Mobileye

1999: betting on one camera instead of a sensor array

Mobileye was founded in May 1999 in Jerusalem by Amnon Shashua, a professor of computer vision at the Hebrew University, and Ziv Aviram. Its central bet was both technical and commercial: that a single, relatively cheap camera, plus smart enough software, could detect in real time a car braking in front of you, a pedestrian stepping into the road or a drift out of the lane, without an expensive sensor array. In those years the accepted assumption in the auto industry was the opposite: active safety would require expensive distance sensors, and a camera was considered too weak a source of information. The ability to turn a flat image into an estimate of distance and speed within fractions of a second is exactly what Mobileye built, and it packaged that ability in a dedicated chip called EyeQ that carmakers could install on their existing production lines. This combination (one component, at a reasonable price, that meets carmakers' safety requirements) is what turned the technology from an academic research project into a mass-market product. The first systems were built into vehicles from many manufacturers around the world, including BMW, Ford, Nissan, Volkswagen and Porsche, and spread further from there. EyeQ chips have since gone through several generations: the first was unveiled in 2004 and entered series production in 2008, and in January 2022 the company unveiled EyeQ Ultra, a chip for autonomous driving, expecting full production in 2025.

From 2.7 million chips a year to 35 million

In December 2021 Mobileye announced that it had passed the mark of 100 million EyeQ chips shipped since 2004, and according to the company this is also the number of cars that left the factory with its technology installed. In the same announcement it said that more than thirty leading carmakers use it, and that it is found in more than three hundred models on the market. The shipping rate alone shows the growth: 2.7 million units in 2014, 19.3 million in 2020, and about 35.6 million units in 2025, according to the company's reports. Revenue in 2025 came to about $1.9 billion, again according to the company's reports, compared with about $1.65 billion in 2024. Numbers like these make Mobileye one of the Israeli companies whose product is physically installed in the largest number of devices in the world, even though almost no one driving such a car knows it is there, and no brand appears on the screen. This also says something typical about the place of Israeli industry in the world: not a consumer brand on the box, but a component inside that the whole industry knows. In China, too, there is a company developing AI chips for cars: Horizon Robotics, whose founder Yu Kai previously headed Baidu's autonomous driving division, makes the Journey series of chips for driver-assistance and self-driving systems, and in October 2024 it went public on the Hong Kong stock exchange.

An IPO, an acquisition and a second IPO, with control staying put

In July 2014 Mobileye went public on the New York Stock Exchange and raised about $890 million, the largest IPO by an Israeli company in the United States up to that time. In March 2017 Intel announced it would acquire the company for about $15.3 billion, and the deal was completed in August of that year. It was then the largest acquisition of an Israeli tech company. Intel did not absorb the company into itself but ran it as a separate division, and in October 2022 it brought it back to the stock market, this time on Nasdaq, in an IPO in which 41 million shares were sold and about $861 million was raised, at a valuation of about $17 billion. A point that is easy to miss: the second IPO did not make Mobileye an independent company. Intel kept all the Class B shares, and with them about 99.4 percent of the voting rights. In other words, the stock is publicly traded, but the one who makes the decisions is still the one who made them before. The logic behind such a structure is simple: a partial IPO lets the controlling owner discover the market value of the business, and raise money for it, without giving up control. For a public investor, this means buying exposure to the company's results, but almost no ability to influence its decisions.

REM: a road map built from the data of millions of cars

One of the company's lesser-known assets is REM, short for Road Experience Management. The idea is simple to explain and hard to carry out: every car driving with an EyeQ chip can send back a tiny summary of what the camera saw (lane markings, traffic signs, road edges), and from these summaries, collected from millions of drives, a detailed map of the road itself is built. The data is highly compressed: according to the company, only about ten kilobytes per kilometer driven, small enough to send over the cellular network at no significant cost. By January 2021, more than 7.5 billion kilometers of roads had been mapped this way. The value of such a map is not navigation but precision: an autonomous driving system needs to know exactly where the lane runs and exactly where the sidewalk begins, at a level of detail that ordinary navigation maps do not provide. It is also a clean example of the network effect in this industry: the more cars drive, the better the map gets, and whoever already has more cars on the road gains an advantage that is hard to catch up with. Note that the map is not built from dedicated cars sent out to scan roads, but from ordinary cars that are driving anyway. That is the difference between a mapping project, which costs money for as long as it runs, and an asset that grows by itself. A similar idea is behind Nexar, whose dashcams are installed by drivers themselves: according to the company, the road data they collect, after faces and license plates are blurred in the camera itself, helps improve maps and detect road hazards. Not every company in the field shares this approach: Israel's Imagry develops driving software that does not rely on pre-made maps and makes decisions in real time based on road conditions, and in November 2022 it received approval to operate autonomous buses on a public road in Nahariya.

Two parallel approaches: camera only, or true redundancy

Mobileye develops two different approaches in parallel, and the difference between them lies at the heart of the debate over autonomous driving. In the camera-only approach, all the information comes from vision sensors. This is the cheap approach, suited to driver-assistance systems in vehicles made by the millions. In the approach the company calls True Redundancy, two separate systems run independently of each other, one based on surround cameras and the other based on radar and laser sensors, and each of them is supposed to be able to drive on its own. The advantage is not just extra accuracy but a different kind of safety: two systems built on different physical principles are less likely to fail at the same moment and in the same way, so the failure of one does not drag down the other. The company's advanced products are built on this basis, including Mobileye SuperVision for advanced driving systems and Mobileye Drive for autonomous driving in ride services. The choice between the two approaches is not only technical but economic. A redundant system costs much more per vehicle, so it makes sense in an autonomous taxi that works many hours a day and earns money every one of those hours, and less in a private car that drives one hour a day and sits parked the rest of the time.

2025–2026: layoffs, heavy losses and a move into robotics

The business picture is far from simple. According to the 2024 reports, revenue was about $1.65 billion, but in that same year an operating loss of more than $3 billion was recorded, a huge gap between the scale of activity and the bottom line. In December 2025 the company laid off about 200 employees, about five percent of its workforce, amid falling demand. In January 2026 it announced its intention to acquire the Israeli humanoid robotics company Mentee Robotics in a deal valued at about $900 million, an attempt to expand from autonomous vehicles into broader physical robotics. On the other hand, the company points to a large future pipeline: according to its reports to the capital market at the end of 2025, the revenue it expects over the next eight years in the automotive field adds up to about $24.5 billion. This number should be read with caution. It is a measure that the company itself calculates and publishes, reflecting carmakers' purchasing intentions, not revenue already on the books. The robotics move should also be read as a stage and not a done deal: there is a distance between announcing an intention to acquire and actually completing the acquisition, and in this industry it is sometimes measured in a year or more. On 23 July 2026 the company announced that its founder, Amnon Shashua, would step down as CEO once a successor is appointed; the board offered him the role of chairman at that point. As of early November 2026, no successor had been named.