WEBVTT 00:00:00.000 --> 00:00:05.120 NVIDIA is trying to strong ARM its way into more devices than ever before, and yes, if 00:00:05.120 --> 00:00:09.600 you can forgive that terrible, terrible pun, today's video is going to be about NVIDIA's 00:00:09.600 --> 00:00:15.640 acquisition of ARM. Now, at first, the two companies don't seem to be likely partners given that they mostly 00:00:15.640 --> 00:00:21.920 operate in different spaces, NVIDIA in dedicated graphics and ARM in more general purpose systems 00:00:21.920 --> 00:00:28.360 on a chip for smartphones and mobile devices. But like many tech acquisitions, this one is about NVIDIA getting a foothold in what 00:00:28.360 --> 00:00:36.560 of things will be a lucrative market in the future. Not only are ARM designs at the heart of nearly all smartphones, both iOS and Android-based, 00:00:36.560 --> 00:00:42.320 but they're in IoT devices, embedded systems, and even in AI applications, which is something 00:00:42.320 --> 00:00:49.080 right in NVIDIA's wheelhouse. You see, even though NVIDIA is better known for consumer graphics, GPUs are actually powerful 00:00:49.080 --> 00:00:54.560 chips for machine learning, too, because they're very good at processing parallel workloads. 00:00:54.560 --> 00:00:58.720 That is, workloads that can be chopped up into pieces with the processor working on multiple 00:00:58.720 --> 00:01:05.240 pieces simultaneously. And NVIDIA has already leveraged this fact to make products for deep learning in self-driving 00:01:05.240 --> 00:01:14.000 cars, retail, and surveillance. But ARM gives them a huge leg up in the AI market. 00:01:14.000 --> 00:01:18.540 Although the phrase machine learning might conjure up images of large-scale server farms 00:01:18.540 --> 00:01:23.640 working on incredibly complicated problems, AI on a much smaller scale is starting to become 00:01:23.640 --> 00:01:30.440 more common and more important. In recent years, ARM has released machine learning capable processors for smartphones 00:01:30.440 --> 00:01:35.160 and IoT devices with the goal of allowing these gadgets to learn on their own without 00:01:35.160 --> 00:01:41.780 having to rely on the cloud. Think about photo-enhancing algorithms for phones or object detection on an autonomous 00:01:41.780 --> 00:01:47.080 car that will work even if you're in an area with little to no network coverage. 00:01:47.080 --> 00:01:51.240 But because NVIDIA has been in the GPU business for such a long time, the company is in a 00:01:51.240 --> 00:01:56.880 good position to use ARM as a vehicle to get more advanced AI solutions into many more 00:01:56.880 --> 00:02:02.800 devices. Although NVIDIA has actually tried to get its ARM-based Tegra chips into phones in the 00:02:02.800 --> 00:02:07.000 past, they didn't have much success partly because of price. 00:02:07.000 --> 00:02:10.760 But this time, NVIDIA is just buying the whole enchilada. 00:02:10.760 --> 00:02:16.080 On top of that, remember that ARM's business model is not to make the chips themselves, 00:02:16.080 --> 00:02:20.320 but to license out their designs to a huge array of other companies. 00:02:20.520 --> 00:02:24.880 Because ARM has so many licensing agreements in place already, NVIDIA might be able to 00:02:24.880 --> 00:02:28.360 quickly start competing in lots of different markets. 00:02:28.360 --> 00:02:36.360 And it isn't just about the small gadgets either. In fact, NVIDIA is planning to build an ARM-based supercomputer for AI research and machine learning, 00:02:36.360 --> 00:02:41.080 and yes, it's similar to that thing in sci-fi where a machine figures out how to make more 00:02:41.080 --> 00:02:48.960 of itself. But such a momentous merger isn't just going to happen without the world raising some eyebrows. 00:02:49.040 --> 00:02:52.920 Although this is the case with a lot of large mergers due to concerns about one company 00:02:52.920 --> 00:02:58.080 dominating a certain market, there's also another wrinkle in that even though ARM is 00:02:58.080 --> 00:03:03.320 a private firm, they've been called the Switzerland of the semiconductor industry. 00:03:03.320 --> 00:03:07.640 Because the company is rather neutral about who they license their chip designs to. 00:03:07.640 --> 00:03:12.720 For example, both Samsung and Qualcomm use ARM-based chips even though they're directly 00:03:12.720 --> 00:03:19.720 competing in the smartphone space. So the concern is that although NVIDIA has promised to keep ARM neutral for the time 00:03:19.720 --> 00:03:25.320 being, NVIDIA may not want to license ARM technology to companies it's in more direct 00:03:25.320 --> 00:03:32.760 competition with down the line. A possibility that may hurt NVIDIA as the deal still has to clear regulatory hurdles 00:03:32.760 --> 00:03:36.840 in order to be approved. But if it does get approved, who knows? 00:03:36.840 --> 00:03:41.080 Maybe one day you'll be able to carry around a full-on gaming rig in your pocket, assuming 00:03:41.080 --> 00:03:44.160 it doesn't literally burn a hole through your trousers. 00:03:44.160 --> 00:03:51.600 So thanks for watching guys, if you liked this video give us a thumbs up, hit subscribe and be sure to check us out in the comments section where you can give us tips and suggestions 00:03:51.600 --> 00:03:53.480 for topics that we should cover in the future.