A Cabinet Decision With Clear Institutional Intent
On June 16, 2026, Israel's cabinet approved a National AI Program anchored by three core commitments: sovereign computing infrastructure targeting 100,000 processing units, a domestically built quantum computer, and a National AI Institute linking government, academia, and industry. Overseeing it all is a new National AI Directorate housed in the Prime Minister's Office and led by Brig. Gen. (res.) Erez Askal — former commander of Unit 9900, the IDF's elite visual intelligence unit, and former head of the IDF's Digital Transformation Directorate.
The $20–30 billion figure attached to the program represents analyst estimates for the computing infrastructure build-out, not a formally appropriated budget. The directorate itself received NIS 120 million for 2026, with 20 staff positions and NIS 13 million in annual operating budget from 2027 onward. The gap between those numbers and the infrastructure estimate is real and intentional — this is a planning framework and institutional commitment, not a signed check.
The Governance Choice That Matters Most
The most consequential decision in the June 16 resolution may be structural rather than financial: placing the AI Directorate in the Prime Minister's Office rather than the Ministry of Innovation, where prior AI policy had lived under the Bennett-Lapid coalition. PM-level placement gives the directorate cross-ministerial authority to move budget and cut coordination delays in ways a ministry-confined body cannot.
Calcalist's investigation into the directorate's formation raised legitimate accountability concerns: Askal was appointed without a competitive public tender, using an "urgency" exemption granted by the Civil Service Commission. Critics noted overlapping interests between policy advisors and commercial beneficiaries of the planned infrastructure rollout. These are the right questions for any governance structure operating at speed, particularly in a small country where tech entrepreneurs and policymakers overlap heavily. Institutional safeguards serve as protection, not friction.
That said, the Nagel Committee — which reviewed Israel's AI readiness — concluded that "Israel today is not in a good position to accelerate the field" as global competitors deploy hundreds of billions. Centralization at the PM level is defensible when the alternative is coordination deadlock that bleeds more time.
Why the 2021 Predecessor Casts a Long Shadow
Israel has been here before. In 2021, the government launched a five-year, NIS 5.26 billion National AI Program with moonshots including a national GPU cluster and regulatory sandboxes for healthcare, finance, and transport. By April 2025, only about one-fifth of that budget had been spent. No operational national compute cluster existed. The promised scientific computation center remained incomplete. Of a planned series of regulatory sandboxes, just two had been approved.
Israel Tech Insider summarized the record plainly: "Blueprints don't build power. Compute, capital, and clarity do." The 2021 failure was not one of ambition but of implementation architecture — diffuse authority, unclear ownership, no mechanism to enforce timeline accountability. The Nagel Committee's warning was delivered against the backdrop of a program that was already supposed to be solving exactly this problem.
What's Different — And Verifiably So
Several structural changes in the 2026 program are materially better than its predecessor. Compute allocations now have operational line items: 5,000 advanced GPUs will be accessible annually for research and industry from 2027 through 2032, and 1,000 Nvidia B200 accelerators have already been allocated — 70% to industry, 30% to academia. Specificity matters. The 2021 program's failures were partly a function of aspirational targets without operational-level commitments.
Askal's military background suggests someone accustomed to delivering capabilities under operational pressure rather than managing policy committees. Whether those skills transfer to peacetime procurement is an empirical question, but it is a different profile than prior leadership.
Israel's commercial AI ecosystem has also matured dramatically since 2021. Its 342 generative AI startups collectively raised over $20 billion. High-tech accounted for a record 18.3% of Israel's GDP in 2025, contributing roughly half of the economy's 2.9% expansion that year.
Labor Market: The Harder Policy Problem
Beyond infrastructure, the program confronts a social challenge that cannot be solved by procurement. The Jerusalem Post estimates between 1 million and 4 million Israeli workers will need partial or complete reskilling as AI diffuses across the economy. The June 16 resolution commits to a national labor market policy mechanism — appropriate in scope, but the design of that mechanism will determine whether it functions. Israel's track record in defense-to-tech workforce transitions is strong; mass civilian reskilling at this scale has no equivalent historical precedent.
Regulation: Israel's Real Comparative Advantage
On regulatory design, Israel is ahead of most peers. The government's December 2023 "Responsible Innovation" AI policy adopted a decentralized, risk-proportionate model — no sweeping horizontal statute, but sector-specific guidance built on existing ministerial frameworks. This deliberately contrasts with the EU AI Act's blanket compliance requirements. Regulators who create proportionate sandboxes attract experimentation; those who create compliance overhead attract legal teams.
The evidence suggests the approach is working. A 2026 study found 95% of Israeli tech workers use AI regularly and 78% use it daily. On Anthropic's AI Usage Index, Israel ranks first globally — its users engaging with AI nearly five times more intensely than their population share would predict. That organic adoption rate is the foundation the national program needs to build on.
A Bet Worth Taking — If Implementation Follows
The 2026 program is architecturally sounder than its 2021 predecessor. PM-level authority, military-grade execution leadership, and specific compute line items all reduce — without eliminating — implementation risk. But Israel's AI program history means the credibility gap is real. The genuine test arrives when annual Knesset budget cycles force trade-offs between AI infrastructure, defense expenditure, and social programs under ongoing fiscal pressure.
Israel's core bet is that excellence, speed, and domain depth in cyber-AI, physical AI, and deepfake defense can compound advantages that raw compute volume cannot replicate. That is how Israel built a world-leading cybersecurity industry from a similar starting point. Whether it works again depends entirely on whether the new directorate executes rather than plans.