A Strategy Built on Integration, Not Scale
On August 13, 2026, Israel's National Artificial Intelligence Directorate published its national AI strategic plan, the operational follow-through on Government Resolution 4255, "Accelerating Artificial Intelligence in Israel and Establishing Global Leadership," which the cabinet approved on June 16, 2026. The plan, branded internally as "Lived AI," makes an explicit strategic choice: rather than compete with the US and China to train the largest foundation models, Israel is wagering its advantage lies in being the first country to fully integrate AI into public life — healthcare, education, and government services — while leaning on its existing strengths in cybersecurity and physical AI.
The headline commitments are large. The Directorate, led by Brig. Gen. (Res.) Erez Askal, wants a national computing base of at least 100,000 advanced AI accelerators (starting with a dedicated 5,000), a domestically built quantum computer under a tender called Project Nexus, an advanced chip fabrication facility at 5-nanometer technology or below, and five regional "acceleration reactors" — labs anchored to existing centers of excellence, each paired with a business or government partner, pursuing goals from productivity to national security. New AI-security laboratories, built jointly with the National Cyber Directorate, will focus on protecting AI systems, defending against malicious AI use, and countering synthetic forgery (deepfakes).
The Case for a State-Led Push
There is a real argument for this kind of top-down infrastructure push, and it deserves a fair hearing before dismissal. Compute, chips, and quantum capacity are precisely the kind of long-horizon, capital-intensive infrastructure that markets underprovide on their own — the same logic that justifies state-backed investment in energy grids or fabs elsewhere. Israel's tech sector is genuinely strong (Google's acquisition of Wiz for $32 billion was the largest deal in Israeli high-tech history, and cyber investment hit $3.8 billion in 2024), but a small country competing against trillion-dollar US hyperscalers and a state-subsidized Chinese chip sector has a plausible case for coordinated public investment to avoid being permanently priced out of frontier compute. Pairing that with a National Ethics Council for algorithmic transparency in public services, rather than bolting ethics on after deployment, is also a sequencing choice worth crediting.
Where the Plan Strains Against Its Own Numbers
The steelman only goes so far, because the plan's own arithmetic undercuts its rhetoric. Askal told a Knesset committee in July 2026 that implementing the full national strategy would require roughly NIS 5 billion ($1.66 billion) annually. The Directorate's actual 2026 budget is NIS 3 billion — and according to Globes' reporting on the plan's rollout, the Directorate absorbed a budget cut, receiving NIS 1 billion instead of a planned NIS 3.6 billion increase. Reaching 100,000 processors at the plan's own estimated cost of roughly $13.5 billion is, in the framing of one industry source Globes quoted, "too little and too late; a drop in the ocean" relative to what has actually been funded. A more realistic near-term outcome, per that reporting, is 3,000–6,000 additional processors — a rounding error against the headline target.
This is not a new pattern. Globes notes this is the fourth Israeli government AI plan in seven years to promise large sums and deliver a fraction: 2019 and 2020 plans recommended NIS 10 billion and NIS 5 billion respectively, and each time only about NIS 1 billion was ultimately allocated. Unresolved questions compound the funding gap — the plan does not name a location or firm construction path for the supercomputer, has not decided whether to build an independent Hebrew-language model, and does not address grid-connection freezes that are already stalling data-center buildout in Israel.
The Governance Question Regulators Should Not Wave Away
A separate, more pointed concern involves how the Directorate itself was stood up. Reporting from Calcalist (Ctech) found that the Prime Minister's Office appointed Askal without a competitive tender, placed the Directorate inside the PMO rather than under the Ministry of Innovation, Science and Technology, and now controls its NIS 120 million budget directly rather than through Innovation Minister Gila Gamliel — with hiring and contracting exemptions layered on top, justified by unspecified security considerations. Calcalist also flagged advisory ties between the Prime Minister and tech entrepreneur Dovi Frances, who pushed the Nagel Committee's compute recommendations ahead of the government's decision. None of this proves misconduct, but centralizing AI infrastructure procurement inside a single office, exempted from normal tender rules, is exactly the kind of arrangement that invites the appearance of favoritism even when the underlying policy goal — faster compute buildout — is sound.
The Proportionate Read
Our editorial instinct favors letting states build infrastructure and set light-touch rules rather than pre-emptively regulate AI applications — and on that axis, "Lived AI" gets the emphasis right: deployment and compute over restrictive rulemaking. But infrastructure ambition needs matching fiscal discipline and procurement transparency, not a fifth iteration of the same unfunded mandate. If Israel wants the credibility that comes with a 100,000-accelerator national compute base, the Directorate needs a funded, auditable path to get there — and a tender process that survives scrutiny — rather than another five-year vision statement layered on a one-year budget.