Lazy Digest #5: $6B for 100 Engineers, and LinkedIn Turns Out the Lights
The biggest Israeli tech number this week wasn’t a funding round. It was a rumor, and it changed both its price and its buyer inside three days.
Monday: Decart is fielding offers around $5 billion, possibly from Nvidia. Thursday: Anthropic, $6 billion. Same company, same week, a billion dollars of drift. Nobody has confirmed anything.
Underneath the noise, the actual disclosed week was the thinnest since this digest started: three funding rounds, $116 million total. And in the same five days, LinkedIn closed its Israeli R&D center entirely.
Lazy Digest #5 — Israeli tech, one week, researched with AI and filtered by a human who reads the primary sources. Estimates are labeled as estimates. Let’s go.
The Rumor That Gained a Billion Dollars
Monday, August 10. Globes’ Assaf Gilead reported that Decart had entered sale negotiations after receiving offers around $5 billion, with Nvidia believed to be among the bidders — three months after Nvidia invested at a $4 billion valuation.
Thursday, August 13. Bloomberg and Reuters reported that Anthropic is in advanced talks at about $6 billion. It would be Anthropic’s largest acquisition ever and its first in Israel. Elon Musk publicly dismissed a separate report that SpaceX was circling.
Decart was founded in 2023 by Dr. Dean Leitersdorf (CEO) and Moshe Shalev (CPO), roughly 100 people split between Tel Aviv and San Francisco, about $450 million raised across Radical Ventures, Atreides, Valor, Zeev Ventures, Sequoia, Benchmark, the venture arms of Toyota, Adobe and eBay, and angels including Andrej Karpathy and Michael Eisner. It builds real-time generative video and world models — Oasis, Lucy — and an optimization stack that extracts more throughput from the same AI silicon.
Now the paragraph that most coverage skipped, and it’s the one that matters. Globes’ own sources, people who know the company, cast doubt on the whole premise: low sales, a product focus that has already shifted once, and a suggestion that Decart is interesting to buyers mainly for the density of engineers and AI researchers it has assembled rather than for anything it currently sells. The comparison drawn was Nvidia’s earlier Israeli purchases — Run:ai and Deci AI.
Read that honestly and the story stops being about generative video. Somebody is preparing to pay five to six billion dollars for roughly a hundred people and a set of inference tricks. That is a talent transaction wearing a product transaction’s clothes, and at that price it’s the largest one this ecosystem has ever produced.
Two things follow. The good one: if it closes, a frontier lab plants its first serious R&D operation in Israel, and the thing being validated is Israeli strength in compute efficiency — not cyber, for once. The uncomfortable one: two buyers and two prices circulated in 72 hours for a company whose revenue nobody will name. In issue #1 I wrote that 2026’s buyers were shopping for capabilities rather than businesses. This is that thesis with three more zeros on it.
Hold it loosely. Nobody commented, and deals at this size die quietly all the time.
Housekeeping, since it’s relevant: this digest’s research is done with AI, and this week the AI in question is made by one of the reported bidders. It doesn’t change a word above, but you should know it.
Three Rounds. That’s the Week.
CTech’s running 2026 funding tracker logs exactly three Israeli rounds between the 9th and the 15th. For comparison: issue #3 counted $278 million in agent-security alone, issue #4 about $315 million across everything. This week: $116 million, three companies.
Mid-August is thin, and announcement calendars empty out. But three is three.
Corma — $60M seed, out of stealth Monday. A foundation model built specifically for defensive cybersecurity, on the argument that general-purpose models keep getting better at attacking companies and remain badly equipped to defend them. Founded 2025 by CEO Alon Pluda, Tel Aviv and San Francisco, mixing DeepMind and Google research people with 8200 and industry cyber operators. Fortune reported Sequoia leading with Khosla and Coatue; CTech’s write-up named no lead, so treat the lead as Fortune’s reporting.
The line worth keeping is Pluda’s own framing to Calcalist: Corma doesn’t replace anyone and isn’t a product — it provides services. A $60 million seed to train a model from scratch and sell it as a service is a long way from the seat-license business Israeli software was built on.
Attestable — $20M seed, out of stealth Tuesday. Led by TLV Partners and Altimeter’s Jamin Ball, with Netz, Cerca, Zero Knowledge Ventures, Halcyon Futures, and an angel list that by now functions as an Israeli cyber credit rating: Wiz’s Assaf Rappaport, Armis’s Yevgeny Dibrov and Nadir Izrael, Meta board member Charlie Songhurst.
Three mathematicians founded it in 2025 — CEO Yogev Bar-On, ex-RAND, who advised the US government on AI and national security; CTO Shahar Papini, ex-StarkWare and then Ilya Sutskever’s Safe Superintelligence; VP R&D Shahar Samocha. About 15 people. They use zero-knowledge proofs to let an organization verify which computations an AI system actually ran, and on what data, without exposing model weights or the data itself.
Bar-On’s stated customer set is the frontier labs — securing them against attacks by foreign actors. Fifteen people selling cryptographic proof to OpenAI and Anthropic is an absurd sentence, and it’s also roughly what Irregular does further down this page. Israel has quietly acquired a second-order industry: companies whose customers are the four or five organizations building the models.
Remepy — $36M Series A, Wednesday. Co-led by O.G. Venture Partners and M Ventures, the corporate arm of Merck KGaA, with NFX, Vine, Qumra, TechAviv, 97212, Key1 and others. Total now $62 million. Founded 2022 in Ramat Gan, ~60 people, co-led by Michal Tsur — who co-founded Kaltura — and Or Shoval.
The product is the interesting part: “hybrid drugs,” a prescription medicine paired with personalized AI-driven therapeutic software, dispensed on a normal prescription. The money funds a global Phase III for Hybridopa in Parkinson’s starting Q4, after Phase IIa showed improvement in both motor and non-motor symptoms, plus a Merck collaboration extending into rare tumors.
Strategic pharma money on the cap table is the same signal I flagged with Sanofi and QuantHealth in issue #4: a corporate investor in this sector is a customer telling you in advance that it intends to buy.
Three rounds, and all three sell something other than software seats. That’s not a coincidence, it’s the year.
The Fund That Exited Before It Closed
Team8 raised $365 million — $265M for Capital Fund III and over $100M for follow-ons, taking the firm to roughly $2 billion across eight funds since 2014. Managing partners Sarit Firon and Liran Grinberg, with Ori Barzilay and Hadar Siterman Norris. Checks of $5–15M into seed and Series A, AI-native, across cyber, infrastructure, fintech and digital health.
The detail I’d underline: Fund III already recorded an exit before its final close. Team8 co-led Koi Security’s ~$48 million round; Palo Alto Networks bought Koi for about $400 million in February. A fund that returns capital before it finishes fundraising is not a normal artifact of a slow market.
Firon’s framing in the announcement is the most quotable line of the week and, I think, correct: building breakthrough technology has never been easier, and building a company that endures has never been harder. Every AI-native seed in this ecosystem is a wager on the second half of that sentence.
Ofir-SouthUp Ventures — a new $20–30M fund for the Western Negev. Named for Ofir Libstein z"l, head of the Sha’ar HaNegev Regional Council, killed on October 7th. Chaired by Dr. Ami Appelbaum, former chairman of the Israel Innovation Authority, with Itai Horstock. The LP list reads like nothing else in Israeli venture: kibbutzim Kfar Aza, Yakum and Ma’agan Michael alongside Dov Moran, Shlomo Kramer, the Bronickis, the Zim fund and the Jewish Community Foundation of San Diego.
First investment, before final close: ZutaCore, the liquid-cooling company for AI chips, ~140 people in the Sapirim industrial zone. An OfirTech park of about 3,000 square meters is planned for Q1 2027.
It would be easy to file this as a memorial project. It isn’t structured like one — it’s an early-stage fund with a geographic mandate, and its first cheque went into a company selling into the global data-center buildout. The Negev’s problem was never a shortage of goodwill. It was a shortage of cap tables.
Subtraction: LinkedIn Turns Off the Lights
LinkedIn is closing its Israeli R&D operation entirely and laying off nearly all of its roughly 50 employees, with a handful staying temporarily to wind things down. Calcalist’s Meir Orbach broke it Tuesday.
The center opened in February 2022, immediately after LinkedIn’s first-ever Israeli acquisition: Oribi, Iris Shoor’s no-code analytics company, for an estimated $80–90 million. The purchase and the center were both championed by LinkedIn CPO Tomer Cohen. The Tel Aviv team went on to build the measurement and attribution tooling inside LinkedIn Marketing Solutions. Although LinkedIn is Microsoft-owned, the Israeli operation always ran as a separate entity from Microsoft’s own local R&D. The official statement is the standard one about focusing teams on highest-impact priorities.
Four years from acquisition to closure. And that timeline is the thing to sit with, because “get acquired by a multinational, become their Israeli development center, employ 50–200 people for a decade” is one of the load-bearing stories this ecosystem tells about why exits are good for the country. It is a real story — it built Israeli Intel, Israeli Microsoft, Israeli Google. But it has an expiry date that nobody prices in at signing.
SciPlay cut 20 jobs at its roughly 300-person Israeli hub on Monday. That operation — built partly on the 2014 Dragonplay acquisition — is about a third of SciPlay’s global workforce and develops some of its biggest titles. Small numbers, same category as Papaya, Moon Active and Playtika before it.
In issue #4 I noted the first week since this digest began with no standalone Israeli layoff, and warned against reading it as a turn. One week later: two.
The Data Under the Layoffs
The Israel Innovation Authority and Zviran published a survey on Tuesday covering 210 companies and about 130,000 employees — over 80% of Israelis working at tech firms with 50+ staff. It is the best picture of this labor market anyone has published this year.
Headline: employment is roughly stable. Companies hired about 8% of their workforce in H1 and laid off about 2.8%, with 4.3% leaving voluntarily.
Then the split, which is the actual news. Software layoffs averaged 6.6%. Hardware: 1.1%. Medical and pharma sat at 2.7%. AI embedded in products went from 21% to 30% of companies in six months. And more than a third of firms expect to hire less in H2.
Two more numbers worth carrying. Roughly 17.6% of companies doing broad layoffs, and 28% of those slowing hiring, named the exchange rate as a cause. And Globes reported the same week that importers are buying dollars aggressively while institutions sell, with July inflation landing at 1.5% — which reduces the pressure for rate cuts, which keeps the shekel strong.
That’s the loop I described in issue #3 when the government wrote its NIS 1 billion fast-track cheque. The cheque is still open through November 19. The force it was written against has not weakened.
And the supply side is now leaking. An Israel Tax Authority study covered on the 13th found emigration among the top 10% of earners rose from 0.3% to over 0.5% by 2024 — high-tech departures up around 150%. Income tax paid by emigrants in their final year before leaving went from roughly NIS 500 million a year pre-2019 to about NIS 1.2 billion in 2023–24. Sustained, the Authority’s own projection is a loss approaching NIS 3.5 billion a year within five years, in a country where the top fifth of earners provide up to 92% of income tax.
Issue #2 reported 16,300 Israeli tech job seekers, a peacetime record. This is the other half of that story: some of them stopped seeking here.
The Earnings Issue #4 Promised
I ended last week saying monday.com’s Q2 would be the first quantified read on whether the summer’s cuts turned into margin. It reported Monday, and the answer is yes — with an asterisk the whole sector should read.
monday.com: revenue $364.6 million, up 22%. Non-GAAP operating margin 17%, against 15% a year earlier, and that’s after roughly 210 basis points of FX drag — the shekel again, showing up in a US-listed P&L. Beat on revenue and earnings. AI product ARR doubled from Q1 and made up 17% of net new ARR. Net dollar retention 109%, and 115% among customers above $50,000.
The cuts became margin. Exactly as designed.
The stock fell about 7.8% to around $86, down roughly 63% over a year. Because the Q3 guide implies 16–17% growth against the 22% just delivered, and margin guidance of about 16% — below the quarter just reported.
So: 620 people, a restructure, a genuine margin expansion, a double beat, and the market marked it down anyway, on the shape of the next four quarters. In issue #4 I wrote that a beat is no longer a story and the growth rate is the story. This is that, in its purest available form. If you’re on a board weighing the monday.com playbook, the honest read is that it works operationally and buys you nothing with investors unless the growth line holds.
Similarweb on Wednesday made the opposite trade and I’d argue got the better week. Revenue $77.2 million, up only 9% — but the first positive GAAP operating profit in company history, $6.5M non-GAAP operating profit, $8.7M normalized free cash flow for an eleventh consecutive quarter, ARR past $300 million, RPO up 26%. Three seven-figure multi-year contracts signed in the quarter worth over $60 million in total contract value, including an expansion that created the company’s third eight-figure account. Net retention back to 100%.
Nine percent growth and a first profit, versus 22% growth and a guide-down. Two Israeli companies, three days apart, at opposite ends of the same trade-off.
Defense went the third way: no trade-off at all. Elbit Systems beat with an order backlog at a record $32 billion, 73% of it from outside Israel — up from $30.2B a quarter earlier. That international share is the number to watch, because it means the defense-tech cycle Israeli startups are riding is an export cycle, not a wartime-procurement cycle, and export cycles outlive wars.
Though not without discipline. Next Vision’s shares fell sharply on ostensibly excellent results, because gross margins narrowed. First real sign that public investors are starting to price defense on margin quality rather than backlog growth — worth noting for the several Israeli defense startups currently drafting listing documents.
Defense Kept Buying
Ondas won the Israeli Defense Ministry’s “Digital Bat” tender on Tuesday, to develop a next generation of low-cost tactical attack drones for IDF infantry battalions — platform, autonomy, mission integration, production readiness and C2 compatibility. Neither side disclosed a figure; Israeli reporting called it multi-million. Ondas is Nasdaq-listed and American, but its Israeli drone business was assembled from Airobotics and Iron Drone, and CEO Eric Brock framed the award as validating the company as a prime contractor.
The context nobody put in the headline: Ondas lost a piece of this same program in July, when Calcalist reported that Israeli startups Kela and eyeAtop won the separate tender for Digital Bat’s autonomous command-and-control layer, beating Ondas for it. So the national attack-drone program is now split across an American prime and two young Israeli software companies. That’s a procurement structure that would have been unthinkable five years ago, and it’s the single most concrete evidence I’ve seen that the Defense Ministry means what it says about buying from startups.
Also in the window: Elbit will open a joint UAV factory in Serbia in September under a five-year, $1.63 billion contract — Israeli platforms plus local co-production, which is becoming the standard shape of European rearmament deals. And the Bank of Israel publicly warned on the proposed NIS 400 billion of additional defense spending over thirteen years. Defense tech’s domestic demand is ultimately a fiscal question, and the central bank has now said so out loud.
Everything Else Worth Your Time
Irregular got named by three frontier labs in two weeks. CNBC linked the Tel Aviv company to disclosed incidents at OpenAI, Anthropic and Meta in which AI models reached the public internet during cybersecurity evaluations meant to run sealed off from it. Irregular’s account: all three traced to the same misconfiguration in the test environment, first spotted by Anthropic, not a sandbox escape, no open issues, white paper coming with recommendations.
Founded 2023 as Pattern Labs by CEO Dan Lahav (ex-IBM AI research) and CTO Omer Nevo (ex-Google Research), who met at debating championships. About 35 people. Raised $80 million last September led by Sequoia and Redpoint at a reported ~$450M, with Omri Casspi’s Swish Fund, Assaf Rappaport and EON’s Ofir Ehrlich.
By CNBC’s own count, Irregular is one of a handful of organizations on earth capable of running advanced cyber evaluations on frontier models, alongside METR and Apollo Research. Getting named in three incident disclosures is not obviously good PR. It is extraordinary commercial positioning: it establishes, on the record, that three of the largest AI companies in the world route this work through a 35-person Israeli lab.
OurCrowd named Cali Chill permanent CEO and, more interestingly, announced a strategy change: away from broad early-stage diversification, toward later-stage, higher-conviction positions and faster liquidity including secondaries, citing a slow IPO market. Chill had been acting CEO since founder Jon Medved stepped back last year following an ALS diagnosis. The firm runs over $2.6 billion across 500+ companies and 73 exits.
Last week this same firm appeared here for realizing an estimated $200 million from BioCatch. Two weeks before that, for writing off Hailo. Now it’s restructuring around concentration and liquidity. Portfolios don’t have moods, but managers do learn.
Fortissimo sold CTS’s pharma business to Dexcel for an estimated NIS 250 million (~$83M), against the NIS 187 million it paid for the whole company four years ago. Founding-family shareholder Sigal First sold her 20% alongside. The Kiryat Malachi real estate and the Flourish supplements business stay behind, and Dexcel is not committed to keeping the plant open indefinitely. Not a tech story — included because it was the only closed Israeli M&A in the window, in a week when the rumored deal was worth 70x more than the real one.
Wonderful is reportedly raising $500 million at a $5 billion valuation, per Globes, with Insight Partners expected among the leads again. The enterprise AI-agents company raised $165M at $2B three months ago. Headcount reportedly went from ~90 a year ago to ~630; revenue from about $1M in August 2025 to $7M at year-end. Unconfirmed, and the valuation would double in a quarter.
TheMarker reported Google is building an Edge-AI chip team in Israel — for robots and autonomous devices, staffed partly by engineers from Hailo, whose acquisition by Microchip I covered in issue #3. Single-source exclusive, no Google confirmation, corroborated mainly by job listings, so hold it lightly. But put it next to the LinkedIn closure in the same week and the direction is hard to miss: multinationals are subtracting Israeli software and adding Israeli silicon.
The AI Directorate published Israel’s national AI action plan in English — 28 pages, sovereign compute, a National AI Institute, quantum. The follow-through on Project Nexus from issue #4. Same instinct I wrote about then, now with a document attached.
What I’d Do With This Week
If you’re a founder with a strong team and thin revenue: the Decart reports are the most encouraging thing you’ll read this year, and you should still be careful with them. Somebody is talking about $5–6 billion for roughly 100 people. But the reason that number exists is a global scramble for inference and optimization talent by four or five buyers, and it does not generalize to your category. Team density is being priced like an asset class in exactly one part of the stack.
If you sell software seats: monday.com just ran the full experiment for you. Cut 20%, expand margin two points against FX headwinds, beat on both lines, and lose 8% of your market cap on the guide. The restructure works. It does not buy you a re-rating. Decide whether you’re doing it for the P&L or for the share price, because it only delivers one of those.
If you’re weighing a multinational acquisition offer: the LinkedIn–Oribi arc is four years from $80–90 million to zero employees. That path is still worth taking and the money is still real — but negotiate retention, mandate and site commitments as though the center has a shelf life, because the evidence this week says it does.
If you’re hiring or job-hunting: the 6.6%-versus-1.1% split is the most actionable number in this issue. Software is being made leaner, hardware and defense are not. Elbit’s backlog is $32 billion. If you have any hardware, chips, sensing or systems background in your history, this is the year it’s worth more than your SaaS résumé.
If you’re building for AI labs: Attestable sells cryptographic proof to model developers. Irregular sells them adversarial evaluation. Both are Israeli, both are tiny, both have customer lists of maybe five names, and both got market validation in the same week. It’s a real category now, and it’s early enough that nobody has a defensible position in it.
If you’re watching the ecosystem number: three rounds, $116 million. Against a rumored $6 billion exit. That ratio — one week, admittedly thin, mid-August — is the concentration story I keep flagging, in its most extreme form yet. Breadth is still the thing to watch, and this week breadth lost badly.
See you next week. Same lazy hands, same working head.
With 💜, Alex Sherman
Research assisted by AI, every claim read and filtered by a human. Primary sources: CTech / Calcalist — including its running funding, M&A and layoffs trackers — Globes, The Times of Israel, the Jerusalem Post, TheMarker, Bloomberg, Fortune, and company filings from monday.com, Similarweb and Team8. Not confirmed, reported only: the Decart sale talks (both the $5B/Nvidia and $6B/Anthropic versions — neither Decart, Nvidia nor Anthropic commented), Wonderful’s $500M raise, and Google’s Israeli Edge-AI chip team (TheMarker exclusive, single-source). Corma’s lead investor is Fortune’s reporting, not CTech’s. The CTS–Dexcel price and the Ofir-SouthUp fund size are Calcalist estimates. Ondas’ contract value was not disclosed by either party. Revenue and growth figures from Wonderful and Irregular are press-reported and unaudited. Disclosure: this digest is researched using AI tools including Anthropic’s, and Anthropic is a reported party to the week’s largest story. Previous issues: #1, #2, #3, #4.