Comments Locked

16 Comments

Back to Article

  • ikjadoon - Monday, March 13, 2017 - link

    I usually don't post grammatical or spelling comments. But, this is rough:

    >Intel will comments a tender offer for all issued and outstanding ordinary shared of Mobileye

    This sentence could've been posted on a Google-translated blogspam website and I wouldn't have batted an eye.
  • Ian Cutress - Monday, March 13, 2017 - link

    Sorry, early morning and written in a rush. Will fix :)
  • ikjadoon - Thursday, March 16, 2017 - link

    No worries, we appreciate it. I think this announcement by Intel wasn't as expected as we usually are, given the amount of leaks.

    And I appreciate the slide deck: it's always confidence-inspiring to see what's coming out of the horse's mouth, too.
  • Meteor2 - Monday, March 13, 2017 - link

    Fascinating. MobilEye has partnered with Delphi (auto components, annual revenue $17bn) aiming to offer OEM self-drive systems by 2019, with retail after that.

    This basically puts Intel into a direct fight with Nvidia for self-drive tech. I don't count Waymo or Uber as their LIDAR-based approaches are dead-ends compared to MobilEye's and Nvidia visual cameras-fused-with-radar-and-sonar approach.
  • shabby - Monday, March 13, 2017 - link

    Has intel stopped innovating? Seems they're buying up companies that are "sure wins" in the future, yet are late doing so.
  • HomeworldFound - Monday, March 13, 2017 - link

    It's sensible. If cars are going to drive themselves, they'll need the technology to do so. There's a nice space for a CPU/SOC and connected services, and it's better that those come from a great company as opposed to some cheap Chinese company.
  • name99 - Monday, March 13, 2017 - link

    That's not the issue. The issue is why does Intel imagine that they are better positioned to take advantage of this market than other investment alternatives for their cash?

    The auto market is essentially identical to the mobile and IoT markets. It has zero installed base, so no interest in x86 legacy code. Intel competes on a level platform with ARM, while dragging the excess costs (in design time and turn-araound time) of x86.
    Likewise the market is large companies (automakers) with specific needs, not individuals. ARM's approach of being able to tailor a SoC (12 high-end cores, 6 low-end cores, 4 DSPs, 2 GPUs, and a partridge in a pear tree) to exactly what Mercedes Benz wants makes sense. Intel's approach is to provide some pre-configured more-or-less random collection of elements on a SoC (driven by marketing on the one hand, and whatever binning yields on the other hand) and Mercedes is expected to just accept whatever in the line-up best meets their needs.

    I don't see anything in Intel today that shows that they have learned from their mobile and IoT failures, and I don't see anything that explains why auto will play out differently.
  • Meteor2 - Monday, March 13, 2017 - link

    Intel has zillions of dollars outside the US which it doesn't want to lose a large portion of to tax, so, from the perspectives of a shareholder, why not buy up sure-fire bets?

    But Intel have stated that their future growth isn't going to come from CPUs, it's going to come from new areas like 5G and AI.
  • prisonerX - Monday, March 13, 2017 - link

    Intel has been innovating?
  • Stan11003 - Tuesday, March 14, 2017 - link

    This is a pretty old tactic for intel. The very core based chips we have today came from a start-up israel lab.
    http://www.seattletimes.com/business/how-israel-sa...

    It makes sense with a company the size and complexity of Intel acquiring star ups helps them stay nimble without risk.
  • zodiacfml - Monday, March 13, 2017 - link

    Intel's on a buying spree. It is a good area to expand on though and they have it quite correct that within a decade, the industry will be worth it. They will not dominate it though. As the gaming market continues to grow, a GPU will provide the best value for such product/service.
  • Yojimbo - Monday, March 13, 2017 - link

    It looks like a mix of a CPU, an ASIC and a more general parallel processor might be the way things go for computer vision AI applications. NVIDIA seems to be optimizing their Denver CPU core for automotive, they have their GPU for a general parallel processor, and they have developed an ASIC for convolutional neural network heavy lifting. The critical components are close to their core competencies, they have a strong software development platform, and they have good momentum. Intel has CPUs, obviously. They have a few initiatives where they are working towards the ASIC (Movidius and Nervana), and they seem to hope to compete with GPUs by using FPGAs.

    I give NVIDIA the advantage but Intel definitely has good prospects. Mobileye has established business relationships and industry expertise. It's unknown how well Intel can integrate the various pieces they've acquired, how Intel's hardware ends up performing, or how easy it will be to develop for. At the moment Intel's position seems stronger to me than Qualcomm's, the third company going after the market. Qualcomm may end up dominating the AI inferencing in the majority of IoT devices, though. These devices will either only need ARM processing or will rely on mobile connectivity to a cloud server for most of the actual processing. But I think devices requiring significant edge processing will use processors from NVIDIA or Intel. Intel has their own mobile modems so they do have at least that advantage over NVIDIA.
  • Meteor2 - Monday, March 13, 2017 - link

    I agree with your analysis, but what strikes me about self-driving is that Tesla says it's a software problem, and the hardware is already good enough. They put a DrivePX2 and a bunch of cameras, sonars and a radar in every vehicle they're producing and say that's enough to enable self-driving with the right software. If Tesla are right, and I wouldn't bet against them, it does make you wonder about the prospects of MobilEye/Intel, who *aren't* shipping hard way 'good enough' for self-drive. Will auto makers figure out the software side themselves as Volvo and Audi are trying, like Tesla? Will they just wait for a COTS solution? Or will Tesla take over the world??
  • shabby - Monday, March 13, 2017 - link

    I wouldn't believe everything elon says, while i do applaud what he's doing with his money he's also trying to earn money selling his wares, and that means making outlandish claims. Missions to mars... please.
  • p1esk - Monday, March 13, 2017 - link

    Nvidia developed an ASIC for convolutional neural network heavy lifting? What are you talking about?
  • asmian - Monday, March 13, 2017 - link

    I have to seriously wonder about the drive to make these vehicles reliant on the cloud. While weather and traffic status updates are obviously good things for a self-driving car to be pulling from the net, I seriously hope the driving logic is going to be completely local to (inside) the car. What is all this proposed mass of data being used for if not generating local driving decisions?

    Quite apart from making it a dangerous moving brick if the internet connection goes down (even a latency glitch could be critical for avoiding an accident), the vulnerability to hacking or surveillance/government override like all current IoT products is far too scary to make me comfortable with getting in one, frankly. I'd want to know the cord can be cut completely and it can still be an autonomous, fully functional self-driving car. If the software ain't that good, then it's not ready.

    If we're set to end up with JohnnyCabs, it'd damn well better be Johnny driving, not some remote server. :/

Log in

Don't have an account? Sign up now