On June 16, the government approved the National Artificial Intelligence Program, which sets ambitious goals for Israel in the coming years, chief among them the expansion of its sovereign computing capacity. That same day, the head of the National AI Directorate, Brig. Gen. (res.) Erez Eskel, appeared before a heated discussion of the Knesset’s Interior and Environmental Protection Committee, which focused on advancing a fast-track planning process for the construction of data centers.
The discussion revealed that the ambition to turn Israel into a global computing powerhouse is likely to be complex to implement. The government’s objectives faced opposition from local authorities and civil society organizations, alongside pressure from industry to rapidly advance the construction of the necessary infrastructure. Thus, the newly approved national program received its first practical test, highlighting the planning and regulatory challenges it will face in implementation.
The urgency of the discussion stemmed from the gap between the future needs of “the cloud” and today’s infrastructure limitations: massive data centers require extensive real estate and exceptionally large amounts of electricity.
Under Amendment No. 168 to the Planning and Building Law, which was discussed by the committee, “a data center intended for artificial intelligence uses” would be classified as a national infrastructure project, similar to ports and prisons.
The significance of this is that particularly large data centers, which consume at least 50 megawatts of electricity and contain advanced processing units (GPUs), would be able to undergo an accelerated planning process through the National Infrastructure Committee (NIC), rather than through the regular planning committees.
An Amount Equivalent to the Consumption of an Entire City
The National Artificial Intelligence Program assumes that Israel’s energy production capacity will undergo significant expansion in the coming years, including through cooperation with the United States, the United Arab Emirates, and India. According to the plan, this would involve an overall addition of 7 to 10 gigawatts, an increase of nearly 50% compared with current production levels.
Representatives from the National AI Directorate told the committee that achieving the flagship goal of the national AI program, the establishment of a national computing system containing 100,000 GPUs, the chips essential for training artificial intelligence models, would require an electricity capacity of approximately 300 megawatts, an amount equivalent to the consumption of an entire city. Against this backdrop, they supported the proposed legislation, which seeks to fundamentally change the way the state approaches data centers.
Local authority heads warned that the fast-track process would weaken local planning powers and give private developers greater control over land and energy resources.
Galit Shaul, head of the Emek Hefer Regional Council, asked: “What sane country gives up its right to plan?” Oshrat Gani Gonen, head of the Southern Sharon Regional Council, said: “We are selling the State of Israel to developers.”
Alongside concerns about harm to local planning, the discussion also raised questions about the electricity grid’s ability to meet the demand created by data centers. Ron Eifer, who represented the Ministry of Energy, said that the electricity sector could absorb “another thousand or slightly more megawatts of data centers,” but added that “the electricity sector needs to take significant steps in order to meet demand.”
The Society for the Protection of Nature in Israel and Lobby 99 also joined the concerns regarding the gap between the electricity demands of data centers and the capacity of the electricity grid.
Ofir Paar from Lobby 99 said that in large areas of Israel, the grid is “stretched to the limit,” and that data centers “on the scale of a small city” could increase electricity costs and even delay the construction of essential infrastructure such as hospitals.
However, Eifer emphasized that the Ministry of Energy “does not want to become a barrier to progress and to turning Israel into a global leader in the field of AI.” The Electricity Authority also presented a public consultation during the discussion, including an overview of the expected impacts of the planned computing infrastructure, alongside a series of regulatory proposals to address these challenges.
A Technological Arms Race
According to government and industry representatives, the legislation is urgently needed because many countries around the world are already working to secure advanced computing infrastructure for artificial intelligence. They argue that countries that fail to establish these infrastructures in time risk falling behind in the technological, economic, and security race.
“This is an arms race. Just like weapons, just like nuclear capabilities, that is how it is with artificial intelligence,” said Yonit Goldberg, founder of the Israeli Data Center Association. According to her, AI infrastructure has become a strategic resource over which countries are competing.
A similar position was presented in a statement published by the World Economic Forum in April, which stated that “AI infrastructure is now a strategic national infrastructure, with importance comparable to electricity grids, ports, or oil pipelines.”
Lior Haimovitz, deputy head of the National Economic Council, emphasized that Israel is lagging behind leading countries in terms of infrastructure. Goldberg elaborated that, according to her assessment, the gap stands at two years behind the United States, one and a half years behind Singapore, and one year behind Germany and the United Kingdom.
The head of the National AI Directorate, Erez Eskel, echoed these concerns, saying that Israel must rapidly secure the computing infrastructure that will “determine our future and our security.” According to him, artificial intelligence is “the new oil,” and its importance is critical to national security, the economy, and the country’s continued prosperity.
These statements align with a State Comptroller report by Matanyahu Englman on national artificial intelligence infrastructure, published last week. The report stated that infrastructure is not merely a technical matter, but a factor that determines what a country is capable of developing, and that artificial intelligence is already changing the way governments operate.
The State Comptroller’s report, which compared Israel’s AI capabilities with those of developed countries in Europe and around the world, described Israel as having very high national potential due to its strengths in research and human capital. However, according to the report, this potential has not been fully realized, partly due to the lack of coordinated government leadership and delays in building the necessary physical and regulatory infrastructure.
The State Comptroller noted that Israel ranks 26th in infrastructure in the Global AI Index, and pointed to a significant gap in budget implementation: out of the 270 million shekels already approved for advanced computing, only 11% has been utilized.
During the committee discussion, a major gap emerged between two key processes that must advance simultaneously if the country is to meet its goals regarding supercomputing infrastructure. Goldberg said that the construction time for data centers in Israel is approximately two years, but she also joined concerns that the pace of construction would not align with processes dependent on the public sector, including expanding electricity capacity, upgrading transmission infrastructure, and completing planning and regulatory procedures. “We do not have time to wait for the regulator,” Goldberg concluded, “because otherwise there will be no event here.”
In April, the International Energy Agency (IEA) published a report highlighting a severe infrastructure “bottleneck” around the world, illustrating the concerns raised during the committee discussion. According to the report, “the speed of the AI revolution stands in growing contrast to the speed of the physical, social, and economic systems that underpin it.”
The report emphasizes that grid connection queues are “long and complex,” with actual waiting times reaching “five to ten years in many jurisdictions.” In other words, even if data centers can be built within approximately two years, the concern raised by the committee chair, MK Yitzhak Kroizer—“will we build data centers and then not have enough electricity to operate them?”—is already becoming a reality in many Western countries.
The discussion also raised the possibility that challenges faced by other countries could provide Israel with an advantage, even if only temporarily. The new National AI Program argues that “at this stage, the global regulatory environment is still being shaped. This reality creates a unique opportunity for Israel to influence the formation of the rules of the game.”
Goldberg presented Ireland as an example of a country that encouraged extensive data center development, until heavy pressure on the electricity grid led authorities to begin restricting projects. According to her, this has resulted in a current state of stagnation, as the country waits for a new strategy promised by the government. Against the backdrop of these regulatory upheavals, Goldberg noted, “companies have started leaving Ireland.”
Regulatory Approaches: Examples from the Netherlands and Singapore
The Netherlands presents a complex example: over the past decade, it has shifted from being a global hub attracting data centers to a country adopting a more cautious and stringent regulatory approach.
Similar to Ireland, the Netherlands made decisions that abruptly halted the expansion of the sector. In 2019, construction of data centers in Amsterdam was suspended; in 2022, parliament blocked Meta’s large-scale data center project; and in early 2024, a nationwide ban came into effect on the construction of facilities with a capacity exceeding 70 megawatts.
Lobby 99 argues that Israel should learn from the Netherlands and adopt a policy based on advance planning and proactive preparation of the electricity sector for growing demand. According to the organization, such preparation is necessary to prevent harm to supply reliability and increases in electricity prices for the wider economy.
However, the Dutch case also serves as a warning about the costs of prolonged regulatory freezes. In 2024, five years after the suspension in the Amsterdam region, the Dutch Data Center Association (DDA) argued in its annual report that the moratorium was still affecting sector growth: “The moratorium caused international demand to decline sharply and move to other locations across Europe.”
Singapore also imposed a construction freeze in 2019 due to concerns over electricity shortages and limited land availability. The freeze ended in 2022 and was replaced by a “Green Roadmap,” which transformed it from “a growth ceiling into a quality filter.”
Today, rather than imposing an outright restriction on data center construction, every new project in Singapore must meet strict requirements, including world-class energy efficiency, the use of green technologies, and efficient utilization of land, water, and energy resources. In this way, responsibility for improving the efficiency and capacity of the electricity system is placed not only on the supply side, but also on those consuming the energy.
The Singaporean strategy, considered a prominent example of how a small country can achieve significant influence in the field of AI, focuses on transforming the island into a global center for “applied innovation.” This approach resembles the proposed “National Institute for Artificial Intelligence” outlined in Israel’s new National AI Program.
The emphasis on building a center where AI technologies move from research into real-world applications, with the aim of leveraging research strengths to address challenges across a range of fields, capitalizes on Singapore’s advantages, similar to Israel’s, in research and innovation.
At the same time, leadership in applied innovation requires less raw computing power than the development and training of entirely new AI models. This allows Singapore’s strategy to better align with the limitations of its electricity system.
Ultimately, as part of its strategic pragmatism as a small state, Singapore has so far chosen not to pursue full infrastructural sovereignty in the field of AI. Instead, it has sought to balance dependence on infrastructure and models linked to China with dependence on infrastructure and models linked to the United States.
This approach stands in direct contrast to Israel’s national AI strategy regarding sovereignty and security through independent computing capacity. Israel’s new National AI Program defines independent computing infrastructure as reaching “the required technological threshold that will ensure access to resources and reduce dependence on international supply chains.” The aim is to provide the state with “broad independent capability.”
From this objective stem, among other things, the target of 100,000 GPUs in the sovereign AI computing system, as well as the 300 megawatts of electricity required to support it.
Proposals from the Electricity Authority
The Electricity Authority joined the discussion with a series of regulatory proposals tailored to Israel’s circumstances. First, regarding the accumulation of connection requests amounting to thousands of megawatts, the Authority noted that some are classified as “duplicate” requests—meaning commercial speculation that places unnecessary pressure on the system.
To address this issue and ensure that only serious projects occupy positions in the queue, the Authority proposed a screening and feasibility assessment process. This would require applicants to demonstrate a connection to the land and provide an approved construction plan for the facility consuming the electricity.
In addition, the Authority seeks to change the economic model governing positions in the connection queue, in order to prevent projects without genuine feasibility from blocking other developers.
Under this proposal, developers would be required to pay a fee for “reserving a place on the grid,” charged from the submission of the application until electricity capacity is actually provided. This would require companies to assess their ability to execute projects before entering the queue.
Two of the Authority’s key proposals relate to the location of data centers, an issue that received significant attention during the committee discussion. The Authority seeks to redirect high electricity demand away from the densely populated and congested central region toward areas where the electricity grid has greater capacity to accommodate large consumers. To achieve this, it is promoting incentives for electricity consumers requiring more than 50 megawatts that choose to connect in peripheral regions.
Another proposal addresses the distance between a large consumer and its power source, which requires the construction of transmission lines and grid infrastructure. Expanding these systems requires significant time, planning, land, and funding.
The Electricity Authority is considering allowing direct agreements between private electricity producers and nearby data centers, without routing electricity through the national grid. This model, known as a “behind-the-meter consumer” arrangement, would allow data centers to receive direct electricity supply from existing or newly built power stations, reducing lengthy planning timelines and easing pressure on the national electricity system.
At the same time, the government aims to transform data centers, as much as possible, into energy hubs in their own right. To achieve this, the pathway allowing the construction of independent generation and storage facilities within data center sites will be expanded. This would enable connection and generation requests to be assessed as a single integrated application submitted to the system operator.
Another proposal from the Electricity Authority seeks to avoid the need to build additional power plants solely to meet peak demand periods, such as the hottest days of summer.
Large data centers would be required to contribute to grid stability through a mechanism of “flexible management and backup,” requiring them to reduce their electricity consumption from the grid during designated peak-demand hours, up to a total of 200 hours per year, and rely on self-generated backup capacity. According to the Authority, this approach could help maintain reliable electricity supply for all households, to some extent even without constructing additional power plants.