Views: 47
The AI Shovel Paradox: Why Israel’s AI Future Lies in Software and Silicon, Not Server Farms
The oldest cliché in Silicon Valley is that during a gold rush, the people who make the most money are the ones selling shovels.
A chart published by Bank of America Global Research over the past week seemed to prove the point. It projects the combined free cash flow (FCF) of the major hyperscalers—Amazon, Microsoft, Alphabet, Meta and Oracle—falling below zero for the first time as they pour hundreds of billions of dollars into AI infrastructure. At the same time, the combined FCF of leading semiconductor companies continues to climb sharply.
The chart quickly became a Rorschach test. AI bears saw proof that Big Tech is overbuilding. AI bulls saw evidence that chip companies are becoming the most profitable businesses ever created. Both interpretations miss the more interesting point.
The economics of AI are shifting. Today, more of the value is flowing to the companies supplying the infrastructure than to the companies financing it.
But there’s an important nuance. Nvidia doesn’t sell AI—it sells the ability to deliver AI. Semiconductor revenues are ultimately a function of hyperscaler capital expenditure, not AI adoption itself. AI usage could continue growing for years while infrastructure spending slows simply because Microsoft, Amazon or Meta decide they have already built enough capacity. Infrastructure cycles are governed by annual capital budgets; software adoption compounds over much longer periods.
For Israel, that’s the real lesson. The opportunity isn’t to predict when the infrastructure cycle peaks. It’s to build the technologies that remain valuable regardless of whether hyperscaler capex grows by 70% or 20%.
Israel’s AI paradox
Bank of America’s latest Global Research report places Israel firmly in the next tier of AI economies behind the US and China. The rankings reveal a country with exceptional intellectual capital but clear physical constraints.
Israel ranks #1 globally in AI talent concentration, with AI professionals representing 2.1% of LinkedIn members. It ranks #5 in cumulative private AI investment, with roughly $19 billion invested between 2013 and 2025, #5 in the number of AI companies, and top 10 globally in long-term AI potential. The report also highlights Israel’s entrepreneurial ecosystem and R&D intensity as key competitive advantages.
The latest AWS Engines of Growth report reaches a similar conclusion from a different angle. Based on a survey of more than 3,400 founders across 20 countries, it found that 31% of Israeli startups founded in the past five years are AI-native, the highest proportion in the study, ahead of the United States. Rather than simply adding AI features to existing software, Israeli founders are increasingly building companies where AI sits at the core of the product from day one.
These rankings reinforce something Israel has demonstrated repeatedly over the past three decades. Our advantage has never been industrial scale. It has been talent density.
A small geography brings researchers, founders, engineers, multinational R&D centres, investors and early customers into unusually close proximity. Knowledge moves quickly between universities, military technology units, startups and global technology companies. That concentration of expertise continues to produce disproportionately important companies in cybersecurity, developer tools, enterprise software and semiconductors.
The problem is that AI increasingly rewards not only intelligence, but infrastructure.
The infrastructure gap
The same Bank of America report shows where Israel falls behind. While it ranks among the world’s strongest AI ecosystems for talent and entrepreneurship, it drops to #12 in short-term AI readiness, #19 in energy readiness and #22 in data-centre costs.
Those rankings reflect structural realities rather than temporary weaknesses. Israel lacks abundant low-cost energy, large-scale industrial capacity and the natural resources required to compete in the infrastructure arms race now unfolding between hyperscalers and governments. Countries such as South Korea, Canada and the UAE enjoy advantages that are difficult to replicate through venture capital alone.
None of this argues against sovereign AI infrastructure. Israel should absolutely invest in domestic compute for defence, research and national resilience, and the government’s recently announced AI strategy is an important step in that direction. But national infrastructure policy is not the same as venture investing. Israeli startups are unlikely to outperform Microsoft or Google by building hyperscale data centres or training frontier foundation models. Their comparative advantage lies elsewhere: in the software, silicon and specialised technologies that make global AI infrastructure faster, cheaper and more efficient.
That distinction matters because it defines where Israeli venture capital can create asymmetric returns.
Build the shovels, not the mines
Israel has spent decades building capabilities that fit remarkably well with the current phase of AI.
Companies such as Mellanox, Mobileye, Habana Labs and Tower Semiconductor demonstrate that Israel can become strategically important without owning the largest manufacturing base. Mellanox became critical to modern data-centre networking before its acquisition by Nvidia. More recently, Tower Semiconductor announced a $3 billion expansion in Japan, supported by government incentives, to increase production of silicon photonics for AI and data-centre connectivity. The model is instructive: Israeli engineering, global manufacturing.
The next decade of AI will require much more than larger GPUs. It will require faster networking, optical interconnects, specialised processors, memory optimisation, advanced packaging, inference acceleration, security and power management. These are precisely the kinds of engineering problems where Israeli companies have historically excelled.
The opportunity extends beyond hardware.
As foundation models become increasingly commoditised, value shifts further up the stack into orchestration, evaluation, observability, AI security, governance and vertical applications. Enterprises will use multiple models simultaneously, choosing different models based on cost, latency, privacy and performance. They will need software that routes workloads, evaluates outputs, manages permissions, protects proprietary data and integrates autonomous agents into existing business processes.
Those software layers are far less dependent on the capex cycle than chip manufacturing. They benefit from AI adoption regardless of whether hyperscalers increase infrastructure budgets by 70% or pause spending for a year. That makes them particularly attractive for Israeli founders building long-term businesses.
A playbook for Israeli VCs
If Israel’s comparative advantage is intellectual property rather than industrial scale, venture investors should align their capital accordingly.
- Avoid competing with hyperscalers on infrastructure ownership. Capital-intensive businesses built around owning compute, operating large data centres or training general-purpose frontier models face structural disadvantages in Israel. That does not make infrastructure unimportant, but it does make it a difficult place for venture-backed startups to generate differentiated returns.
- Double down on silicon and infrastructure technology. Chip architecture, silicon photonics, networking, specialised processors, inference optimisation, memory technologies and advanced packaging all allow Israeli companies to participate directly in the global AI buildout without replicating the industrial footprint of Taiwan, South Korea or the United States.
- Own the orchestration layer. As enterprises move from experimenting with AI to operating fleets of models and autonomous agents, orchestration, evaluation, observability, AI security and governance become increasingly strategic. Israel’s exceptional concentration of AI talent makes it well positioned to build these control layers rather than another commodity application.
- Focus on sectors where technical depth matters more than scale. Cybersecurity, defence, healthcare, industrial software, financial services and robotics reward proprietary technology, domain expertise and trust. These markets are less likely to be won by whoever owns the largest compute budget and more likely to be won by whoever builds the most reliable solution.
For Israel, the answer looks increasingly clear. Our future won’t be determined by the number of GPUs installed within our borders or the size of our data centres. It will be determined by how much of the global AI stack is designed here.
Countries with abundant energy and industrial capacity will build many of the world’s AI factories. Israel doesn’t need to win that race.
It should focus on designing the chips, networking, security, orchestration software and specialised applications that make those factories work. In the AI era, the biggest opportunity may not be owning the mine—it may be designing the shovels everyone else depends on.
He is a former General Partner at Google Ventures (GV) in Europe, former head of Google for Entrepreneurs in Europe, and founding head of Campus London, Google’s first startup hub. Eze writes on Israeli tech, venture capital, artificial intelligence, and founder strategy.
He is also the founder of Techbikers, a nonprofit that brings together the startup ecosystem on cycling challenges in support of Room to Read.