It is hard to argue with Erez Askal's diagnosis. Israel is one of the strongest countries in the world in AI development, but almost all the computing power its researchers, startups and government ministries rely on is rented from foreign cloud companies. The day prices rise, US policy changes or someone decides Israel is not a priority customer, the whole industry will feel it. Sovereignty in compute is not a slogan, it is insurance.
But a national plan is measured by three things: how much money it gets, who is responsible for it and when it is supposed to be done. In Askal's speech, according to reports, there was none of these. Even the number of processors is unclear: some reports speak of 100,000, others of 300,000. That gap, in a plan meant to cost tens of billions of dollars, is not a technical detail.
And history is not encouraging. Israel has announced national programs in cyber, AI and quantum before, and many were partly funded, delayed or buried between governments. Meanwhile, countries like the UAE and Saudi Arabia have turned AI into a national project with multibillion-dollar budgets and direct partnerships with Nvidia, OpenAI and the cloud companies.
There is also a practical question the plan must answer: electricity. 2 gigawatts is a huge amount relative to the size of Israel's power sector, which already struggles to meet peak demand and the cost of building new plants. In the US, governors were elected last week after promising to curb data centers over power bills. In Israel, a giant data center near a city will raise the same questions, and a plan that does not address them in advance will get stuck.
The vision is right, and the timing is urgent. But if Israel wants AI sovereignty, it has to start with the boring part: a protected multi-year budget, a single body with authority to execute and a realistic power plan. Until then, the hundreds of thousands of processors will stay on a slide, and Israeli tech will keep renting its future from Amazon, Google and Microsoft.
