Wait... let's rewind. Why does Intel have an "Israeli automotive hub" from acquiring "Mobileye, the autonomous driving company" in 2017?
It makes sense when a large company acquires a smaller one, or founds a division from scratch, because it can provide that smaller company/company with uniquely valuable resources. E.g. Google/Alphabet founding Waymo because of Google's existing expertise in ML.
But how does expertise in processors translate to any kind of advantage for autonomous driving? Presumably the microprocessors in a self-driving car are one of the most commoditized parts... who cares if it's Intel, AMD or ARM?
Does anyone "in the know" have a clue around Intel's strategy here?
I can only guess they want to repeat the "Intel Inside" marketing campaign of the 1990's -- i.e. if they lost mobile, they don't want to lose cars too. But then wouldn't they want to form strong subsidized partnerships, the way they did with Intel Inside? Developing their own technology seems like it would make the autonomous car industry view them as competitors, not partners, and thus less likely to use their chips.
NVidia has product lines specifically for this field. Tesla proudly announced to the world that they made their own chip for autonomous driving. "Autonomous driving" research/development right now means lots of Machine Learning, image processing, ... High-performance stuff that benefits from GPU-like designs, all the while the systems for that need to be reasonably energy-efficient and build in a way that they'll survive in an automotive setting. They're not a competitor to car companies because car companies typically don't want to design chips themselves, they'll happily buy them if they can (Tesla is a really odd one out in that regard, and even they'll use superior hardware made by someone else if it makes sense). If it takes off across manufacturers, Intel very much would like a slice of that pie instead of leaving it all to NVidia&Co.
Sure, NVidia has chips that fit the needs of that field. Intel though has bought the actual company.
My only explanation is that Intel may believe they would have trouble breaking into that market with their own chips, and buying the company means they can make it dogfood Intel's products. If successful, it strengthens their sales argument to others.
Their competitor here is Nvidia, which has quite good hardware. Also, yes the ISA doesn't matter that much in the field, because no 3rd-party user-installed apps on automotive by design...
MobilEye's key advantage was they fabbed their own chips. Back when you couldn't really run a neural net in a car, they essentially kind of hard coded the evaluation of a certain neural net into a special chip.
Not sure I described this 100% accurately, but it's something like that. Hence the Intel acquisition. Their tech is pretty good honestly, used to power Tesla Autopilot 1 before they decided to build their own solution.
Actually one of Mobileye’s big strengths is their custom chips. They are purpose built and very efficient, which allows them to mount them under the windshield (important for their main business, ADAS), and they are built with the needs of the software in mind. I don’t know if Mobileye uses any Intel fabs, but it’s not such a crazy stretch to say one could help the other.
Intel does more than processors though. OpenCV itself originated at Intel and they still have an image processing library and several models of depth cameras with an SDK for face recognition, a physically based renderer and probably many other stuff.
It makes sense when a large company acquires a smaller one, or founds a division from scratch, because it can provide that smaller company/company with uniquely valuable resources. E.g. Google/Alphabet founding Waymo because of Google's existing expertise in ML.
But how does expertise in processors translate to any kind of advantage for autonomous driving? Presumably the microprocessors in a self-driving car are one of the most commoditized parts... who cares if it's Intel, AMD or ARM?
Does anyone "in the know" have a clue around Intel's strategy here?
I can only guess they want to repeat the "Intel Inside" marketing campaign of the 1990's -- i.e. if they lost mobile, they don't want to lose cars too. But then wouldn't they want to form strong subsidized partnerships, the way they did with Intel Inside? Developing their own technology seems like it would make the autonomous car industry view them as competitors, not partners, and thus less likely to use their chips.
What am I missing?