Israel's AI Czar Courts Chipmakers and Data Centers to Defend Its Tech Edge
Israel built its reputation as the "Startup Nation" on software, cyber and the fabless design work that turns Tel Aviv office parks into chip laboratories for the likes of Nvidia and Intel. Now the country's newly appointed artificial-intelligence chief wants something it has never had: a leading-edge fab of its own, and enough domestic compute to keep its defense establishment and its startups from depending on foreign hardware. The pitch is explicit and, by the standards of Israeli officialdom, unusually blunt about what is at stake.
Brigadier General (Res.) Erez Askal, who leads Israel's newly created National Artificial Intelligence Directorate inside the Prime Minister's Office, told Bloomberg he is in active talks with Taiwan Semiconductor Manufacturing Co., Samsung Electronics and Intel to build a semiconductor fabrication plant on Israeli soil capable of producing chips at two nanometers or smaller. Alongside the fab, he is pitching a data-center cluster able to support 100,000 GPUs within five years — a scale that would put Israel roughly on par with the largest AI "gigafactories" the European Union has planned.
The plan's specifics
The chip and data-center push sits inside a broader National AI Program that Israel's cabinet approved on June 16, 2026, on Prime Minister Benjamin Netanyahu's recommendation. The program spans human capital, "Cyber AI," "Physical AI" and a national quantum-computing effort, but its most expensive component is sovereign compute. Analysts and experts have pegged the infrastructure cost at $20 billion to $30 billion or more, though the official resolution does not commit to a specific top-line figure.
Askal's directorate is not starting from zero on the data-center side. A $300 million AI data center opened in Modi'in in October 2025 with 4,000 Nvidia GPUs, and about a quarter of that facility's capacity is earmarked for a national supercomputer run under the Israel Innovation Authority. In February 2026, the cabinet approved fast-track measures stripping away regulatory barriers that had been slowing construction, and roughly 1 gigawatt of new electricity capacity for data centers is already being built.
The directorate itself runs on a budget of about $1 billion, which Askal plans to scale to $1.5 billion by 2027. His strategy leans on tax incentives and on positioning the Israeli government as an anchor customer for compute — a way to guarantee demand and de-risk the projects for private and foreign investors.
The fab is the harder ask. Israel has hosted Intel manufacturing for decades, but leading-edge capacity has stalled: Intel's planned "Fab 38" expansion has been suspended, and the country has never produced chips at the two-nanometer frontier now dominated by TSMC. Askal is candid that Israel has been left behind here.
What officials are saying
Askal has described the country's failure to secure a cutting-edge fab as a "very big failure," an admission that frames the current courtship of TSMC, Samsung and Intel as catch-up rather than expansion. He has also acknowledged direct competitive pressure from Gulf states, particularly the United Arab Emirates and Saudi Arabia, which are pouring petrodollars into AI infrastructure with fewer land and energy constraints than Israel faces.
On the data-center reforms, Netanyahu called the February fast-track approval a "major breakthrough," language that signals how central compute has become to the government's economic and security agenda. The prime minister's office has cast the overall program as a bid to make Israel "one of the world's leading countries in artificial intelligence" while strengthening technological independence and national resilience.
The sovereignty race Israel is joining
Israel is late to a game that a dozen governments are now playing at once. The logic driving Askal's directorate — that a nation cannot be a security or economic power in the 2020s if its AI runs entirely on someone else's chips in someone else's country — is the same logic behind the EU's gigafactory program, the U.S. CHIPS Act subsidies that drew TSMC to Arizona, and the sovereign-compute funds standing up across the Gulf. TSMC alone still fabricates roughly 90% of the world's most advanced logic chips, almost all of it on an island in one of the planet's most contested regions. Every fab that comes online elsewhere, whether in Arizona, Dresden or potentially Israel, chips away at that concentration risk.
For Israel the calculus has an extra edge. Its defense and intelligence agencies are among the most demanding AI users in the world, and the same U.S. export controls that restrict advanced chips to rivals can also complicate supply to allies. A domestic fab and a national GPU cluster would insulate both the military and the country's dense startup ecosystem from foreign supply shocks. But Israel cannot match the Gulf on raw scale — it has a fraction of the land and far tighter power. Askal's answer is specialization over size: high-value, leading-edge capacity rather than sprawling commodity capacity.
The constraints are unforgiving. Israel's Electricity Authority is sitting on a backlog of roughly 27 gigawatts in data-center connection requests against the 1 gigawatt now under construction — demand outrunning supply by more than an order of magnitude. Grid limits and security risks complicate both the data centers and any leading-edge fab, and no chipmaker has yet committed to building one.
What to watch
The decisive signal will be whether any of the three courted chipmakers — TSMC, Samsung or Intel — publicly commits to a two-nanometer fab in Israel; so far the talks remain talks. Watch, too, for how fast the directorate's budget actually climbs toward its 2027 target, whether the government's anchor-customer model draws private capital, and whether Israel's grid can be expanded quickly enough to make 100,000 GPUs anything more than a slide in Askal's pitch deck. In a region where the UAE and Saudi Arabia are writing far larger checks, Israel's wager is that quality, defense integration and speed can substitute for money and megawatts. The next year of fab negotiations will show whether that bet holds.