scale up house desktop
scale up house
Doron Aaronsohn
Opinion

Israeli founders should build deeper

"The new golden age of Israeli tech will not be built on another dashboard," writes Doron Aaronsohn, CMO at Sela. "It will be built on technology that is genuinely difficult to copy."

Wall Street keeps swinging between fear and relief over what AI will do to software. The lesson for Israeli founders does not change with the stock prices: the moat has moved down the stack, to complexity.
In early October, it looked as if US software stocks had put the worst behind them: on October 6, the S&P 500 software and services index climbed to its highest level since November 2025, after posting its best quarterly gain since 2020. Only a few months earlier, the same index had lost more than 26% between late January and April, in a selloff traders nicknamed the "SaaSpocalypse," driven by fears that AI would make traditional software obsolete.
1 View gallery
דורון אהרנסון סמנכ״ל שיווק גלובלי של חברת sela
דורון אהרנסון סמנכ״ל שיווק גלובלי של חברת sela
Doron Aaronsohn
(Niv Kantor)
The relief did not last. At the end of October, ServiceNow had one of the worst trading days in its history after its quarterly results revived fears that AI will erode its business, and within about a week its shares were down more than 30%. Fortinet fell about 20% the same day after cutting its service revenue forecast, pulling Palo Alto Networks and CrowdStrike down with it, and Salesforce slid as well. Within a few weeks, some of the same companies that had led the recovery were leading the decline.
For investors, this is a volatile market - and not every drop was about AI; Fortinet's reflected a slowdown in service revenue. For founders, the swings point to a question that no quarterly report will settle: where in the technology stack is value still hard to copy?
When a product becomes a feature
If the market is questioning the software giants, with their massive installed bases, years of proprietary customer data and global distribution, a startup founded today faces even tougher economics. Generative AI has made interfaces and business logic cheap and fast to build. Products that once took years and large engineering teams to build can now be shipped in a fraction of the time and cost. That cuts both ways: if you can build your product in a quarter, so can ten competitors, and so can the incumbent whose customers you are trying to win. A thin application layer on top of someone else's model or cloud is no longer a company. It is a feature waiting to be absorbed.
The question investors ask has changed accordingly. It is no longer "how many features does the product have?" but "what in this architecture is genuinely hard to replicate?"
Follow the money: the value is moving down the stack
Israel's own deal flow tells the story. More than 50 M&A deals involving Israeli startups and tech companies were announced in the first half of 2026, and the biggest buyers were paying for depth, not for interfaces. Palo Alto Networks paid a reported $400 million for Koi, a two-year-old company that secures the AI agents, plugins and software running on employees' devices. Cisco acquired Astrix, which secures non-human identities, for about $400 million. Two Israeli makers of optical interconnects, the hardware that moves data between AI chips, drew major buyers as well: Credo agreed to acquire DustPhotonics for up to $1.3 billion, and Molex agreed to acquire Teramount for $430 million.
The funding market is pointing in the same direction. Of the $2.77 billion Israeli startups raised in September, nearly 70% came from five mega-rounds, including Cyera ($400 million at a valuation above $12 billion), Island ($400 million for its enterprise browser and AI-agent control layer) and Upwind ($300 million in cloud security, since expanded to $400 million). Even the month's largest round, Wonderful's $550 million for enterprise AI agents, is a bet on deep integration into customers' systems rather than on a chat window. What these companies have in common is not a clever interface. It is hard engineering at the infrastructure level: data, security, connectivity and control.
The new moat is complexity
This is where the opportunity lies for Israeli founders. After years when many startups crowded into segments like marketing tech and HR tech - exactly the categories most exposed to AI-driven commoditization - it is time to return to our home turf: deep tech and infrastructure. These are the areas where engineering complexity is itself a strategic defense, and where Israel's talent base - much of it shaped in elite technology units - has a genuine edge.
The openings are concrete: optimizing GPU and cloud costs as AI workloads explode, securing data and AI agents operating inside the enterprise, and building the data infrastructure that serves as the plumbing of the AI economy. Incumbent software companies will struggle to build these layers themselves because of technical debt and rigid architectures, which is exactly why they keep buying them. A startup building AI infrastructure today can start with an advantage over companies valued in the billions, and a clear path to an exit.
For founders deciding where to build, three questions are worth asking before writing the first line of code:
  • Could a well-funded competitor rebuild this in six months with today's AI tools? If the answer is yes, the product is not defensible. Proprietary data, deep integration and hard engineering compound over time; the interface does not.
  • Who would want to buy this, and why can't they build it themselves? Strategic buyers pay for what they cannot build in-house.
  • Will the platform you build on eventually offer what you sell? If the cloud providers or model companies can absorb your product as a feature, you are building on borrowed ground.
Wall Street will keep changing its mind about software, sometimes within weeks. The direction for the next generation of Israeli startups should not change with it. The new golden age of Israeli tech will not be built on another dashboard. It will be built on technology that is genuinely difficult to copy.
Doron Aaronsohn is the CMO at Sela, a cloud and AI service provider.