efficiency-enhancing techniques such as quantization, model compression, and Mixture of Experts, or MoE, architectures, have made it possible to perform the same tasks with less computation. Further advances in these technologies are likely to continue. Some inference workloads may also be distributed to end-user devices or edge-computing facilities, meaning that not all demand will necessarily be concentrated in large-scale data centers.

At the same time, lower computing costs may accelerate the adoption of AI itself. As generative AI is incorporated into a wider range of services—including search engines, office software, architectural and engineering design, healthcare, manufacturing, logistics, and autonomous driving—inference workloads will occur on a continuous basis. Even where electricity consumption per individual task decreases through efficiency improvements, total demand may still increase if the number of uses grows at a faster rate.

More importantly, decisions concerning urban planning and infrastructure cannot be based solely on average future demand. Electricity supply systems, land, and supporting infrastructure must be planned with potential peak demand in mind.

Even where computational efficiency improves in the future, electrical receiving facilities and transmission networks cannot be developed within a short period only after demand has materialized. The construction of power plants, substations, and transmission lines may require several years or, in some cases, more than a decade.

In parts of the United States and Europe, the rapid development of data centers has reportedly outpaced the reinforcement of transmission networks and the processing of grid-connection applications, resulting in connection delays and restrictions on new development. Similar connection delays may therefore arise in Japan in the future and may, in some locations, already be occurring.

Data center demand in the AI era should therefore not be assessed solely in terms of improvements in computational efficiency. Planning should instead be based on multiple scenarios incorporating factors such as:

• growth in computing demand;
• facility utilization rates;
• electricity-supply capacity;
• land-use conditions; and
• the development of transmission networks.

This scenario-based approach is also important for land-use planning. For example, next-generation AI data centers could be expressly addressed in municipal master plans for city planning and Location Optimization Plans as facilities that support, but also place substantial demands on, regional infrastructure.

Incorporating such facilities into long-term land-use and infrastructure planning would make it possible to consider their siting not only as an individual development decision, but also as part of the broader spatial structure and infrastructure strategy of the region.

Data Center Development Is Also Advancing in Japan

Data centers in Japan are not distributed evenly across the country. Instead, they are heavily concentrated in the Tokyo and Osaka metropolitan regions.

According to materials published by the Ministry of Internal Affairs and Communications and the Ministry of Economy, Trade and Industry, 510 data center buildings had been identified in Japan as of 2023, with a combined server-room area of approximately 1.68 million square metres. Of this total, the Kanto region accounted for approximately 1.07 million square metres, or 64 percent, while the Kansai region accounted for approximately 410,000 square metres, or 24 percent. Together, the two regions represented approximately 88 percent of the national total.

Location of Data Centers in Japan, 2023
Source: Ministry of Internal Affairs and Communications and Ministry of Economy, Trade and Industry, “Expert Meeting on the Development of Digital Infrastructure, Including Data Centers,” Secretariat Briefing Materials for the Seventh Meeting

Major Data Center Development Plans from 2024 Onward
Source: Ministry of Internal Affairs and Communications and Ministry of Economy, Trade and Industry, “Expert Meeting on the Development of Digital Infrastructure, Including Data Centers,” Secretariat Briefing Materials for the Seventh Meeting

By number of buildings, however, the Kanto region accounted for 38 percent and the Kansai region for 16 percent, giving the two regions a combined share of 54 percent. Because their share of total server-room area is substantially higher than their share of building numbers, the concentration is not simply a matter of there being more facilities in the Tokyo and Osaka metropolitan regions. It also indicates that comparatively large data centers are disproportionately located in these areas.

Several factors explain this concentration, including the scale of communications demand, access to domestic and international telecommunications networks, the availability of electricity supply, and proximity to major corporate customers. In the Tokyo metropolitan region, particularly in Inzai City, Chiba Prefecture, a major cluster of large-scale data centers has already formed. Development plans announced for 2024 and later likewise show continued concentration in the Tokyo and Osaka metropolitan regions.

How Should Data Centers Be Classified by Use under the Building Standards Act?

Data centers are not established as an independent building use under the Building Standards Act. Based on publicly available building-confirmation documents, interpretations issued by local governments, and materials relating to legal disputes, data centers have variously been treated as offices, warehouses, factories, or “other” uses.

An important distinction that is sometimes overlooked is that requirements applicable to an individual building—often referred to as individual building provisions, such as requirements for fire-resistant buildings—are separate from the provisions governing whether a building may be located on a particular site. The latter are generally referred to as collective provisions, because they regulate buildings in relation to surrounding land uses and the wider urban area.

The issue that has recently become particularly significant concerns whether a data center may be located within a particular use district under Article 48 of the Building Standards Act.

The Building Standards Act, however, contains no definition of a data center. Moreover, local governments had already developed their own interpretations before the siting of large-scale data centers emerged as a major policy issue.

Because each local government with a building official—an individual qualified to conduct building-standards conformity assessments—may determine the applicable use based on the circumstances of an individual project, differences in administrative practice have arisen across Japan. The Japan Conference of Building Administration, an organization composed of building-administration authorities, has likewise not presented a definitive national policy on how the use of data centers should be determined.

This is therefore an important issue not only for individual development projects, but also for considering the broader structure of building administration in Japan.

Table 2. Current Building-Use Classifications and Siting Restrictions Associated with Lithium-Ion Batteries

Potential classification under the Building Standards Act	Similarities to a data center	Principal use districts in which siting may be permitted	Effect of limits on the quantity of electrolyte contained in lithium-ion batteries
Office	Monitoring, maintenance, management, and information-processing activities	Permitted in Category II Medium-to-High-Rise Oriented Residential Districts, Category I Residential Districts, Category II Residential Districts, Quasi-Residential Districts, Neighborhood Commercial Districts, Commercial Districts, Quasi-Industrial Districts, Industrial Districts, and Exclusively Industrial Districts. Large-scale facilities may generally be located only from Category II Residential Districts upward.	Siting restrictions may apply where the relevant quantity exceeds 5,000 litres in residential use districts, 10,000 litres in commercial use districts, or 50,000 litres in Quasi-Industrial Districts.
Warehouse, excluding a warehouse used for warehousing business	Servers, racks, and related equipment are accommodated within the building	Permitted in Category II Medium-to-High-Rise Oriented Residential Districts, Category I Residential Districts, Category II Residential Districts, Quasi-Residential Districts, Neighborhood Commercial Districts, Commercial Districts, Quasi-Industrial Districts, Industrial Districts, and Exclusively Industrial Districts. Large-scale facilities may generally be located only from Category II Residential Districts upward.	Same as above.
Factory	Continuous information processing by machinery, with electrical and cooling equipment accounting for a substantial part of the facility	Depending on scale and operational characteristics, siting may be permitted in Quasi-Industrial Districts, Industrial Districts, and Exclusively Industrial Districts.	In a Quasi-Industrial District, siting restrictions may apply where the relevant quantity exceeds 50,000 litres.
Other use, not falling within any existing classification	The facility does not clearly correspond to any established use category	Determined by the designated administrative agency based on the actual characteristics and operation of the facility.	Classification as “other” does not avoid restrictions based on the quantity of hazardous materials. Depending on the quantity of batteries and electrolyte, the range of use districts in which the facility may be located may still be limited in substantially the same manner as for an office or warehouse.

Large-scale data centers developed in response to the expansion of generative AI possess characteristics that cannot be fully captured by conventional building-use classifications.

In addition to server rooms, these facilities integrate electrical receiving and transformation equipment, emergency generators, cooling equipment, and telecommunications facilities. The facility as a whole continuously provides information-processing services. Although relatively few people may be present inside the building, the facility consumes substantial amounts of electricity, while cooling systems, heat discharge, and emergency generators may affect the surrounding environment in ways that differ considerably from ordinary offices or warehouses.

Campus-type developments comprising multiple buildings operated as a single integrated facility have also become increasingly common. Such developments function at the level of the entire site rather than as a collection of independent buildings.

Mechanically assigning these facilities to an existing use category may describe how an individual building is occupied or operated. It does not, however, provide an adequate basis for assessing the facility’s wider effects on regional infrastructure, surrounding land uses, and the urban environment.

Regulatory Reforms Being Advanced by the National Government

The Japanese government has also begun reviewing the institutional issues surrounding data centers. In the Recommendations on Regulatory Reform compiled by the Council for Promotion of Regulatory Reform on June 29, 2026, the “acceleration of domestic siting of next-generation AI data centers” was identified as an independent reform item.

At this stage, therefore, the issue is no longer merely under general consideration. The recommendations already specify concrete measures and implementation schedules. The government has stated that domestic computing infrastructure must be secured for AI processing in which low latency, privacy, and data security are particularly important.

At the same time, the electricity consumption of AI servers can fluctuate significantly during computation. Lithium-ion batteries may therefore be installed near server racks in order to absorb and level these fluctuations, and next-generation AI data centers are expected to contain large quantities of storage batteries. Existing regulations under the Fire Service Act and the Building Standards Act were not designed with this type of facility configuration fully in mind, creating regulatory obstacles to the domestic siting of such facilities.

Three Issues Identified by the National Government

1. Aggregation of Hazardous Materials Contained in Lithium-Ion Batteries

The first issue concerns the aggregation of hazardous materials contained in lithium-ion batteries.

The electrolyte contained in lithium-ion batteries constitutes a hazardous material under the Fire Service Act. Where the aggregate quantity of electrolyte in batteries installed within a facility exceeds the designated quantity, regulations applicable to hazardous materials storage facilities or handling facilities may apply.

Under an existing notice issued by the Fire and Disaster Management Agency, batteries whose safety has been confirmed through specified combustion testing may be excluded from the calculation of the aggregate quantity of hazardous materials. In principle, however, the relevant test assumes that the batteries are tested in a sealed condition. AI server equipment may require openings for cooling and may therefore be unable to satisfy those testing conditions.

The international UL 9540A test method allows testing with openings in place, but Japan does not yet have a sufficiently developed domestic testing environment capable of conducting all of the necessary tests. This has also been identified as an issue.

2. Lack of Coordination between the Fire Service Act and the Building Standards Act

The second issue is that the regulations under the Fire Service Act and the Building Standards Act do not operate in a coordinated manner.

Even where the safety of lithium-ion batteries has been confirmed and their electrolyte is excluded from aggregation under the Fire Service Act, the separate restrictions on the storage of hazardous materials under Article 130-9 of the Order for Enforcement of the Building Standards Act continue to apply.

As a result, a data center containing a large quantity of lithium-ion batteries may face restrictions on siting in Category II Medium-to-High-Rise Oriented Residential Districts, Category I and Category II Residential Districts, Quasi-Residential Districts, Neighborhood Commercial Districts, Commercial Districts, and Quasi-Industrial Districts. The facility may consequently be required to locate in an Industrial District or an Exclusively Industrial District.

In other words, even where safety performance has been verified, the determination made under the Fire Service Act is not reflected in the siting restrictions imposed under the Building Standards Act. This creates an inconsistency between the two regulatory systems.

3. Limited Regulatory Options for Fire-Extinguishing and Smoke-Exhaust Systems

The third issue is that the available options for fire-extinguishing equipment and smoke-exhaust equipment are not fully aligned with the characteristics of next-generation AI data centers.

Telecommunications equipment rooms above a specified scale are generally required to install fire-extinguishing systems using inert gas, halogenated agents, or dry chemicals. In fires involving lithium-ion batteries, however, water-based fire-extinguishing systems may be effective because of their cooling function.

The existing regulations do not yet provide a sufficiently developed framework for selecting water-based fire-extinguishing systems for telecommunications equipment rooms. Moreover, even where the use of such systems becomes permissible, the current regulations under the Building Standards Act do not provide the same exemption from smoke-exhaust equipment requirements that may apply where inert-gas or similar fire-extinguishing systems are installed.

Consequently, separate smoke-exhaust equipment may still be required even where its necessity from the standpoint of fire safety is limited.

The recommendations also identify as a regulatory issue the fact that data centers may be restricted from locating in Quasi-Residential Districts, Commercial Districts, and Quasi-Industrial Districts even where access to electricity and telecommunications infrastructure is available.

However, access to electricity and telecommunications infrastructure is a separate question from land-use compatibility from the perspective of urban planning and community development.

In particular, next-generation AI data centers equipped with large quantities of storage batteries, emergency generators, and cooling systems should not necessarily be permitted in residential or commercial use districts solely because existing infrastructure is available. It may instead be appropriate to establish an urban-planning mechanism through which their location can be coordinated in light of surrounding land uses, infrastructure capacity, and local environmental conditions.

Planned Regulatory Reforms

The recommendations indicate that these issues should be addressed not merely through separate amendments to the Fire Service Act and the Building Standards Act, but through an integrated review involving the relevant ministries and agencies.

Table 3. Planned Regulatory and Institutional Reforms

Item	Responsible authority	Planned reform	Schedule	Relevant legislation or instrument
Exclusion of storage batteries from aggregate hazardous-material quantities	Fire and Disaster Management Agency	Revise the relevant Fire and Disaster Management Agency notice to align domestic treatment with UL 9540A and allow storage batteries whose safety has been confirmed to be excluded from the calculation of aggregate hazardous-material quantities. A domestic testing method for batteries tested with openings in place will also be considered.	Measure to be implemented during FY2026	Fire Service Act and Fire and Disaster Management Agency notice
Restrictions on the storage of hazardous materials under the Building Standards Act	Ministry of Land, Infrastructure, Transport and Tourism	Consider exempting lithium-ion batteries that qualify for exclusion from aggregation under the Fire Service Act from the application of Article 130-9 of the Order for Enforcement of the Building Standards Act.	Review to begin in the first half of FY2026, followed by prompt implementation once a conclusion is reached	Article 130-9 of the Order for Enforcement of the Building Standards Act
Selection of water-based fire-extinguishing systems	Fire and Disaster Management Agency	Examine fire-control effectiveness, water damage, electric-shock risks, and the safety of evacuees and firefighting personnel, and consider the enactment or amendment of an Ordinance of the Ministry of Internal Affairs and Communications to permit the selection of water-based fire-extinguishing systems.	Review to begin in the first half of FY2026, followed by prompt implementation once a conclusion is reached	Articles 13 and 29-4 of the Order for Enforcement of the Fire Service Act
Exemption from smoke-exhaust equipment requirements	Ministry of Land, Infrastructure, Transport and Tourism	Where a telecommunications equipment room is fitted with a water-based fire-extinguishing system and smoke-exhaust equipment is determined to be unnecessary, provide an exemption through measures including amendment of Ministry of Construction Notification No. 1436 of 2000.	Review to begin in the first half of FY2026, followed by prompt implementation once a conclusion is reached	Article 126-2 of the Order for Enforcement of the Building Standards Act and Ministry of Construction Notification No. 1436 of 2000
Cross-ministerial review framework	Ministry of Economy, Trade and Industry and Ministry of Internal Affairs and Communications	Establish a framework in which the Fire and Disaster Management Agency, the Ministry of Land, Infrastructure, Transport and Tourism, and other relevant authorities jointly discuss updates to the applicable regulations and institutional arrangements.	Measure to be implemented during FY2026	Institutional and administrative arrangements
Coexistence with local communities	Ministry of Economy, Trade and Industry and Ministry of Internal Affairs and Communications	Promote effective compliance with the Guidelines for Data Center Coexistence with Local Communities and support voluntary information disclosure by industry associations.	Measure to be implemented during FY2026	Guidelines

Of these measures, the most significant from the perspective of the Building Standards Act is the proposed linkage between safety assessments conducted under the Fire Service Act and the quantity restrictions applicable to hazardous-material storage in individual use districts under Article 130-9 of the Order for Enforcement of the Building Standards Act.

If this reform is implemented, data centers containing lithium-ion batteries whose safety has been technically verified may be permitted in a wider range of use districts than under the current framework.

In practical terms, the reform could make it possible to locate data centers requiring large-capacity storage batteries not only in industrial use districts, but also in certain residential and commercial use districts.

The Underlying Issue of Building-Use Classification Remains Unresolved

The principal focus of the recent recommendations is the review of rules concerning lithium-ion batteries, fire-extinguishing equipment, smoke-exhaust equipment, and limits on the quantity of hazardous materials permitted within individual use districts. The recommendations do not establish uniform criteria for determining whether a data center should be treated under the Building Standards Act as an office, factory, warehouse, or “other use,” such as a computer room or data center.

One relevant example is currently being contested in Inzai City. A certificate of confirmation was issued for a data center within an area subject to a district plan that restricts factories and warehouses used for warehousing business. The underlying use district is a Commercial District, but additional restrictions on factories and related uses are imposed through the district plan. The legality of the development is now the subject of litigation. For the purposes of the confirmation process, the private designated confirmation and inspection body classified the building as “other use (data center).”

Inzai City is a limited designated administrative agency whose authority to examine buildings is restricted to smaller-scale projects. However, a municipal government may establish an ordinance implementing restrictions under a district plan pursuant to Article 68-2 of the Building Standards Act.

It is therefore possible that the designated confirmation and inspection body sought the City’s view as to whether the data center fell within a use restricted by the ordinance. However, the publicly available materials do not confirm whether any such inquiry or consultation actually took place. This point therefore remains an inference rather than an established fact.

The district-plan ordinance in question does not prohibit warehouses in general. It specifically restricts warehouses used for warehousing business. A data center ordinarily differs in character from a warehousing business, which stores goods belonging to other persons under a deposit or custody arrangement. The likelihood that a data center would fall within the category of a warehouse used for warehousing business therefore appears limited.

The central issue in this case is consequently more likely to be whether a data center should be treated as a factory, because it continuously processes information through machinery and equipment, or whether it may be treated as an other use that does not fall within any existing classification.

The current regulatory reforms may ease some of the siting restrictions affecting AI data centers containing large quantities of storage batteries. They will not, however, immediately resolve the more fundamental problem that designated administrative agencies may reach different conclusions regarding building-use classification.

The recommendations also require verification that any relaxation of siting restrictions will not cause adverse effects on traffic or other aspects of the urban environment. They further emphasize the need to address concerns regarding the safety and security of local residents and to secure social acceptance.

The government’s approach should therefore not be understood as a simple relaxation of regulation. Rather, it represents a shift toward performance-based regulation: where safety performance has been verified, a broader range of siting options may be permitted, while appropriate consideration must still be given to the surrounding environment and local communities.

The benefits generated by a data center are not confined to the municipality in which it is located. They may extend across a much wider geographic area. Entrusting the entire siting decision to a single municipality may therefore also raise questions of governance.

The current reform process seeks to assess data centers across multiple regulatory dimensions, including the safety performance of storage batteries, fire-extinguishing methods, smoke exhaust, and impacts on surrounding urban areas, rather than merely assigning them formally to a conventional building-use category. Nevertheless, important issues remain unresolved with respect to building-use classification and the use of urban planning to guide the location of such facilities.

The position may be summarized as follows.

Data centers resemble warehouses in their external form because they accommodate large numbers of servers and racks while generally involving relatively limited continuous human occupancy. This resemblance is particularly apparent in colocation facilities, where customers’ equipment is accommodated and provided with electricity, telecommunications connectivity, and controlled cooling conditions. To that extent, it is reasonable to understand such a facility as a building used to accommodate and protect equipment.

Conventional data centers, however, have never merely stored servers. Their servers have continuously performed the computational processing required for websites, databases, business systems, cloud services, and other digital functions. There are therefore clear limits to treating a data center as equivalent to a warehouse.

In data centers designed for generative AI, this computational function becomes substantially denser and larger in scale. Large numbers of high-performance GPUs and specialized accelerators are installed and operated together with electrical receiving and transformation equipment, cooling systems, electricity-storage systems, and emergency generators. These facilities continuously perform large-scale parallel computation. Functionally, they share certain characteristics with factories in that electricity is supplied to machinery and equipment that operate continuously to produce computational outputs.

The concept of a factory under the Building Standards Act, however, has generally developed with facilities that manufacture or process tangible goods in mind. A data center provides information-processing, telecommunications, or computational services. It should therefore not automatically be treated as a factory. Ordinary data center operations are also not manufacturing businesses, making it difficult to regard them as factories under the Factory Location Act.

A data center is therefore more appropriately understood as a complex facility combining:

• a warehouse-like function of accommodating equipment;
• an office-like function of monitoring and management; and
• a factory-like function of continuous processing.

These functions are even more closely integrated in next-generation AI data centers, where telecommunications equipment, computing equipment, electrical infrastructure, and cooling systems operate as a single facility system.

Rather than mechanically assigning such facilities to one of the existing use categories, one possible approach would be to establish data centers as an independent building use and introduce graduated use restrictions based on factors such as:

• IT equipment capacity;
• electrical receiving capacity;
• power density;
• quantities of storage batteries and emergency fuel;
• cooling method; and
• the overall scale of the facility or multi-building campus.

A further option would be to position large-scale data centers as urban facilities under the City Planning Act and establish a mechanism through which their location could be coordinated from a wider regional perspective.

Conclusion

At present, data centers are not defined as an independent building use under the Building Standards Act. Instead, they are assessed on a case-by-case basis as offices, warehouses, factories, or “other” uses. Next-generation AI data centers, however, differ substantially from conventional facilities. They integrate high-density GPU clusters, large-scale electrical receiving and transformation equipment, extensive lithium-ion battery systems, emergency generators, and cooling equipment, and they perform continuous information processing on a 24-hour basis. In light of these characteristics, it is becoming increasingly difficult to ensure consistent treatment nationwide through project-by-project determinations under the existing system of building administration.

Going forward, next-generation AI data centers should be clearly defined in law, and use restrictions should be graduated according to factors such as IT equipment capacity, electrical receiving capacity, power density per unit of floor area, quantities of storage batteries and emergency fuel, cooling method, and the overall scale of a campus comprising multiple buildings.

Small edge facilities and enterprise data centers may reasonably continue to be treated in a manner comparable to offices or other ordinary uses. By contrast, for large-scale AI data centers with capacities of several tens of megawatts or more, the appropriateness of siting in residential and commercial use districts should be examined carefully in light of their electricity demand, equipment scale, and potential effects on the surrounding environment.

Where exceptional siting in residential or commercial areas is considered on the basis of access to electricity and telecommunications infrastructure, the decision should not be made solely through an individual confirmation process. Instead, a comprehensive review should be conducted through a district plan or city planning decision, taking into account residents’ views, land use in surrounding municipalities, noise, heat discharge, landscape, traffic, disaster prevention, and the burden placed on regional infrastructure.

Even where exemptions or other forms of regulatory relief are introduced under the Fire Service Act, it remains important to distinguish between two separate questions: whether the facility has been confirmed to be safe from the standpoint of fire protection, and whether the proposed land use is appropriate from the perspective of urban planning and community development.

At the same time, next-generation AI data centers are likely to become increasingly indispensable to social and economic activity. Public policy should therefore explain both their public value and their locational impacts to residents and the wider public, while also securing appropriate sites at the scale of the broader region that benefits from the services they provide.

Further examination will be required as national regulatory reforms proceed, particularly with respect to consistency with the use-district system and the development of more effective mechanisms for guiding the location of these facilities.

References

• Ginzburg-Ganz, E., Lifshits, P., Machlev, R., Belikov, J., Krieger, Z., and Levron, Y. (2026). “Technical Challenges of AI Data Center Integration into Power Grids—A Survey.” Energies, 19(1), Article 137.
• Oba, T., and Kamata, Y. (2025). “Urban Planning Challenges Associated with Data Center Siting in the Greater London Area and Policy Implications for Japan” [in Japanese]. Reports of the City Planning Institute of Japan, No. 24, pp. 494–498.
• International Energy Agency (IEA). (2025). Energy and AI: World Energy Outlook Special Report. Paris: International Energy Agency.
• Ministry of Internal Affairs and Communications and Ministry of Economy, Trade and Industry. (2024). “Secretariat Briefing Materials for the Seventh Expert Meeting on the Development of Digital Infrastructure, Including Data Centers,” Document No. 4 [in Japanese].
• Council for Promotion of Regulatory Reform. (2026). “Implementation Measures under the Recommendations on Regulatory Reform,” June 29, 2026 [in Japanese].