The recent clash between the Pentagon and Anthropic over the use of its models for military purposes sent shockwaves through the tech and defense worlds. If a private company can dictate operational terms to a superpower based on its internal policy disagreements, what hope do smaller nations have if they rely entirely on foreign AI? This brings us to “AI sovereignty“, the ability of a state to control its digital infrastructure, data, and core technologies without relying on external actors.
To borrow a definition from IBM:
AI sovereignty is an organization’s or nation’s capacity to control its artificial intelligence (AI) technology stack, including related IT infrastructure, data, AI models and operations.
But what does AI sovereignty actually look like in practice, and is it achievable for a country the size of Israel? To understand this, we need to look at how the modern AI stack is built and who currently controls it.
Where the World is Going: Escaping the Duopoly Stack
In the United States, AI is effectively nationalized by default. The world’s leading frontier labs are all headquartered in the U.S., ensuring near-full incentive alignment between the tech giants and the American government.
However, the rest of the world is realising that in the 21st century, sovereignty is measured not just in military might, but in data, computing power, and the ability to build and deploy intelligence at scale. Global powers are scrambling to ensure they are not left entirely dependent on either the U.S. or Chinese “AI stacks”, the soup-to-nuts chain of technologies ranging from chips to models to applications.
Only the US and China operate near full-stack AI ecosystems.
The Dependency Economy of AI (source)
They design the chips, control fabrication (directly or indirectly), run the dominant clouds, and build the frontier models.
While the United States hosts roughly 75% of global AI supercomputing performance, China has successfully built a parallel digital ecosystem, accounting for approximately 15% of global compute and aggressively pursuing state-backed AI sovereignty under the centralised control of the CCP. Notably, many of the world’s leading open-source models are now Chinese.
We are seeing middle and rising powers take aggressive action to carve out their own sovereign infrastructure:
- U.S. Hyperscalers are increasingly offering “quasi-sovereign” solutions, such as Amazon’s European Sovereign Cloud, which operates exclusively on European soil with European governance
- The UK recently launched a £500 million Sovereign AI Unit designed to build national capabilities while simultaneously signing strategic MoUs with frontier labs like OpenAI, Cohere, and Nvidia to ensure access and influence.
- India, under Prime Minister Narendra Modi, is actively focusing on equitable access and mitigating the sovereignty risks posed by foreign technological dependency, looking to build localized models and leverage platforms like its national Aadhaar system.
Within all this, Israel finds itself in a precarious paradox. Despite being a world-renowned “Startup Nation” and a powerhouse of innovation, the country remains deeply dependent on global infrastructure.
Defining Sovereignty Through Jensen Huang’s 5-Layer Cake
Nvidia CEO Jensen Huang recently conceptualized AI not simply as software, but as essential infrastructure, akin to electricity. He argues that AI operates as an industrial “5-layer cake,” converting raw materials into real-time intelligence:
- Energy: The foundational constraint. Intelligence requires power generated in real time, making energy the absolute ceiling on how much AI a system can produce.
- Chips: Processors designed to efficiently transform energy into massive computational power at scale.
- Infrastructure: The physical layer—land, cooling, networking, and the massive data centers that act as “AI factories” designed to manufacture intelligence.
- Models: The intelligent systems that understand language, biology, physics, and the physical world.
- Applications: The top layer where economic value is ultimately created, such as drug discovery platforms, industrial robotics, or autonomous vehicles.
Every successful application pulls on every layer beneath it, all the way down to the power plant that keeps it alive.
Where Israeli Technology Can Achieve Sovereignty
Israel is caught in a paradox: it is a global innovation powerhouse but remains deeply dependent on foreign digital infrastructure. To transition from a “Startup Nation” to a “Systems Nation,” Israel must attack the sovereignty problem across Huang’s five layers:
- Layer 1: Energy (The Strategic Bottleneck) Israel currently faces an energy and infrastructure bottleneck, with a growing gap between electricity demand and grid development. True digital sovereignty requires an integrated national policy that treats AI data centers as strategic electricity consumers, prioritizing them for grid connections. Without energy sovereignty, there is no AI sovereignty.
- Layer 2: Chips (Subsidies and Strategic Alliances) Israel cannot easily build its own advanced GPU fabs and will remain dependent on global semiconductor supply chains. However, it can secure access. The Israeli government’s Telem program is directly subsidising computing resources, allocating 1,000 Nvidia B200 accelerators for local high-tech firms and academia. By subsidising these resources, Israel ensures that startups can train models locally and retain their intellectual property. In addition, Mellanox, acquired by Nvidia is largely accredited with being the force behind GPUs. It’s an Israeli startup, and several new challengers are emerging (some still in stealth) to develop the new generation of AI chips.
- Layer 3: Infrastructure (The Sovereign Data Center) This is where tangible shifts are happening. To handle sensitive government and defense data, Israel must build isolated, sovereign environments. A prime example is DREAM, an AI company that recently opened Israel’s first sovereign AI data center near Modi’in. Built with NVIDIA B200 clusters, this facility allows national agencies and critical infrastructure operators to train and deploy AI within a fully controlled, localized national environment, meeting strict regulatory and compliance requirements.
- Layer 4: Models (Open Source & Domain-Specific) Competing with U.S. frontier labs to build general models from scratch is prohibitively expensive. Instead, Israel can achieve sovereignty at the model layer by mastering the secure deployment of open-source models within sensitive environments. Additionally, companies like DREAM are developing their own proprietary, domain-specific models tailored for cybersecurity, healthcare, and defense, bypassing reliance on public foundation models.
- Layer 5: Applications (Startup Nation’s Edge) This is where Israel already excels. The local ecosystem is aggressively building AI applications that create structural, productivity-led growth. By building cutting-edge applications on top of a more sovereign infrastructure stack, Israel ensures that its intellectual property remains in-house and secures its economic resilience and export power.
Judging by the AI resilience index by country, Israel has its work cut out:
AI is becoming critical infrastructure
Achieving AI sovereignty across this 5-layer cake is no longer just an economic goal; it is a national security imperative. Local AI-driven productivity has the potential to expand Israel’s economic output, while control over strategic technologies changes the terms of international partnerships, granting Israel the ability to negotiate from a position of strength and independence. As the standoff between the Pentagon and private AI labs demonstrated, relying entirely on the benevolence of foreign tech giants is a risk no nation can afford to take.
I'm a former general partner at google ventures, head of Google for Entrepreneurs in Europe and founding head of Campus London, Google's first physical hub for startups.
I'm also the founder of Techbikers, a non-profit bringing together the startup ecosystem on cycling challenges in support of Room to Read. Since inception in 2012 we've built 11 schools and 50 libraries in the developing world.
- The Global Race for AI Sovereignty: Where Does Israel Stand? - March 16, 2026
- The Web’s Next Customer Isn’t Human - March 13, 2026
- Weekly Firgun Newsletter – March 13 2026 - March 13, 2026
