“Fasten your seatbelts. We’re starting,” opened the message from the National AI Headquarters announcing the national plan in the field, which was approved by the government last week. The full version of the plan, obtained by Davar, sets ambitious targets regarding computing power and expresses an intention for close cooperation with the United States, while providing limited detail on regulatory mechanisms. It also presents a plan to establish a new body called the National Institute for Artificial Intelligence, intended to develop AI applications around national challenges in a variety of fields.
The plan was formulated after the State Comptroller called, in late 2024, for the urgent development of such a strategy. It was led by the National AI Headquarters in the Prime Minister’s Office, which was established last September. “We are in a historic window of opportunity,” said Head of the Headquarters Brig. Gen. (res.) Dr. Erez Eskel. “In the coming years it will become clear which countries manage to establish themselves at the global forefront of the AI era, and which will be forced to rely on technologies, infrastructure, and capabilities developed by others.”
Prime Minister Benjamin Netanyahu stated that the plan is intended to “establish Israel as a global leader” and “ensure technological independence.”
The plan is divided into several “layers,” similar to comparable strategies in other countries: energy, computing power, models, data, human capital, and regulation. This conceptual framework resembles what Nvidia refers to as the “five-layer stack”: energy, chips, data centers, models, and applications.
Hardware: 100,000 Processing Units
The flagship component of the plan is the establishment of sovereign computing infrastructure, including a quantum computer based, “as much as possible,” on Israeli technology.
The ambitious target is to reach 100,000 GPU processing units, which will serve the public sector, the private sector, and academia. These chips, which include such processing units, are essential for training artificial intelligence models.
For comparison, Saudi Arabia’s “Vision 2030” plan set an expectation of up to 600,000 advanced Nvidia GPU chips within the HUMAIN company owned by the Saudi sovereign wealth fund, over a period of about five years. South Korea announced toward the end of 2025 a plan to deploy more than 260,000 GPU chips.
The Israeli plan states that “today, computing infrastructure in Israel is estimated at approximately 10,000 GPUs. This is a limited level of AI infrastructure that does not match the scale required for a country aspiring to lead in artificial intelligence.”
The plan sets an expectation that Israel’s energy production capacity, currently standing at 20–23 gigawatts, will be expanded to 30 gigawatts in order to support the demands of the desired computing power.
Access to Computing Resources: No Defined Priority Order
The plan states that the state should enable “potential accessibility” to computing resources in order to meet national needs through three channels. In the public sector, the state will work to establish “allocation and implementation mechanisms for computing power” that will include all government ministries and relevant public bodies. Access will be provided through a “public cloud” or dedicated infrastructure designed to ensure operational independence.
In academia, access is intended to enable advanced research and development and to remove barriers faced by researchers. Regarding the private sector, the state will work to formulate “strategic partnerships and agreements” to ensure that local industry also has access to significant processing power, in order to maintain its competitive advantage.
According to the full wording of the plan obtained by Davar, the plan does not define an explicit prioritization in the allocation of these resources between the three intended sectors. The task of formulating “allocation and implementation mechanisms” has been assigned to the Head of the National AI Headquarters, who will be required to determine how computing power will be distributed in practice.
The plan aims to “strengthen [Israel’s] position as a significant actor in the global supply chain, with an emphasis on semiconductor manufacturing,” based on the understanding that there is currently a “rare strategic opportunity to connect planning and manufacturing within a single geographic space.”
To this end, the government has tasked the National AI Headquarters with defining the technological threshold for AI chips to be produced in Israel, with a dual objective: “to ensure the supply of chips for Israel’s needs and to position it as an influential player in the global semiconductor industry.”
While most sections of the plan present mainly general objectives, the chapter dealing with hardware, energy, and computing power sets quantitative targets. Like France, the United Kingdom, South Korea, Brazil, and India, Israel also places strong emphasis on computing capabilities.
In addition, the focus on quantum computing aligns with a broader trend in developed countries, however, with one important difference: in most of these countries, it is addressed in a separate program with substantial independent budgets, whereas in Israel the issue is currently integrated only as a sub-component within the artificial intelligence plan.
Regulation: What Works for Europe Does Not Work for Israel
In terms of regulation, the explanatory notes to the plan state that there is still no globally agreed model. Accordingly, Israel is consciously choosing not to follow the broad, comprehensive European approach, but instead to adopt a more sectoral and flexible approach, including “pilots, tailored regulatory pathways… and the removal of barriers where necessary.”
The Israeli plan does include references to the need to develop appropriate regulation, but it does not propose, for example, that the National Headquarters should initiate legislation in the Knesset to define a comprehensive regulatory framework. In practice, very few specific regulatory mechanisms are outlined in the plan.
In contrast to Israel, the AI policies of the European Union and Singapore, for example, treat regulatory certainty as a competitive advantage. These frameworks argue that uniform and clear regulation functions as a “common language” that enables relatively small states, such as Singapore, and complex coalitions like the EU, to coordinate between many actors, rather than each developing capabilities independently. In this way, it is precisely through regulatory certainty and the trust it generates that they are able to attract investment, sign agreements, and influence global norms.
Tool Development: Models Adapted to “Culture and Values”
The plan states that a “National Artificial Intelligence Institute” will be established, operating under a concept of “task-oriented artificial intelligence.” The institute will not directly engage in model development, but rather in creating practical tools. It will operate through “acceleration hubs”—frameworks for developing tools in areas such as health and transportation.
The plan defines a list of 17 aspects to which the National AI Institute will refer in its work. These include, among others, procurement mechanisms, performance indicators, inter-state cooperation, and the development of cyber capabilities.
The plan treats models as a layer that requires governmental centralization, rather than relying solely on the free market. The Head of the National Headquarters is instructed to include in his work plan “consideration… for the production of models adapted to the culture, language, values, and the national, historical, and cultural assets unique to Israel.”
At the implementation level, the plan assigns the new “National Institute” the role of coordinating between a wide range of government bodies, including the Ministry of Finance, the Ministry of Defense, the Israel National Digital Agency, the Israel Innovation Authority, the Ministry of Economy, the Planning and Budgeting Committee (VATAT), and the Council for Higher Education.
The plan stipulates that the central body, the new National AI Institute, will not train the models itself, but will instead direct academic, industrial, and governmental resources toward that goal, resources that are not specified in detail in the plan.
Data: A Government Platform for Data Sharing
The plan frames government data as a “national strategic asset” that requires a shift from a default policy of concealment to a policy of controlled sharing, in order to fuel local artificial intelligence capabilities. Data, the plan states, is “the oxygen required for building, training, and operating AI systems.”
Two tools are to be developed for this purpose. First, the establishment of a centralized government digital platform that will enable data sharing for the training and development of AI models. The second tool is the use of advanced privacy-enhancing technologies (PETs), which allow access to datasets containing sensitive and protected information while preventing the identification of individuals included in them.
Oversight of the entire process is assigned to a “steering committee to be established under the leadership of the Head of the National Artificial Intelligence Headquarters.”
In other words, decisions about who will share which data with whom and under what conditions will not be set by law or by an independent regulator, but rather through a mechanism operating under the Prime Minister’s Office.
This reflects a broader gap between the regulatory approach of the Israeli plan and those of developed countries worldwide, which propose more decentralized regulatory structures, including independent regulators and channels for public oversight, with a stronger emphasis on human rights considerations rather than efficiency alone.
Human Capital: Encouraging Immigration of AI Experts
The plan dedicates an entire extensive chapter to the issue of human capital, with reference to two target groups: the “creators,” meaning the engineers and researchers who develop AI infrastructure, and the “implementers,” which include all workers in the economy who will use AI tools to improve and streamline their work.
The areas of action in this context span all stages of life, starting with the education system, where teaching methods will be updated and emphasis will be placed on science and technology (STEM) subjects; through academia, where scholarships for relevant advanced degrees will be expanded; and up to the labor market, where the state will carry out real-time demand mapping and offer training and retraining pathways for workers in professions that are particularly exposed to technological change or may be displaced.
It is also proposed to establish a center within the Ministry of Aliyah and Integration to assist AI experts who are returning residents or new immigrants in integrating into Israel.
Security System: Integrated Across All Layers
Unlike typical international plans, in the Israeli framework security and intelligence do not appear as a separate implementation domain, but are embedded across all layers. “Welfare and national security” is defined as one of the overarching goals of the plan, alongside economic growth.
The plan includes a section on the bidirectional integration between cybersecurity and artificial intelligence: using cyber capabilities to protect AI systems, and also leveraging AI to detect digital threats and neutralize them more effectively. The plan defines a central national objective as the integration of AI tools into addressing security challenges, while using the security establishment as a “motor” for developing new technologies and training skilled personnel for the wider economy.
The security establishment does not appear only as an end-user of the technology, but as an active partner in management, planning, and execution. At the level of physical infrastructure, the Ministry of Defense participates through the Directorate of Defense Research & Development (MAFAT) in teams determining local chip manufacturing capabilities and the establishment of quantum computing infrastructure.
In the human capital layer, the security system is defined as a “central pillar,” whose role is to cultivate a new generation of experts for the civilian industry.
Even in the international arena, security bodies are part of the teams shaping strategic alliances, such as the “Silicon Pax” alliance.
Global Context: Connection to the United States, Diaspora Jewry as an Asset
The Israeli plan describes the world as shifting from a “global free-market reality” to a reality of alliances, and explicitly states that Israel “cannot afford to remain in the second or third circle, it must establish its place in the first circle.”
For this purpose, in terms of international cooperation, the plan focuses almost exclusively on Israel’s participation in “Pax Silica,” a new U.S.-led initiative that defines a circle of partnership aimed at a high-tech supply chain with minimal dependence on the Chinese axis.
In contrast to the plans of the European Union and India, for example, which seek to define a more diverse set of partners, Israel deliberately prioritizes a strategic alliance with the United States, alongside “expanding cooperation with countries, global corporations, and innovation communities” as a secondary layer.
The plan also refers to “leveraging the unique position of the Jewish diaspora as a bridge for cooperation.” In this context as well, this could strengthen cooperation with the United States, which is home to the second-largest Jewish population in the world.
Timeline and Budget
The implementation timeline for the plan is divided into three stages. The government has defined 2026 as a year for planning and public consultation, 2027 as the year for full budgetary activation, and the years 2027 to 2033 as the five main years of implementation for achieving the infrastructural and operational targets.
The government has also determined that the working plans for each component of the program, such as the national institute, infrastructure, human capital, and data, will be published for public comment by mid-October 2026. The Head of the Headquarters is required to submit for government approval, as part of the 2027 state budget, a detailed work plan for the coming years.
In terms of budgeting, the plan relies on previous allocations, for example, a recent government decision detailing a budget of approximately NIS 2.6 billion for the expected national program, but states that the updated version of the plan will include requests for additional resources required to implement the expanded targets, particularly the goal of 100,000 GPUs.
The plan does not define a budgetary breakdown between its various components. It explicitly states that it “does not include a budget” at this stage, prior to the submission by the Head of the Headquarters of a detailed work plan.
Responding to the State Comptroller’s Requirements – Except Regulation
The criticism by State Comptroller Matanyahu Englman regarding Israel’s preparedness for artificial intelligence evolved from an “alarm call” in November 2024, in which it was stated that “too little and too slowly was being done,” into a more detailed and pointed report published about two weeks ago.
The new report examined Israel in comparison with 11 European countries. It concluded that despite the establishment of the National AI Headquarters, the state has still not succeeded in “translating the national advantage into a comprehensive government action plan,” and that the gap between potential and implementation is widening.
The national artificial intelligence plan, approved last week, constitutes an ambitious response to such criticism. While the Comptroller warned of a “gap between budgeting and implementation” and the absence of a “long-term national plan” anchored in clear governmental responsibility, the government has now approved a full roadmap extending to 2031, transferring leadership to the National Headquarters in the Prime Minister’s Office.
With the creation of the national plan, Israel has responded to most of the Comptroller’s recommendations at the planning level, including a significant expansion of computing infrastructure, treating data access as a national asset, and broad development of AI literacy.
However, the Comptroller’s demands for “risk-adapted regulation” and comprehensive legislation similar to the European model received only a partial response.
