Opinion
Same funds, different risk: What Israeli and American cap tables reveal about AI shock
"Israel does NOT have an early-stage capital problem," writes Dor Lee-Lo, Managing Partner at IBI Tech Fund, "it has a supervision problem."
The most consequential property of Large Language Models, for the business of building software companies, turned out not to be intelligence. It was cost. Coding assistants, agent frameworks and generally available models collapsed the marginal cost of producing software. A product that required a team and eighteen months can now be shipped by a small group of people in weeks. Autonomous agents that read and write across a company's stack began absorbing workflows that horizontal software used to own, which weakened the SaaS pricing models and per-seat logic.
The financial consequence followed quickly: roughly $285 billion came off software valuations in the first quarter of 2026 alone. For venture capital funds, the problem is that the pricing methodology for many software companies is completely undermined. ARR, NRR, gross margin, seat expansion and almost every input in the early-stage model now carries a question mark. That leaves an investor with two ways to respond: move into categories where the old model never applied in the first place, or move toward companies whose outcomes can be estimated for reasons that have nothing to do with the model.
Capital injected into American startups took the first route, at scale. In the first half of 2026, global venture deployed roughly $510 billion into AI. Two companies, OpenAI and Anthropic, absorbed $217 billion of it, about 43% of every venture dollar deployed worldwide. Deep tech now accounts for 36% of global venture capital, and one in four unicorns minted this year came from robotics, defense or aerospace. The response of US capital in America was to push the frontier and fund the creation of categories and industries that did not previously exist.
To understand the situation in the local market, I classified 2026 rounds one by one. With the help of my friend and colleague Claude, I catalogued every Israeli seed and Series A round of $10 million or more listed in the full list of Israeli high-tech funding rounds in 2026 by CTech between January and July: 80 rounds, $2.44 billion, split evenly between seed and Series A, median round $25 million. For each I recorded the amount and stage, then coded two things. First, whether the company sits in a category with an established Israeli venture track record before 2020, or in a frontier category without one: quantum, AI silicon and hardware, robotics and physical AI, energy, neurotechnology. Second, whether CTech's reporting at least one founder with a prior exit or a senior operating role at a company that was acquired or is a major technology firm.
It turns out that the Israeli ecosystem took the second route, and interestingly, with the same investors. US funds concentrated capital behind founders who had built and sold companies before, working on problems the ecosystem already understands in depth, with AI applied to make those solutions significantly better rather than to open new category. The pattern is easiest to see in identity and access, a category Israel has been selling into for fifteen years. In July 2026, Shai Morag raised a $60 million seed for Oak, an AI-native platform meant to replace the patchwork of legacy identity tools, after three prior exits. Greylock, Accel and CRV led. NewCore emerged with $66 million raise to rebuild identity security for the AI era, founded by Dome9's co-founder, Nym's founder, and a former CIO of T-Mobile and Telstra, with the stated ambition of going head-to-head with Microsoft and Okta. Way Security raised a $20 million seed for enterprise identity, founded by two cybersecurity veterans and Venice raised $25 million Series A, to replace static enterprise permissions with dynamic access control. The same shape repeats across security more broadly: Medigate's founders raised $20 million in seed for Act Security and closed a $60 million Series A four months later. Nir Zuk (Palo Alto Networks) and Ehud Shamir (SentinelOne) launched Cylake with a $45 million seed, Frame raised $50 million, founded by one of Wiz's first employees and a former CTO of Team8 and Above Security raised $43 million, run by two serial entrepreneurs, betting AI agents can solve insider threat. None of these is a modest ambition, and none is a new category. Each takes a problem Israeli security companies have worked on for a decade or more and rebuilds it with the new technology, led by someone who has sold a company in that exact market before.
Within this sample, cybersecurity and defense in existing well-established markets account for 45% of deals. AI applications take another 14 rounds and 15.6% of capital. Between them, three familiar categories absorb more than seven of every ten dollars and only 8 deals (10%) are in frontier categories. Three quantum companies, two in energy, one robotics (Enigma), one neuro-AI foundation model, one industrial automation. That is the entire frontier of Israeli early-stage venture for seven months, in a year that American capital is rewriting its own category map. Many of the cap tables carry the same names of US funds, so the same investment committees, but different types of risk taking. That rules out the easy explanations: capital availability, funds’ culture or the shock itself.
A more complete explanation grounded in academic research is that venture risk has two components, the risk selected at entry and the risk managed afterward, and appetite for the first depends on capacity for the second. At 9,000 kilometers, the capacity for the second risk thins, and the investor substitutes stronger selection at entry for weaker governance later. A proven operator in a category the ecosystem has dominated for two decades is the most legible asset on the menu. This predicts both results: concentration in categories the market can already price, and a frontier that exists only where cheques are large enough to buy a committed lead and a long runway.
If this reading holds, the policy conclusion reverses the usual one. Israel does NOT have an early-stage capital problem. In seven months, $2.44 billion arrived in large seed and Series A rounds, most of it from abroad. Israel has a supervision problem. Adding foreign capital to a market where the binding constraint is monitoring capacity does not change what gets funded, it scales the existing allocation. What really moves the frontier share is domestic investors with judgment, conviction and board presence to lead those rounds, and foreign capital syndicating with them. The risk of not building that capacity is not that Israeli founders will stop building excellent cybersecurity companies. They will not. It is that ten years from now, that is still the best thing the ecosystem knows how to produce.
Dor Lee-Lo is the Managing Partner at IBI Tech Fund.
