What if Israel is competing in the wrong AI race?

For the last few years, much of the AI race has been measured by models.

Who has the smartest one? Who has the most compute? Who can spend the billions required to train the next generation? Much of that competition is concentrated among American companies such as OpenAI, Anthropic, and Google, alongside a growing field of Chinese competitors.

At a geopolitical level, it has become another dimension of the technological rivalry between the United States and China.

Israel is unlikely to win that race, but recently I’ve become convinced it doesn’t need to. Earlier this month, I attended Israel’s first conference dedicated to Physical AI, a relatively new term for taking AI out of our screens and putting it into machines that can perceive, navigate, and operate in the real world.

Think autonomous vehicles, drones, or robots working in a factory.

Before one panel, I spoke with Ezra Gardner, CIO and co-founder of Varana Capital, who has spent nearly two decades investing in Israeli deeptech and hardware. “Israel is actually perfectly built to take advantage of physical AI, and I’ve been calling it ‘robotics’,” he told me.

I’d heard this thesis a few times in my reporting for JNS, and in conversations on my podcast, The Spiro Circle.

Consider this: When I interviewed Tal Cohen (who was also at that conference), we discussed what he calls the “deployment gap”: the distance between increasingly capable AI and our ability to deploy it reliably in consequential environments.

More recently, my exit interview with outgoing Israel Innovation Authority CEO Dror Bin explored how and why Israel had resisted calls to pour resources into developing its own LLM. “Developing a large foundation model requires hundreds of billions of dollars, huge infrastructure, energy — which a tiny country like Israel cannot do,” he said.

Instead, Bin described a thesis the Authority had developed several years ago: “The world will have to go back to innovating in the real world, not just the digital world.” He mirrored what was said by other podcast guests, Arik Kleinstein and Amit Spitzer from Glilot Capital. “If somebody [in Israel] is now thinking about creating an LLM in 2026, I suggest not to do it.”

All different conversations, but they each appear to be pointing toward the same idea: Maybe we are defining the AI race too narrowly.

A model provides intelligence, but one of the biggest bottlenecks faced nowadays is turning that intelligence into action, which requires physical infrastructure: semiconductors, sensors, edge computing, energy, and the engineering required to make all of those things work together reliably.

“What is the foundation of Physical AI and robotics?” Gardner asked. “It’s the data centers and the semiconductors behind them.” It sounds much closer to areas where Israel has already spent decades developing expertise, from semiconductors and sensing technologies to autonomous systems. Its cybersecurity and defense industries also provide additional capabilities that could support the transition.

Israel didn’t invent the internet, but became exceptionally good at securing it. It didn’t invent the automobile, but Mobileye became a major player in helping cars understand the world around them.

And now in this AI era, perhaps Israel doesn’t need to create the intelligence at the center of every device, either. I’m beginning to think Startup Nation’s biggest opportunity is in building the infrastructure that allows that intelligence to interact with the world.

Gardner interprets the Israeli government’s National AI strategy ambition of becoming an “Edge Up Nation” as something considerably simpler. “Israel needs to go from cyber to robotics,” he explained. “That’s what they’re saying.”

According to the Israel Innovation Authority’s 2025 report, approximately 1,500 deeptech companies operate in Israel, having raised $28.6 billion between 2019 and 2025. Earlier this month, the Authority also announced a new incubator dedicated to Physical AI, robotics and industrial automation.

There are reasons to be cautious about this shift, however. Gardner said that from an investment point of view, robotics and hardware require more capital and longer development cycles than the software businesses around which much of the Israeli VC ecosystem has been built.

The two leading players, the United States and China, possess the industrial and financial resources needed to crown one of them the winner of the AI arms race. I’m not suggesting that Israel can replicate it (because it can’t), but its strategy must involve knowing what not to compete on.

There are four industries where Gardner believes labor shortages create particularly obvious opportunities for robotics: agriculture, construction, services such as hotels and restaurants, and military and government.

His advice to founders was simple. “Don’t try and imagine your own idea,” he advised. “The ideas are already there. Make a robot that picks fruits and vegetables. Make a robot that folds laundry. Make a robot that replaces a driver in a tank… If you build it, they will come.”

The United States and China may dominate the competition to build ever more powerful models. And Israel doesn’t need to beat them at everything. Startup Nation, after all, has succeeded by becoming uniquely important at particular layers of the technology stack.

The model race will remain between the giants. But Israel has a different race to run altogether.

–

I spend a lot of time speaking to founders and investors across Startup Nation. You can read my work on my independent platform and podcast, The Spiro Circle.