The Tel Aviv AI Map: 24 Companies, a $32B Exit, and 2.9km End to End

The Tel Aviv AI Map: 24 Companies, a $32B Exit, and 2.9km End to End

Google closed its $32B acquisition of Wiz in March 2026, the largest deal in Google's history. The price is not the part that stayed with me. The address is. Twenty three more AI companies sit within three kilometres of it.

I have mapped a few AI clusters now, and this one behaves differently from the rest. Paris and New York are dense. Tel Aviv is dense and it exits, repeatedly, and the operators do not leave afterwards. That combination is the actual story.

Here is the map of Greater Tel Aviv, all 24 companies, and then the mechanism underneath it.

The 24 companies, by sector

Frontier and infrastructure. AI21 Labs in foundation models, with $336M raised from backers including Google and Nvidia. Decart in real-time generative world models at a reported $4B valuation. Run:ai in GPU orchestration, acquired by Nvidia for a reported $700M. Hailo in edge AI processors, now being acquired by Microchip Technology. Firebolt in AI data warehousing.

AI security. Wiz in cloud security, acquired by Google for $32B, closed March 2026. Cyera in AI data security at $12B. Torq, an agentic SOC platform that joined the unicorn club in January. Upwind in runtime cloud security at a reported $1.6B.

Go-to-market and revenue. Gong in revenue intelligence, with ARR past $500M growing more than 55% year over year. Similarweb in digital intelligence (NYSE: SMWB). Taboola in AI recommendations (Nasdaq: TBLA). Lusha in B2B contact data. Explorium in AI data enrichment.

Media and voice. D-ID in AI avatars and presenters. Bria in licensed visual generative AI. Deepdub in AI dubbing and voice localisation.

Dev and work AI. monday.com in AI work management (Nasdaq: MNDY). Qodo in AI code review. Tabnine in AI coding assistance.

Vertical AI. Aidoc in clinical imaging, deployed in close to 2,000 hospitals. Riskified in fraud prevention (NYSE: RSKD). Navina in clinical AI for primary care. Verbit in AI transcription and captioning.

Three things that stood out

It clusters harder than anywhere else I have mapped

18 of the 24 sit within one kilometre of each other, around Sarona. Wiz and Explorium share an address. Gong and Upwind share another. Decart and monday.com are fifty metres apart. End to end, the entire map is 2.9 kilometres.

The exits happen fast, and the people stay

Nvidia bought Run:ai for a reported $700M. Google closed the $32B Wiz deal in March 2026. Microchip signed a definitive agreement to acquire Hailo in July 2026. Three exits from one small map, and the pattern afterwards is consistent: the operators start the next company four blocks away rather than relocating to an acquirer's headquarters.

The go-to-market layer runs deep

Gong, Similarweb, Taboola, Lusha and Explorium are all on this map. If you run outbound, a meaningful slice of your stack was built within a few kilometres of the rest of it.

Why exit velocity compounds

Most clusters treat an acquisition as an ending. This one treats it as a funding event for the next cohort. Three mechanisms make that work.

Capital recycles locally and quickly. A large exit converts several hundred employees into people with liquidity, a validated resume and an existing network, all inside the same postcode. Some fraction of them start companies and another fraction angel invest in the ones that do. The money does not need to be raised from abroad before the next company can begin.

Acquirers plant research centres instead of relocating teams. Nvidia, Google and Microchip are buying capability that is geographically committed. The engineering stays, the acquirer's brand becomes a local employer, and the talent pool gets deeper after the exit rather than thinner. In a cluster where acquisitions mean the team moves to the buyer's headquarters, the map thins after every win. Here it thickens.

Second-time founders skip the slow parts. They already know how to sell to a US enterprise from a different timezone, how to structure the American entity, how to recruit design partners, and how long the security review will take. That knowledge is the difference between an eighteen month first sale and a six month one, and in a 2.9km map it transfers by conversation.

A large slice of the outbound stack has one origin cluster

This is the part that changes how I read the map. Look at the go-to-market layer as functions rather than logos.

Call recording and revenue intelligence: Gong, now past $500M ARR and growing more than 55% year over year. Traffic and competitor intelligence: Similarweb. Contact data: Lusha. Enrichment and signals: Explorium, which is the layer most teams reach for when they start building intent-based targeting. Distribution and recommendation: Taboola.

That is four or five layers of a standard modern stack, built inside a few square kilometres. It is worth knowing when a vendor tells you their product category is crowded. The category may be crowded, but the origin of the serious products in it often is not.

The headquarters trick

Look most of these companies up and you will see a New York or San Francisco address. Seventeen of the 24 register their global headquarters in the US, and Wiz was described as New York headquartered in most of the coverage of the Google deal.

That is a legal and commercial decision about where to sell and raise, not a statement about where the product is built. The engineering is on this map. The same gap shows up in reverse across the US headquarters map, and it is why any cluster analysis based purely on registered addresses will mislead you.

Caveats on the numbers, since several are reported rather than disclosed. The Run:ai price was never officially confirmed and $700M is the widely reported figure. The Hailo transaction was signed in July 2026 and had not closed at the time of writing. Decart and Upwind valuations are reported private rounds. The Wiz price and close date are the firmest numbers here, because they came from a public acquirer.

How Tel Aviv compares

New York built its cluster around proximity to buyers in finance, media and health. Paris built one frontier lab plus a long applied tail. Europe as a whole skewed toward regulated domains where compliance is the moat, and France in particular produced a disproportionate amount of go-to-market software. The most valuable US AI startups concentrate value in the model layer.

Tel Aviv is the security and infrastructure cluster, with a revenue software layer attached, and it recycles faster than any of them.

Frequently Asked Questions

Why did Google pay $32B for Wiz?

Cloud security is the category where Google Cloud was furthest behind AWS and Azure, and Wiz had built the fastest growing product in it with a multi-cloud posture. The deal was announced in 2025, cleared regulators through late 2025 and early 2026, and closed in March 2026 as the largest acquisition in Google's history.

Which Tel Aviv AI companies matter to a sales team?

Gong for revenue intelligence, Similarweb for traffic and competitor data, Lusha for contact data, and Explorium for enrichment and signals. Taboola matters if you buy distribution. Between them they cover several distinct layers of a normal outbound stack.

Are these companies Israeli or American?

Both, in a specific way. Seventeen of the 24 register their global headquarters in the United States for commercial and fundraising reasons, while the engineering organisations remain in Greater Tel Aviv. Registered address is a poor proxy for where a product is actually built.

What makes the Sarona concentration unusual?

18 of the 24 companies sit within one kilometre of each other, and the whole map spans 2.9 kilometres. Several share a single building address. That density means an exit does not disperse the talent, it redistributes it a few blocks away.

RevenueFlow builds AI-native pipeline systems and you pay per qualified meeting, not a retainer. No paying for activity. You only pay when we book you a qualified sales meeting. See if you qualify.