Opinion
The next great opportunity for Israeli startups is hidden inside the AI supply chain
The race to build frontier AI models is increasingly concentrated among a handful of giants. The race to remove the bottlenecks holding AI back is far more open.
Conventional wisdom says the future of AI belongs to whoever builds the smartest model. People picture a handful of US tech giants in a race to refine their ever-complex tech. What's often neglected is the infrastructure on which the models run.
To put it in simpler terms: everyone is talking about Claude. Few are talking about the miles of fiber, the enormous power consumption, the advanced chips, cooling systems, and security infrastructure required to make Claude possible. If AI is the story of our era, then the AI supply chain is the story behind the story.
But the AI supply chain is struggling to keep pace with the rapid advances in AI models, creating bottlenecks at nearly every layer of the infrastructure stack. Solving these constraints will require an infrastructure rebuild cycle unlike anything we have seen in decades. And Israel is uniquely positioned to help lead it.
Israel’s technology ecosystem has spent decades developing expertise in many of the exact technical bottlenecks now emerging across the AI infrastructure stack. This expertise did not emerge by accident. Intel's largest development center outside the United States is located in Israel, while NVIDIA, Apple, Amazon, Google, and Microsoft all maintain major engineering operations here. Together, they have helped build one of the world's deepest concentrations of talent in chip design, networking, and the underlying technologies that power modern computing.
One of the biggest bottlenecks is moving data. As AI clusters grow larger, the challenge is no longer just computing power — it is getting information quickly and efficiently between thousands of chips. Existing networking technologies are approaching their limits in terms of speed, power consumption, and heat, creating a major opportunity for new infrastructure technologies. As network speeds continue to increase, moving data reliably over traditional electrical connections becomes increasingly difficult due to signal degradation, power requirements, and thermal constraints.
These challenges are accelerating industry investment in silicon photonics, advanced RF technologies, and next-generation interconnect architectures designed to deliver higher bandwidth with greater energy efficiency.
The same is true across the stack. Memory bandwidth is increasingly becoming a limiting factor for both AI training and inference workloads. Power delivery and cooling are emerging as critical constraints for next-generation data centers. GPU utilization often remains lower than it should be because orchestration, scheduling, and infrastructure management software have not fully caught up with the scale and complexity of modern AI systems. Security architectures designed for human users are struggling to secure AI agents, machine identities, and autonomous systems operating in real time.
Every one of these bottlenecks creates opportunities for new companies, and many of them align remarkably well with Israel's historical strengths in semiconductors, networking, cybersecurity, infrastructure software, and deep engineering.
The opportunity extends beyond hardware
While foundation models have captured most of the attention, AI-native infrastructure software is emerging as one of the most important categories of the AI era. As AI systems become increasingly autonomous, organizations will require entirely new orchestration, observability, security, and governance layers. The challenge is no longer just building smarter AI. It is building the systems that can control it, secure it, and operate it reliably at scale.
This aligns naturally with Israel’s long-standing strengths in solving complex infrastructure bottlenecks through deep engineering and technical innovation.
Importantly, many of the most attractive opportunities may not look like traditional “AI startups.” Some will resemble infrastructure companies. Others will look like semiconductor enablers, networking companies, developer tooling platforms, or industrial systems businesses. But they will all share one commonality: they will be critical to enabling AI at scale.
The race to build frontier AI models is increasingly concentrated among a handful of giants. The race to remove the bottlenecks holding AI back is far more open. And history suggests that the companies that become essential to an ecosystem often create more enduring value than the companies that dominate the headlines.
Needless to say, providing this value has led to Israel's greatest tech successes.
Everyone is looking at the AI on their screens. The bigger opportunity may lie in the infrastructure behind the curtain, and in the companies solving the bottlenecks that make AI possible at scale.
Lian Swirsky is an investor at Ibex Investors.
