Israel's AI Vision: Four Plans, Flashy Appointments, and Few Results

In recent weeks, as elections approach, the Prime Minister's Office has been promoting the discourse on the AI headquarters established within its office over the past year. At the beginning of the month, Prime Minister Benjamin Netanyahu held a festive press conference where he launched the headquarters' plan and introduced its head for the first time, Brigadier General (res.) Erez Askal, an intelligence officer who previously served as commander of the 9900 mapping unit.

The declarations were promoted, among others, by his personal advisor, Aleph venture capital fund partner Michael Eisenberg, on his social media accounts. In recent days, the Prime Minister even shifted gears on the matter by introducing the new candidate for the 11th spot on the Likud party list, Oren Dobronsky, who has lived in Israel for several years after a period in Silicon Valley and is credited with an exit in the toolbar sector. Dobronsky, a venture capital investor known in recent years primarily for his participation in the 'Shark Tank' program on Keshet 12 and owner of an American hummus chain, was presented in the Likud announcement as the person who will manage the AI sector in the next government.

On Wednesday, the national AI headquarters shifted gears and presented the details of the plan at a special press conference held at the headquarters' new offices in a flashy building in Ramat Gan, led by Askal, surrounded by new employees who arrived with him from the ranks of the security establishment, the Prime Minister's Office, and the Accountant General's office in the Ministry of Finance.

However, behind the festive announcements, flashy appointments, and polished packaging, a cautious and significantly poorer plan is revealed compared to the promises of the past. Examining the details of the new initiative shows that instead of a national revolution that will launch Israel to the forefront of global technology, the new headquarters presents mainly minor steps, leaving the heaviest questions of the AI industry unanswered.

Main points of the plan presented by the national AI headquarters this week:

Introducing AI studies to middle school: Acquiring artificial intelligence skills at an early age in accordance with OECD country practices. However, this is only 30 hours per year.
Purchasing 5,000 processors for Israeli supercomputers: The purchase is intended to prevent external dependence and boost computing power in Israel. About 1 billion shekels will be allocated for this (in addition to the 1.1 billion shekel budget already approved by the government).
Developing an Israeli language model: Discussions are still taking place in the government as to whether the model is indeed needed and who will develop it, while Israeli AI companies are being sold or crashing.
Tender for establishing a quantum computer: The goal is to equip the industry with a cloud system that will provide access to powerful quantum computers. The tender lacks budget details.
City for robotic AI experiments: Establishing a kind of neighborhood where experiments with robotic technology will be carried out in a hospital, school, and other public institutions built specifically for this purpose.

At the heart of the plan: AI lessons and building a quantum computer

Among the initiatives in the plan: AI lessons for 8th and 9th grade students in the amount of 30 annual hours; returning 100 scientists to Israel (after three have committed to return so far); publishing a letter of intent to establish a 'physical AI city', where a complex of a hospital, school, and other public buildings will test technologies in the field of robotics; a tender for establishing a quantum computer with an unknown budget; a general intention to purchase up to 100,000 Nvidia processors within a few years in an undisclosed financial scope; and establishing a special institute with a budget of tens of millions of shekels that will coordinate projects in cooperation with the private sector, to be headed by Orit Korman, former head of the planning department in the C4I Directorate.

This is a very minor plan whose clauses are general and it does not include budget details, compared to the grandiose plan presented last year by Professor Yaakov Nagel — the person asked by Netanyahu to formulate a plan for the defense budget and then also to formulate a strategic plan for the AI headquarters before Askal's appointment.

Nagel supported the establishment within two years of a supercomputer with a budget of 18 billion shekels equipped with 60,000 Nvidia chips, which would be used by academic researchers and startup founders for AI computing; a plan for returning brains and developing AI researchers with a budget of 2 billion shekels; building training programs in the field in high schools and the army with a budget of 1 billion shekels; and establishing an AI institute and building a language model in Hebrew. The total recommended plan: 25 billion shekels spread over five years, in installments of 5 billion shekels per year.

The headquarters, which was supposed to be established with the submission of the Nagel committee's conclusions a year ago and was supposed to start investing billions of shekels as early as October 2025, presented its new plans only now — as the election day bells rang. In fact, it barely touched on the central issues on the agenda: there is no significant reference to building that supercomputer, nor was a sum or location set for its establishment (a debate is raging in government offices whether to build it in a foreign country or in Israel); the headquarters did not decide whether to establish an original Israeli language model; and no solution was presented to the problem that occupies the AI industry and has put it in a turmoil: the temporary suspension of server farm connections to the power grid, which has halted the planning and construction of dozens of server farms on the agenda.

Senior officials formulated plans, but most recommendations were shelved

If this story sounds familiar to you, it is because this is the fourth time in the last seven years that Netanyahu's governments have presented a national AI plan, with no results so far. In 2018, Netanyahu appointed Professor Itzik Ben-Israel and Professor Eviatar Matania to formulate the first plan, after the two proved themselves in building a similar plan for promoting cyber. The plan recommended allocating 10 billion shekels to the field and was ready for action as early as 2019, but due to the multiple election cycles that year, it was submitted only in 2020, and was effectively shelved during the political upheavals.

The Innovation Authority took it upon itself to promote the issue and for this purpose concentrated the voluntary TELEM forum. At the end of that year, Dr. Orna Berry, former Chief Scientist and senior executive at Google Israel, was hired by the forum with the goal of formulating another plan, and Berry recommended an investment of 5 billion shekels in the field. Only a few of her recommendations were eventually adopted, and with a budget of only 1 billion shekels.

The development of the Hebrew language model that Berry recommended was shelved after an investment of many millions after OpenAI launched ChatGPT in 2022, and the supercomputer she recommended was eventually built under the Innovation Authority with a meager budget of only 160 million shekels, with a fairly low amount of graphics processors that is far from satisfying the needs of the industry and academia (about a thousand 'Blackwell' processors). Today, according to estimates, the capacity of the national supercomputer in Modiin stands at 4,000 graphics processors rented from the Nevius company — far from the target set by Nagel of 60,000 processors.

Last year, the TELEM forum, which until then held most of the AI activity, became a thorn in the side of the Prime Minister and officials in the Accountant General's office in the Ministry of Finance, and together with his then-director general Yossi Shelley — the establishment of a national AI headquarters was promoted. Dubi Frances, also a star of the 'Shark Tank' program and an investor who started, like Dobronsky, with investments alongside Oren Zeev, advised on the establishment of the headquarters but was barely involved in it in the last year. After him, Professor Nagel was brought in to prepare a master plan.

Meanwhile, key figures who were part of the plan within the framework of the Innovation Authority, such as Professor Yoav Shoham, one of the senior lecturers at Stanford on artificial intelligence and one of the founders of AI21 who was the scientific advisor to the plan, or Ziv Katzir — head of the TELEM forum at the Innovation Authority, resigned from their positions.

'Too little and too late': Where did 3 billion shekels go

Askal is a celebrated officer who won Israel Security Awards for developments in the field of satellites while being commander of the 9900 mapping unit and was even responsible in the army for the technological upgrade system of the systems. However, he is considered to lack management experience in civilian bodies and has no experience in the high-tech industry or a scientific degree.

'The task assigned to him is so great: to promote the AI industry in Israel and at the same time encourage all the large systems — the Ministry of Education, Health, Transportation — to adopt AI. It's easy to lose your way like this, so even in the headquarters they are starting small with sporadic projects,' says a senior official in the industry. 'What the headquarters presented this week is too little, too late, and is a drop in the ocean.'

The headquarters' budget for 2026 stood at 3 billion shekels, with senior officials in the organization hoping that after the elections it will grow as defined in the Nagel committee to 5 billion shekels, but there is no guarantee that the headquarters will look as it does in the next government that will be formed after the elections.

In the plan revealed on Wednesday, the budget distribution was not presented and senior officials in the headquarters used vague terms of 'tens' or 'hundreds' of millions of shekels for the projects they presented. However, the largest expense item in the headquarters, the purchase of Nvidia-made processors for what is called the 'Israeli supercomputer', is still controversial. While Professor Nagel presented a request for 60,000 processors, the headquarters is interested in purchasing 100,000. This is, as mentioned, compared to only 4,000 currently serving the only civilian supercomputer that Israel has in Modiin through the Innovation Authority.

In the headquarters and the Ministry of Finance, they are adopting a realistic approach: purchasing 100,000 'Vera Rubin' processors from Nvidia (about 135,000 dollars per unit) is an astronomical expense of 13.5 billion dollars, high even by the terms of a multi-year plan. According to estimates, officials in the Accountant General's office in the Ministry of Finance and in the headquarters are settling for the purchase of 3,000-6,000 additional processors for government and academic needs, and encouraging private entities such as Mega Or or SDS of Yossi Sheinfeld to purchase tens of thousands of additional processors, even 60,000, for industry needs.

At this stage, the government plans to build the supercomputer in a foreign country. The decision was criticized mainly due to the creation of a dependence of a critical government resource on a foreign country. Supporters, however, explain that the government server farm can be built much faster than in Israel.

Meanwhile, the AI headquarters experienced for the first time the taste of a budget cut. The headquarters was supposed to receive a budget of 2.5 billion shekels for purchasing graphics processors (in addition to the 1.1 billion shekel budget, which was approved for the Planning and Budgeting Committee and then included in the headquarters' budget), but the Ministry of Finance informed the headquarters that the budget will not grow to 3.6 billion shekels but will be unified — which signals an increase of only about 1 billion shekels.

The Electricity Authority shuffled the cards in the industry

Even if the headquarters and the Ministry of Finance succeed in bringing tens of thousands more graphics processors to Israel, the server farm industry that is supposed to absorb them is currently in trouble following the Electricity Authority's decision to stop granting permits for connecting server farms for a 'rethinking' period of about four months.

The obstacle that the Electricity Authority placed before the industry shuffled the cards and led potential clients of server farm entrepreneurs to renegotiate contracts. Some of them, including new players who were interested in entering Israel — such as Coreweave — even backed out. Existing players were also harmed by the move — including Mega Or of Tzahi Nahmias, whose stock fell by 22% in the last month.

In the AI headquarters, they explain that they are showing involvement in the process and are confident that Israel will be able to provide 2 gigawatts to meet most of the demand. According to them, they are working together with the Ministry of Finance and the Electricity Authority to formulate new criteria for granting electricity permits, based, for example, on financial proof for building large and expensive server farms and on good relations with Nvidia that allow them to obtain the newest processors in the industry — Vera Rubin. In the industry, however, there is criticism that this is reactive activity and that it is not defined as core work in the AI headquarters, which is supposed to be an effective body under the personal auspices of the Prime Minister.

'There is a lack of coordination. It is embarrassing to potential partners'

Another step by the headquarters that arouses sharp criticism in the industry is the publication of the tender for Israeli quantum computing — without presenting a budget. According to the plan's details, the goal is to establish a computer that does not bet on a specific quantum technology (such as trapped ions or superconductors) but allows cloud access to quantum computing capabilities for the benefit of national research and development.

The criticism comes because a regulated national quantum program has already been operating in the Innovation Authority for several years to invest 1.1 billion shekels in national quantum technology — which itself published a call for proposals just a month ago to establish an Israeli quantum computer by 2029 with an initial budget of 100 million shekels for 2026 alone.

Officials in the industry express deep concern about the lack of coordination between the bodies, which leads to confusion among suppliers and potential partners facing duplicate authorities. According to them, this is a clear case where 'the right hand does not know what the left hand is doing', when the national headquarters enters the field with independent and uncoordinated actions, something that not only creates unnecessary duplication but 'embarrasses potential foreign partners', as a senior official in the quantum industry says anonymously.

While the government defined with the support of the Budget Department that the Innovation Authority will lead the quantum policy, with an intention to renew the national quantum program's budget by another 1.25 billion shekels for the next five years, the AI headquarters in the Prime Minister's Office is acting alone.

And while the government is spending huge fortunes on quantum computers through two arms that do not talk to each other, the Ministry of Finance's attitude towards the headquarters is also ambivalent. While in the Accountant General's office they take an active part in formulating the headquarters' activity, including the processor purchasing policy, in the Budget Department they find it hard to be impressed by the plan presented to the public, believing it is full of good intentions but high statements without coverage. Mainly, the department finds it hard to understand the duplication in responsibility for chip purchasing policy that is now split between the TELEM forum that represents MAFAT, the Planning and Budgeting Committee, the Ministry of Science and the Innovation Authority, and the new headquarters that currently employs about 20 people.

On the other hand, senior officials in the public sector who are familiar with the Ministry of Finance's work explain that 'the conservative policy in the Budget Department, in addition to political coalition pressures, leads to the fact that only a few projects are approved and those that are — they take them with both hands because that is what there is. The product is a collection of eclectic plans that are not necessarily implemented as part of a strategy.'

What will be the fate of the Israeli language model

It seems there is no better example of the state of the national AI plan than the initiative to develop an internal Israeli language model — not necessarily in Hebrew. The initiative has existed, as mentioned, since the days of Orna Berry's plan — which was shelved after the launch of ChatGPT in 2022.

In the new artificial intelligence headquarters, they published a call for proposals for developing the model, and they definitely talk about the need to develop an Israeli model that will not depend on Anthropic, OpenAI, and other foreign companies. This is to prevent dependence on these companies and prepare for the possibility of arbitrary decisions to block the access of Israeli institutions to AI models for political reasons in the future.

And yet, even today, after seven years of plans, the headquarters still does not know who will develop the model and whether it will be private companies — Israeli or foreign. Israeli startups like AI21 that developed their own models have weakened and crashed or were sold. The new Israeli hope — Descartes — is on the way to a giant purchase by Anthropic.

And so, while the Israeli government continues to formulate national plans and scatter festive announcements ahead of the elections, the local industry is forced to continue and look for answers in the private sector and international companies — far from the headquarters' promises.