Bought for the Brains
Israel's AI startups increasingly get acquired before they scale — sold for their teams, not their revenue. The 2025 acqui-hire wave, the Habana cautionary tale, the Mellanox contrast, and what it costs a country to keep selling its best engineers early.
Israel's AI startups increasingly get acquired before they scale — sold for their teams, not their revenue. The acqui-hire economy, the Habana cautionary tale, and what it costs a country to keep selling its best engineers early.
There is a recurring pattern in Israeli technology, and the AI era has accelerated it: build a sharp team, prove a hard capability, get acquired before you ever scale into an independent company. In 2025 it became the dominant shape of the Israeli AI exit — and it raises a real question about what the country is building versus what it is selling.
The 2025 acqui-hire wave
Strip out the giant headline deals and the Israeli exit market in 2025 was a story of small, fast acquisitions. PwC Israel counted 22 transactions involving companies founded in just the previous three years — most valued under $50 million, and nearly half focused on AI. The average acquisition size, excluding the Wiz mega-deal, fell about 40% to roughly $160 million.
These are not distressed sales. As PwC framed it, they are strategic acqui-hires — large technology firms making aggressive, often defensive moves to secure small, efficient teams and critical AI capabilities before a competitor does. The acquirer isn't buying a business. It's buying a roomful of engineers who have already solved something hard, and the IP that came with them.
Recent examples span the spectrum: Apple's roughly $1.5 billion purchase of the secretive Israeli sensing startup behind silent-speech recognition; the Dutch AI firm Nebius acquiring Tavily for a few hundred million; a steady stream of cyber-AI teams — Prompt Security, Veriti, Dig Security, Avalor — absorbed into global buyers over the past two years.
The Habana cautionary tale
The risk in the model is best told through one deal that went wrong. In 2020 Intel acquired the Israeli AI-chip startup Habana Labs for roughly $2 billion. Habana had built Gaudi, an AI training processor on a different architecture from NVIDIA's GPUs — Amazon later judged an iteration meaningfully more efficient than the NVIDIA equivalent. Amazon itself was rumored to be circling. Intel outbid and won.
Then it mismanaged the asset. Intel pursued several competing internal AI strategies without fully committing to any, and never let Habana become the NVIDIA challenger it could have been. By 2024, nearly all of Habana's original founders, managers, and engineers had left — most the moment their four-year retention packages vested. Today almost none of the original hardware team remains, and Habana has been folded into Intel's broader operations.
The contrast that defines the era sits right next door: NVIDIA's acquisition of Mellanox, closed around the same time, was integrated successfully and now generates more than $13 billion in annual revenue while employing thousands in Israel. Same country, same window, opposite outcomes. The acquirer's competence, not the Israeli team's quality, decided which Israeli technology shaped the AI era and which one disappeared inside a struggling giant.
The bigger picture
Zoom out and the scale is striking. More than 70 Israeli semiconductor startups have been acquired for a combined $44 billion — Mellanox to NVIDIA, Mobileye and Habana to Intel, Annapurna to Amazon among them. The country built the engineering that sits inside the AI hardware stack, then sold most of it to American owners.
There is a genuinely positive reading. The acqui-hire economy is liquidity. It returns capital to founders and early employees, who recycle it into the next generation of startups and into the family offices and funds now backing Israeli AI. It validates the talent globally and pulls in foreign R&D centers. The flywheel is real and it has made a lot of Israelis wealthy.
And there is the cautionary reading. A country that consistently sells its best AI teams at the sub-$200-million, pre-scale stage is, structurally, a research-and-development supplier to the global giants rather than a builder of independent global champions. The enterprise value of the scaled company accrues elsewhere. The Habana story is the warning inside the warning: even the value you do sell can be destroyed by an acquirer who doesn't know what to do with it.
Why it matters
The acqui-hire wave is the clearest expression of Israel's central AI-economy tension: extraordinary at producing talent and capability, weaker at retaining the scaled enterprise value that talent creates. The 2025 data — dozens of young AI teams sold early — shows the pattern intensifying, not fading, as the AI gold rush makes engineering teams the scarcest resource in the world.
The question for the next decade is whether more Israeli AI companies choose to stay independent and scale — the AI21 path — or whether the country remains content to be the place the giants come to shop for brains. Both are lucrative. Only one builds lasting Israeli enterprise value.
Part of The Israeli AI Economy, Olam's complete map of Israel and AI. Related: Fewer Bets, Bigger Checks · Israel's Model Layer · The Israelis Inside the Machines.
