Intel’s Mobileye plans a third route to robotaxi, unlike Tesla, Waymo

Mobileye is an Israeli company, an intel unit, which has a predominant position in the Driver’s Assistance Market (ADAS) among heritage automobiles. They made the original Tesla Autopilot, but broke away from Tesla after a fatal collision. They have had their ambitions in “Pilot” supervised systems for many years, and also plan to make current self-direction systems up to Robotaxis.

Mobileye business is selling chips and software for vehicles, and while she briefly flirted with a plan to run her own robotaxi business, he currently hopes to do so, under the name “Mobileye Drive” in partnership with automotive VW Commercial vehicles, “Verne” of Rimac and some Chinese sellers. Under the name “Mobileye Chauffeur” they have some partners hoping to establish systems of part -time autonomy, where a car can drive themselves without human attention in free but a human driver is on alert when entering areas that the system does not taken.

When it comes to ADA more basic, Mobileye systems are in a large variety of vehicles. System is the most popular system in that area. Their “Mobileye” surveillance system is in a variety of cars and races with Autopilot, Ford BlueCruise and Byd NOA, to name some. Despite this, Mobileye’s financial results this week disappointed, resulting in a 20%market haircut. (Part of Mobileye is public. Most belong to Intel.)

At CES 2025, the Director General of Mobileye Amnon Shashua presented a contrast between their plan to achieve complete self-direction (sometimes mistakenly called “Level 5”) and plans, as they saw them, Waymo and Tesla. Waymo is the undisputed leader, being the first company to make a self-driving car, but Tesla probably gets more press, though it currently far away.

Shashua made progress in two axes, borrowing names from data science and computer perception. He called security performance as “precision” and a number of factors that control the usefulness of the system, such as geographical width (service area) and affordability as “memory”.

While security performance ranges from 0 to 100%, only values ​​very close to 100%-where you are ready to “bet your life” in the system as it runs better than most people-are true self-direction and can lead to a motor revolution. Lower levels are still useful (and in fact, are almost all Mobileye business), but they need a completely attentive human driver whose task they can make a little easier.

The next dimension is important, however, as you want to be able to use it in many places and want people to be able to access it and handle it. You really want both of the true revolution. If you want to sell a customer car, it would have better car almost anywhere. On the other hand, a taxi service that only serves a big city like New York or San Francisco can be a lucrative business. Indeed, a system that simply drives highways can be a valuable convenience of luxury cars, and Mobileye is building it. (Mercedes already sells such a system, but is currently very limited.)

Shashua describes Waymo’s path as working first to meet the full security requirement, and then expanding “memory.” Waymo has work vehicles in 3 cities and has announced the expansion in several others for this year and the other. Some describe the growth of Waymo as very slow, limited by the way they have chosen.

Tesla started with their Autopilot Driver’s help product, and is improving it over time, hoping to make it reach the level of safety needed. If they do this, then they can expand quickly where it is available almost anywhere. Their past access failed, but they declare that they have completely re-elected their system and are now doing better, though it is still extremely weak in security performance.

Some are critical of this strategy. Sterling Anderson, co -founder of Aurora and a leader spent on the Tesla Autopilot team, calls it “trying to build a ladder on the moon.” Many self-direction teams rejected the idea of ​​starting with Adas and improving it; Instead they worked to build self-direction from scratch. Tesla and Mobileye oppose it is a moon scale.

Mobileye progresses a third-way, “zig-zag”. They started with Adas and built their supervision product along a path similar to that of Tesla. Now, however, for their driver’s product, which will allow driving “eyes” (but not “driver” uncontrolled operation) they plan to limit the circumstances they can run but break over that critical level of security (“precision”). They hope from there, with the backdrop they had to oversee most roads, they can more easily increase the system in Mobileye drive, which wants to do the full task of driving, with anyone in the vehicle, in most roads.

Mobileye has taken a different path from Tesla. They make the processing chips he and join them with cameras (and sometimes radar) with their automobile customers. Unlike Tesla, they believe that radars and lidari can improve security performance and allow independent systems to act as reinforcements and obstacles in vision systems that are not perfect. (Tesla believes in full power in Silicon.) Mobileye was building his own loyalty, but has dropped that project and will work with innoviz. However, they are at full speed ahead, on an image radar-an added resolution radar that can do many of the things a lidar can do, and some cannot, how to see all weather. Waymo design uses more cameras, more image radar and more and better than anyone. They believe it is stupid to fight to save money on the first day, as computer/electronics products always get free on the scale. Their purpose, which they achieved, was to first take the safety and drive of the car.

Mobileye also has a different approach to the mapping I have discussed in previous articles. Their chips are in many tens of millions of cars, so they constantly collect data on all roads every day, though they receive much less car data.

Which road?

If the movement from Adas to the self-direction is really a scale on the moon, then both Tesla and Mobileye are being distracted, though more tesla. At the same time, thinking about the vanity of this has moderated both with the advent of his new powerful techniques, and the long period that has received teams aimed only at self-direction to get success. Some of them (Waymo, navigation type, Gatic, Nuro, May, Baidu, Pony, Weride and Autox) have received some level of success-defined as service on public roads with passengers and no security drivers) although Waymo and Baidu seems to be at a much higher phase of it.

At the same time, almost everyone has thrown most of their early models and systems and have begun again, almost from scratch, and in some cases many times. Thinking about the role of different sensors and what to do with their data has changed. The role of the maps has changed many-teams trying to avoid maps in many different areas from road maps at lane level. Others began with dedicated cartography machines making very detailed maps, but they all also run “outside the map” as needed in any construction area. More map data is coming from low -cost crowd sources, or for free because any car that hopes to drive without a map is a car that can make a map while driving, and the fleet can remember it for next time. Waymo, with its origin in Google, the #1 map enterprise uses that expertise, but is not finding the mapping to be an obstacle to expansion in other areas-there are many other obstacles to expand beyond mapping for all.

staleSo you have built a robotax, now where is your infrastructure?from Brad templeton

It is not clear if Mobileye Zig-zag is the best plan. The difficult truth is the results that matter today, not just plans. Mobileye should offer cars on the highway that works and work more than automobiles and taxis to show that they have thresholds. Everyone does. Perhaps it is not possible to make it work and suddenly work everywhere. There are many things that need to be done to drive into a city-which means betting your client’s life and bet your business-just making the software quite good. Two companies had serious incidents (including fatal) with pedestrians, and both projects were closed and stripped for pieces. The shares are very high, but so is the reward. The “Revolution” in Shashua’s graph is true, and ideally many many companies will reach it and compete. (Surely not “level 5”, though. Due to the reduction of returns, trying to complete the long tail of all roads with a certified security guarantee is far from in the future, if ever comes.