WEBVTT - The Rise and Rise of ARM Holdings 0:00:04.400 --> 0:00:07.800 Welcome to tech Stuff, a production from I Heart Radio. 0:00:12.119 --> 0:00:14.840 Hey there, and welcome to tech Stuff. I'm your host, 0:00:14.960 --> 0:00:17.760 Jonathan Strickland. I'm an executive producer with I Heart Radio 0:00:17.800 --> 0:00:21.080 and I love all things tech. And we are back 0:00:21.480 --> 0:00:26.480 with the story about ARM ARM Limited in later ARM Holdings, 0:00:27.120 --> 0:00:29.840 and the family of processors that grew out of a 0:00:29.880 --> 0:00:34.240 British computing company from the nineteen seventies. If you haven't 0:00:34.280 --> 0:00:36.879 listened to the first part of this, I recommend you 0:00:36.920 --> 0:00:41.000 go back and listen to that one first because it's sequential. 0:00:41.400 --> 0:00:43.680 But we left off in our last episode with the 0:00:43.800 --> 0:00:48.960 launch of the ARMS six family of microprocessors, confusingly the 0:00:49.040 --> 0:00:53.000 fourth generation of them, and the failure of the Apple 0:00:53.120 --> 0:00:58.160 Newton product, which partly prompted ARM CEO Robin Saxby to 0:00:58.240 --> 0:01:02.120 adopt a strategy to license our architecture technology to other 0:01:02.200 --> 0:01:08.520 companies back in in other words, to license the the 0:01:08.600 --> 0:01:13.120 intellectual property to other fabricators, so ARM would not be 0:01:13.720 --> 0:01:18.360 a manufacturing company. It would design microprocessors, and it would 0:01:18.400 --> 0:01:23.039 design instruction sets for those microprocessors, and then license that 0:01:23.120 --> 0:01:28.199 information to other companies that could then make these chips themselves, 0:01:28.480 --> 0:01:31.360 and even make them and sell them to other customers. 0:01:31.360 --> 0:01:33.800 That's kind of how this could work, and ARM would 0:01:33.840 --> 0:01:37.800 end up receiving an upfront licensing fee and some royalties 0:01:37.840 --> 0:01:42.600 as well, joining v L s I, which had previously 0:01:42.680 --> 0:01:46.760 been the sole fabricator of ARM chips, where the Japanese 0:01:46.760 --> 0:01:50.480 company Sharp and the British semiconductor company g E C 0:01:50.720 --> 0:01:54.360 plus E Semiconductors. Now this would just be the beginning 0:01:54.440 --> 0:01:58.200 of companies manufacturing our microprocessors. And as I mentioned, the 0:01:58.280 --> 0:02:01.320 licensing deal had two parts them. First was the upfront 0:02:01.440 --> 0:02:04.240 licensing fee for the technology, so this was a flat fee, 0:02:04.840 --> 0:02:08.280 and if a company like Sharp wanted to manufacture ARMED chips, 0:02:08.280 --> 0:02:10.720 it would first have to pay that flat fee in 0:02:10.840 --> 0:02:13.240 order to do so. But on the back end was 0:02:13.280 --> 0:02:16.120 a royalty fee which might not kick in for a 0:02:16.160 --> 0:02:18.520 few years, in fact, in some cases up to five 0:02:18.639 --> 0:02:21.480 years after products were being sold. That's when the royalties 0:02:21.480 --> 0:02:24.040 would start to kick in. So as companies would sell 0:02:24.360 --> 0:02:29.240 Armed chips or selling products that used ARM designed chips, 0:02:29.680 --> 0:02:33.839 then ARM would end up earning a small amount on 0:02:34.000 --> 0:02:39.320 those sales. Would prove to be an enormous year for ARM. 0:02:39.360 --> 0:02:41.880 Not only did it see the launch of the ARMS 0:02:41.919 --> 0:02:45.080 six in the form of the first Apple Newton, which 0:02:45.400 --> 0:02:49.320 admittedly is perhaps not the most auspicious of beginnings, and 0:02:49.400 --> 0:02:52.680 not only did it transition into a sort of intellectual 0:02:52.720 --> 0:02:57.079 property company, it also landed a deal with Texas Instruments, 0:02:57.160 --> 0:03:00.920 a company famous for lots of electronic devices is including 0:03:00.960 --> 0:03:06.000 popular scientific calculators. T I would license the ARM architecture 0:03:06.040 --> 0:03:08.520 and would become not only a valuable partner for ARM, 0:03:08.919 --> 0:03:11.560 but also a sort of almost like a sales rep. 0:03:11.639 --> 0:03:14.120 Like I was saying. You know, companies would make ARM 0:03:14.200 --> 0:03:18.359 processors and then some sometimes sell those to other companies. 0:03:18.840 --> 0:03:23.200 So when hand set manufacturer Nokia began looking for a 0:03:23.240 --> 0:03:29.240 processor solution for its upcoming mobile phones, Texas Instruments suggested ARM. 0:03:29.320 --> 0:03:32.320 Nokia balked at the idea at first because the cost 0:03:32.360 --> 0:03:35.440 of the ARMS six chips was higher than the company 0:03:35.440 --> 0:03:38.720 wanted to pay for components for its handsets. It would 0:03:38.720 --> 0:03:42.200 mean either Nokia would have to eat that cost, thus 0:03:42.280 --> 0:03:46.160 cutting into profits, maybe even leading at selling at a loss, 0:03:46.160 --> 0:03:49.360 which isn't really acceptable, or they would have to price 0:03:49.400 --> 0:03:52.200 the handsets at a level higher than the market would 0:03:53.040 --> 0:03:57.080 maybe support so they could price themselves out. The phones 0:03:57.080 --> 0:03:59.560 would be too expensive, no one would buy them. ARMS 0:03:59.560 --> 0:04:01.920 response was quick, which was helped by the fact that 0:04:01.960 --> 0:04:05.200 the team at ARM was still relatively small. The company 0:04:05.240 --> 0:04:07.720 could actually be really nimble and respond to things quickly, 0:04:08.200 --> 0:04:11.400 and they got to work designing a new instruction set, 0:04:11.680 --> 0:04:15.360 one that was just sixteen bits rather than thirty two. 0:04:15.400 --> 0:04:17.960 If you need to have a refresher on that, listen 0:04:18.000 --> 0:04:22.039 to the last podcast. But it lowered the memory demands 0:04:22.279 --> 0:04:25.360 for the technology and thus reduced the cost of the 0:04:25.400 --> 0:04:29.280 components required, and it was called the ARM seven T 0:04:29.640 --> 0:04:32.599 d M I. The T d M I stands for 0:04:32.600 --> 0:04:36.360 a thumb debug fast Multiplier and enhanced i c e. 0:04:36.960 --> 0:04:39.400 The only bit I really want to talk about among 0:04:39.640 --> 0:04:42.880 those right now is thumb and that was a special 0:04:42.920 --> 0:04:46.359 instruction set, and it was this instruction set that allowed 0:04:46.360 --> 0:04:50.000 for the sixteen bit approach. To go into more detail 0:04:50.279 --> 0:04:53.600 would frankly require a deeper understanding than I possess, So 0:04:53.720 --> 0:04:58.039 rather than fumble about with attempted explanations, I'll just leave 0:04:58.040 --> 0:05:01.000 it at this. Essentially, it was is able to take 0:05:01.720 --> 0:05:05.880 thirty two bits set of instructions and dumb it down 0:05:05.960 --> 0:05:11.560 to sixteen bit. That is drastically oversimplifying what it did, 0:05:11.920 --> 0:05:15.680 but it's it serves its purpose for this episode, Texas 0:05:15.720 --> 0:05:19.080 Instruments licensed the technology from ARM and then sold the 0:05:19.120 --> 0:05:22.280 chips to Nokia, and then Nokia then used that chip 0:05:22.400 --> 0:05:26.920 to power the Nokia sixty one ten hand set, which 0:05:27.040 --> 0:05:31.640 was phenomenally successful in Europe. Nokia would introduce a version 0:05:31.760 --> 0:05:34.240 of this phone for the North American market with some 0:05:34.400 --> 0:05:37.280 minor changes to the hand set. This one was called 0:05:37.279 --> 0:05:40.760 the six nine, and the candy bar style phone would 0:05:40.760 --> 0:05:43.400 become one of the early big successes in cellular phones. 0:05:43.440 --> 0:05:46.000 I remember having a Nokia. I don't think it was 0:05:46.080 --> 0:05:48.080 this model. I think it was actually a little later, 0:05:48.480 --> 0:05:51.320 but I remember playing Snake on it. You guys remember 0:05:51.320 --> 0:05:54.919 Snake Anyway. More than that, the success of the Nokia 0:05:55.000 --> 0:05:58.359 sixty one ten led to the big success for ARM 0:05:58.400 --> 0:06:00.919 as the ARMS seven technology. She became the go to 0:06:01.120 --> 0:06:04.520 architecture for mobile phones, and to call it a success 0:06:04.640 --> 0:06:09.360 really understates things. More than ten billion with a B 0:06:09.880 --> 0:06:13.200 ARMS seven chips were produced since they were first introduced 0:06:13.200 --> 0:06:18.640 in nineteen and the Nokia phone itself wouldn't debut until now. 0:06:18.720 --> 0:06:21.080 Keep in mind those chips are being made by companies 0:06:21.080 --> 0:06:24.159 all around the world. ARM is licensing the design to 0:06:24.240 --> 0:06:27.440 those companies now in the meantime, ARM was developing a 0:06:27.480 --> 0:06:32.480 new generation of this technology, ARM eight, which confusingly marks 0:06:32.520 --> 0:06:36.680 the fourth generation of ARM processors, would emerge in nineteen 0:06:36.839 --> 0:06:40.360 nine six. While it followed the same process as the 0:06:40.520 --> 0:06:43.880 ARMS seven ten line, it packed in twice the performance 0:06:43.880 --> 0:06:49.400 with a more optimized design. It also revamped the instruction pipeline, which, hey, 0:06:49.400 --> 0:06:52.360 that opens up another opportunity to talk about how that works. 0:06:53.120 --> 0:06:57.279 The idea behind an instruction pipeline is to maximize the 0:06:57.400 --> 0:07:01.359 usefulness of a processor by delivering LLL instructions to the 0:07:01.400 --> 0:07:06.160 processor simultaneously. So rather than have parts of the processor 0:07:06.360 --> 0:07:09.080 kind of going dormant because they're not needed for a 0:07:09.080 --> 0:07:12.560 particular process, the goal was to keep all the parts 0:07:12.560 --> 0:07:15.280 of the processor busy in an effort to reduce the 0:07:15.320 --> 0:07:18.400 amount of time that the processor takes to complete a 0:07:18.480 --> 0:07:22.880 certain task or program. Pipelines do this by delivering sequential 0:07:22.920 --> 0:07:27.000 steps for specific tasks that different processor units can handle. 0:07:27.680 --> 0:07:31.920 The ARMS seven architecture had a three stage pipeline divided 0:07:32.000 --> 0:07:36.360 into the instructions for fetch, decode, and execute. These are 0:07:36.680 --> 0:07:41.480 basic instructions. The ARM eight pipeline introduced a five stage 0:07:41.480 --> 0:07:45.679 approach which added instructions for memory access and writing to memories, 0:07:45.720 --> 0:07:49.000 so essentially reading from memory and writing to it. Really, 0:07:49.080 --> 0:07:52.280 what this means is that the design optimizes how instructions 0:07:52.400 --> 0:07:54.840 arrive at the processor and how much more of the 0:07:54.840 --> 0:07:57.840 processor is in use at any given moment to complete 0:07:57.880 --> 0:08:02.400 tasks faster. As Night was coming to a close, ARM 0:08:02.600 --> 0:08:06.160 was entering a new phase of its existence. It had 0:08:06.280 --> 0:08:10.120 up to that point been a privately held company. The 0:08:10.160 --> 0:08:15.360 income for that year measured up to around two point 0:08:15.480 --> 0:08:19.160 nine million pounds, which if we converted that to US 0:08:19.240 --> 0:08:23.240 dollars and adjusted for inflation, is somewhere around seven and 0:08:23.280 --> 0:08:26.520 a half million dollars. But to be accurate, I really 0:08:26.560 --> 0:08:29.840 should just give a range of values, because we're dealing 0:08:29.920 --> 0:08:33.440 with both, you know, translating from one form of currency 0:08:33.520 --> 0:08:36.640 to another, plus adjusting for inflation. So really it could 0:08:36.640 --> 0:08:39.199 be anywhere between six point four and nine point one 0:08:39.240 --> 0:08:42.280 million dollars, which is pretty big range. The company itself 0:08:42.320 --> 0:08:47.400 was worth in excess of twenty million pounds, a princely sum. Indeed, 0:08:47.760 --> 0:08:51.439 it was time to take the company public. Now I've 0:08:51.480 --> 0:08:55.040 talked about this process in other episodes of Tech Stuff, 0:08:55.240 --> 0:08:59.280 but generally speaking, the process of taking up private company public. 0:09:00.000 --> 0:09:02.199 First of all, It involves an awful lot of paperwork, 0:09:02.480 --> 0:09:05.880 as people have to determine the value of the company 0:09:05.960 --> 0:09:08.920 when all its assets and debts are taking into account, 0:09:09.559 --> 0:09:12.920 and that value in turn guides how many shares and 0:09:12.920 --> 0:09:15.920 how much money per share will be on offer when 0:09:15.920 --> 0:09:18.800 the company goes public and people can can buy a 0:09:18.880 --> 0:09:21.880 stake in the company. The idea here is that the 0:09:22.000 --> 0:09:24.520 number of shares and the price per share will reflect 0:09:24.559 --> 0:09:27.440 the value of the company. You know, you can't just 0:09:27.600 --> 0:09:31.520 keep printing out shares at the same price, because it 0:09:31.679 --> 0:09:35.440 all has to kind of link back to how much 0:09:35.480 --> 0:09:38.319 is the company actually worth. It's actually a pretty complicated thing, 0:09:38.880 --> 0:09:41.960 and it's partly dependent upon our perception of a company's 0:09:41.960 --> 0:09:45.680 worth as opposed to a you know, hard, concrete, universal, 0:09:45.760 --> 0:09:48.760 quote unquote real number. Now, at this point, I think 0:09:48.760 --> 0:09:52.559 it's good to reflect a moment on Acorn Computers. If 0:09:52.559 --> 0:09:55.760 you listen to the first episode, you know that ARM 0:09:56.040 --> 0:09:59.320 spun off from Acorn Computers, which was a company that 0:09:59.480 --> 0:10:02.800 used the risk based processors and the personal computers that 0:10:02.880 --> 0:10:07.520 manufactured the Archimedes computers. Well, things had taken a turn 0:10:07.559 --> 0:10:10.280 over at Acorn as ARM was making a move to 0:10:10.320 --> 0:10:15.000 go public. Much of Acorn's management was focused on that process. Now, remember, 0:10:15.320 --> 0:10:19.200 ARM was originally a joint venture between Acorn, Apple Computers 0:10:19.280 --> 0:10:22.319 and v L s I, which was a fabrication company. 0:10:22.400 --> 0:10:27.040 Acorn management changed and after the company experienced a substantial 0:10:27.120 --> 0:10:31.839 loss in the new management decided on a massive restructuring 0:10:32.080 --> 0:10:35.760 that involved selling off several divisions to other companies and 0:10:35.840 --> 0:10:40.760 ultimately renaming Acorn Computers itself. The company transformed into a 0:10:40.800 --> 0:10:43.959 new one called Element fourteen. Now, if you take a 0:10:44.040 --> 0:10:47.440 quick glance at the periodic table, which I imagine you 0:10:47.520 --> 0:10:50.400 keep handy just as I do, you will see that 0:10:50.480 --> 0:10:55.840 Element fourteen is Silicon Clever. Element fourteen wouldn't be around 0:10:56.040 --> 0:10:59.320 in that form for very long as an Element fourteen 0:10:59.360 --> 0:11:02.240 the company Silicon, As it turns out, it's still in 0:11:02.280 --> 0:11:07.120 good shape. But Broad Calm Corporation acquired Element fourteen in 0:11:07.240 --> 0:11:11.000 two thousand. Broad Calm, in turn would later be acquired 0:11:11.040 --> 0:11:15.559 by Evago Technologies, though it still operates as Broad Calm Corporation. 0:11:15.960 --> 0:11:19.760 But this all reminds us that, as Quigon Jen once 0:11:19.840 --> 0:11:25.600 said famously, there's always a bigger fish. While Acorn Computers 0:11:25.679 --> 0:11:30.840 effectively vanished due to acquisitions, ARM continued to go strong 0:11:31.120 --> 0:11:34.400 The company listed on both the London Stock Exchange and 0:11:34.480 --> 0:11:39.520 the NASDAQ on April seventeenth, changing its name to ARM 0:11:39.679 --> 0:11:43.360 Holdings in the process. The reason for listing on both 0:11:43.360 --> 0:11:47.800 stock exchanges was due to a mixture of practicality and appreciation. 0:11:48.120 --> 0:11:52.040 America was leading in the text space, so listing on 0:11:52.080 --> 0:11:56.200 the NAZDAC just made sense, but the ARM executives also 0:11:56.240 --> 0:11:59.920 wanted Acorn shareholders back in England to remain involved with 0:12:00.080 --> 0:12:03.680 the company. Interest in the company pushed the stock price higher, 0:12:03.679 --> 0:12:06.880 and before long, ARM went from being a twenty million 0:12:06.960 --> 0:12:10.360 pound company to a company that was valued at more 0:12:10.400 --> 0:12:14.800 than a billion dollars. However, not all of that would 0:12:14.880 --> 0:12:17.440 end up being good news, at least in the short term. 0:12:17.640 --> 0:12:20.480 ARMS rise and valuation was in step with a general 0:12:20.520 --> 0:12:24.719 trended technology that proved to be unsustainable, and I'll get 0:12:24.760 --> 0:12:27.800 back to that in just a moment. Another thing that 0:12:27.880 --> 0:12:32.520 happened in was the release of the ARM nine group 0:12:32.600 --> 0:12:36.600 of processor cores. This move also saw a departure from 0:12:36.640 --> 0:12:40.840 the way ARM had been designing processors. Up through ARM eight. 0:12:41.120 --> 0:12:44.600 The company had followed a von Neumann model sometimes called 0:12:44.920 --> 0:12:48.440 the Princeton model, and the basic concept of this model 0:12:48.480 --> 0:12:50.800 is that you have a system that has a control unit, 0:12:51.240 --> 0:12:55.559 a logic unit, and a memory unit. Input would prompt 0:12:55.640 --> 0:12:59.480 the system to perform operations on data, which results in output, 0:12:59.720 --> 0:13:02.079 which in turn gets sent to some sort of output 0:13:02.120 --> 0:13:04.920 device like a display or a printer. And it's named 0:13:04.960 --> 0:13:08.280 after John von Neumann, who described such a system in 0:13:08.320 --> 0:13:11.679 a paper called the First Draft of a Report on 0:13:11.720 --> 0:13:16.319 the EDVAC. EDVAC was an early electronic computer that ran 0:13:16.360 --> 0:13:20.520 on binary data. With a von Neumann system, a processor 0:13:20.679 --> 0:13:25.600 cannot both fetch instructions and run a data operation at 0:13:25.600 --> 0:13:29.160 the same time because both of these processes share the 0:13:29.240 --> 0:13:32.240 same bus. Now, you can think of a bus as 0:13:32.280 --> 0:13:35.800 a data pathway, and it connects different components in a 0:13:35.840 --> 0:13:40.880 computer system or a circuit, and you can't have two 0:13:40.960 --> 0:13:44.439 things share the same pathway at the same time. Typically, 0:13:45.040 --> 0:13:47.320 I often think of it similar to like pipes in 0:13:47.360 --> 0:13:52.640 a plumbing system. The ARM nine family of processors changed 0:13:52.679 --> 0:13:57.640 to a Harvard architecture, which has dedicated buses for stuff 0:13:57.679 --> 0:14:01.520 like memory and fetching instructions. And since this approach has 0:14:01.600 --> 0:14:05.840 paths specifically for each of those tasks, there's no bottleneck 0:14:05.920 --> 0:14:09.240 if you need to do both of them simultaneously. Now, 0:14:09.320 --> 0:14:13.280 for certain implementations that doesn't really matter that much, so 0:14:13.440 --> 0:14:17.280 von Neumann approach is perfectly fine in some cases, but 0:14:17.360 --> 0:14:21.160 as you get into circuit complexity, being able to fetch 0:14:21.160 --> 0:14:24.440 instructions and read or write data to memory can really 0:14:24.480 --> 0:14:27.560 speed things up. ARM recognized that it was time to 0:14:27.600 --> 0:14:32.160 transitionto a more robust circuit design as lightweight computational devices 0:14:32.160 --> 0:14:36.000 were putting heavier requirements on processors. The ARM nine chips 0:14:36.240 --> 0:14:40.440 also saw improvements in heat production, meaning they were producing 0:14:40.760 --> 0:14:45.920 less heat than their comparable ARMS seven predecessors. ARM ten 0:14:46.280 --> 0:14:49.120 would not be far behind. It launched in late nine, 0:14:50.440 --> 0:14:53.920 and as you would expect, the new micro architecture included 0:14:53.960 --> 0:14:58.320 some benefits over older designs, but frankly, there wasn't anything 0:14:58.400 --> 0:15:00.760 so incredible that I feel that I should really break 0:15:00.840 --> 0:15:03.000 it out in this episode. We'll just keep in mind 0:15:03.040 --> 0:15:05.760 that the company was keeping up its process of research 0:15:05.800 --> 0:15:08.960 and development and then licensing the resulting designs to various 0:15:09.000 --> 0:15:13.600 fabricators around the world, because otherwise this episode gets way 0:15:13.640 --> 0:15:18.359 too dry away too fast. In the company made an acquisition. 0:15:18.520 --> 0:15:23.480 It purchased a software consulting company called Micrologic Solutions. Like 0:15:23.680 --> 0:15:26.920 ARM Holdings, this company was also based out of Cambridge, England. 0:15:27.280 --> 0:15:31.680 The following year it made three more acquisitions. Euromps that 0:15:31.800 --> 0:15:34.480 was a company out of France that designed smart cards. 0:15:34.520 --> 0:15:38.200 Those are integrated circuit cards have an embedded circuit chip. 0:15:38.640 --> 0:15:41.960 They act as a kind of authorization device. The second 0:15:42.160 --> 0:15:44.680 was a Lot Software. It was a company that made 0:15:44.720 --> 0:15:48.840 debugging software, and the third was Infinite Designs, another English 0:15:48.840 --> 0:15:52.320 design company, And in two thousand one the company purchased 0:15:52.320 --> 0:15:56.960 a division out of Neural Micrologics, another debugging company in England. 0:15:58.280 --> 0:16:01.640 But around this time the X sector had become the 0:16:01.800 --> 0:16:06.200 new gold Rush, with entrepreneurs launching new companies by the day, 0:16:06.320 --> 0:16:08.280 many of which were tied in some way to the 0:16:08.320 --> 0:16:11.960 growth of the Internet. In particular, in the late nineties, 0:16:12.040 --> 0:16:15.160 the general public was really just starting to understand that 0:16:15.240 --> 0:16:18.840 the Internet was a thing. Previously, the Internet had largely 0:16:18.880 --> 0:16:22.840 been the stuff of research labs, college computer labs, some 0:16:23.040 --> 0:16:27.680 specific industries, the American military. But then we got the 0:16:27.680 --> 0:16:31.240 Worldwide Web in the early nineties, and slowly but surely, 0:16:31.960 --> 0:16:34.280 the average person began to catch on to what the 0:16:34.320 --> 0:16:37.360 Internet was. And one thing lots of people were sure 0:16:37.360 --> 0:16:40.000 about was that the Internet was the future of all 0:16:40.080 --> 0:16:45.720 commerce and business. You could argue convincingly that they were 0:16:45.920 --> 0:16:50.200 all right. But the problem was, back then there wasn't 0:16:50.240 --> 0:16:53.720 as much of an understanding around how this would come about. 0:16:54.360 --> 0:16:57.160 It was just generally considered to be a foregone conclusion, 0:16:57.520 --> 0:17:00.040 and so numerous businesses popped up with the intent of 0:17:00.120 --> 0:17:02.800 cashing in on the tech craze in general and the 0:17:02.840 --> 0:17:07.679 Internet in particular. Investors went bonkers, and so did some 0:17:07.800 --> 0:17:09.720 of these companies, as many of them would go on 0:17:09.800 --> 0:17:14.000 to lavishly spend money on stuff like office spaces and 0:17:14.080 --> 0:17:18.280 perks without you know, figuring out how the company would 0:17:18.320 --> 0:17:23.639 ultimately make money. And this trend proved to be unsustainable 0:17:23.680 --> 0:17:26.920 in the long run. How's that for a cliffhanger. I mean, 0:17:26.960 --> 0:17:30.600 you guys know what's coming, but still, you know, structurally 0:17:31.119 --> 0:17:35.159 it's a cliffhanger. Anyway, we'll be back to talk about 0:17:35.160 --> 0:17:46.840 it after this quick break. Eventually, due to a few 0:17:46.920 --> 0:17:50.760 different factors, reality would pull the rug out from under 0:17:50.840 --> 0:17:54.840 the tech sector and we got the great Dot Com crash. 0:17:55.680 --> 0:17:58.760 Companies that have been valued in the millions of dollars 0:17:59.119 --> 0:18:03.280 lost all their value rapidly as investors lost confidence, and 0:18:03.400 --> 0:18:06.440 ventures that just showed no signs of having a business 0:18:06.480 --> 0:18:09.719 plan or means of making revenue. But it wasn't just 0:18:09.880 --> 0:18:13.320 the questionable companies that suffered the effects of the crash 0:18:13.720 --> 0:18:17.719 rippled through the tech sector, hitting more established companies with actual, 0:18:17.840 --> 0:18:22.760 solid business plans, and that included ARM. Now part of 0:18:22.760 --> 0:18:26.040 the issue was that ARM itself was seeing its value 0:18:26.119 --> 0:18:31.280 inflated well beyond the company's own earnings. In before the 0:18:31.359 --> 0:18:33.920 dot com crash, the value of the company was more 0:18:33.960 --> 0:18:37.960 than three hundred times what it was earning. But even 0:18:38.000 --> 0:18:40.719 after the crash, ARM was still hitting targets, but its 0:18:40.800 --> 0:18:45.160 value was plummeting because the market was readjusting. The world 0:18:45.240 --> 0:18:48.159 went into a recession, and that recession affected ARM just 0:18:48.240 --> 0:18:51.480 as it was affecting other companies, and it was, according 0:18:51.520 --> 0:18:55.760 to some employees at the time, a pretty rough time 0:18:55.800 --> 0:18:58.200 to be working for ARM. To be fair, it was 0:18:58.200 --> 0:19:01.800 a pretty tough time to be working just about everybody 0:19:01.840 --> 0:19:05.960 around this time. Robin Saxby, the CEO who had sort 0:19:06.000 --> 0:19:10.040 of guided ARM into the intellectual property phase of its existence, 0:19:10.359 --> 0:19:14.879 transitioned into the role of chairman of ARM, and David 0:19:15.040 --> 0:19:19.600 Warren Arthur East, better known as just Warren East, came 0:19:19.680 --> 0:19:22.840 in as the new CEO of the company. East had 0:19:22.880 --> 0:19:28.040 previously worked at Texas Instruments until n And joined ARM 0:19:28.119 --> 0:19:32.160 and created a consulting business division within ARM. He then 0:19:32.280 --> 0:19:34.920 rose to the rank of vice president of Business Operations. 0:19:35.400 --> 0:19:39.800 Then later on he became the chief operating Officer or CEO, 0:19:40.040 --> 0:19:42.520 and then he became the CEO, and he would lead 0:19:42.520 --> 0:19:45.600 the company for more than a decade, which included guiding 0:19:45.720 --> 0:19:48.000 ARM not just out of the two thousand one recession, 0:19:48.400 --> 0:19:51.880 but then a subsequent recession in two thousand nine. These 0:19:51.880 --> 0:19:56.280 were enormous recessions. It wasn't like ARM did something wrong. 0:19:56.400 --> 0:20:01.200 It was more that these were global events. And while 0:20:01.320 --> 0:20:04.440 he would move on from ARM in two thousand thirteen, 0:20:04.560 --> 0:20:07.240 I wouldn't worry so much about him because now he's 0:20:07.240 --> 0:20:11.240 the CEO of Rolls Royce Holdings. So one of the 0:20:11.240 --> 0:20:14.720 things East did in order to try and and re 0:20:14.960 --> 0:20:17.560 center the company after the dot com crash was to 0:20:17.600 --> 0:20:20.560 create five year road maps with a plan on how 0:20:20.600 --> 0:20:23.760 to guide the business beyond just the short term, you know, 0:20:23.800 --> 0:20:27.399 more of a long term look at production and and 0:20:27.480 --> 0:20:32.120 research and development and business plans, which I think is phenomenal. Meanwhile, 0:20:32.800 --> 0:20:35.639 the business was beginning to change again because in the 0:20:35.640 --> 0:20:38.240 early days, a big part of The work around designing 0:20:38.280 --> 0:20:42.880 processors was getting the individual components to a smaller size 0:20:42.960 --> 0:20:45.320 so that you could fit more of them on a 0:20:45.400 --> 0:20:49.399 single chip without making the chip any bigger. Manajorization was 0:20:49.440 --> 0:20:51.560 really the goal, while at the same time you had 0:20:51.600 --> 0:20:54.080 to keep an eye on stuff like heat management, because 0:20:54.119 --> 0:20:58.200 packing more components closer together usually means that a powered 0:20:58.240 --> 0:21:01.600 processor is going to generate more heat, and heat and 0:21:01.640 --> 0:21:06.080 electronics are not a great pair. And at a certain point, 0:21:06.240 --> 0:21:10.000 the microprocessors were reaching a stage where the need to 0:21:10.000 --> 0:21:13.680 make them more powerful was starting to diminish, at least 0:21:13.720 --> 0:21:16.679 for the time. The microprocessors could be shrunk down to 0:21:16.720 --> 0:21:20.320 a smaller form factor because the components were smaller, but 0:21:20.400 --> 0:21:24.200 there wasn't a need to keep the next generation processors 0:21:24.240 --> 0:21:28.200 at the same size to cram more of them on there. 0:21:28.200 --> 0:21:30.359 In other words, you didn't need to keep the chip 0:21:30.480 --> 0:21:34.520 itself the same size. If your chip measured a centimeter 0:21:34.680 --> 0:21:38.640 square and you were able to shrink down the components, 0:21:39.040 --> 0:21:41.840 but you didn't need to add more components, you could 0:21:41.840 --> 0:21:45.399 make that square smaller. Maybe it's you know, nine tenths 0:21:45.400 --> 0:21:49.399 of a centimeter instead point nine centimeters. Well, companies began 0:21:49.440 --> 0:21:53.159 to take the opportunity to build out software based approaches 0:21:53.240 --> 0:21:56.199 to chips and build out what are called system on 0:21:56.240 --> 0:22:00.920 a chip or s OC solutions, and that really made 0:22:00.920 --> 0:22:05.119 it useful to have these very very tiny microprocessors as 0:22:05.200 --> 0:22:07.320 part of it. A system on a chip, by the way, 0:22:07.400 --> 0:22:11.520 is an integrated circuit that creates an entire system on 0:22:11.600 --> 0:22:14.679 a single chip, and that includes stuff like a central 0:22:14.720 --> 0:22:19.480 processing unit UH, internal memory, input and output ports, and more. 0:22:19.920 --> 0:22:23.320 The system on a chip approach cuts down energy consumption, 0:22:23.600 --> 0:22:26.840 it can add more capabilities to smaller devices that don't 0:22:26.920 --> 0:22:30.760 have the physical space to fit multiple chips. So the 0:22:30.800 --> 0:22:33.399 system on a chip approach is what would ultimately allow 0:22:33.480 --> 0:22:36.800 for the evolution of something like smartphones and other small 0:22:36.920 --> 0:22:41.080 computational devices. Many companies interested in developing system on a 0:22:41.160 --> 0:22:45.720 chip systems for various devices from communication systems to AI 0:22:45.800 --> 0:22:49.639 platforms look to ARM processors to provide the low power, 0:22:49.960 --> 0:22:55.160 high efficiency processing part of that system. So again, it 0:22:55.240 --> 0:22:57.879 wasn't that ARM was creating systems on a chip, but 0:22:58.040 --> 0:23:03.639 rather the designs of ARMS micro architecture was an important 0:23:03.760 --> 0:23:06.040 part of the system on a chip that was made 0:23:06.040 --> 0:23:09.080 by these other companies. ARM Holdings have been releasing new 0:23:09.119 --> 0:23:14.240 micro architectures pretty frequently leading up to the dot com crash, 0:23:14.320 --> 0:23:18.960 But after ARMED ten, which released in late it would 0:23:19.000 --> 0:23:22.680 be nearly four years before the next generation of processor 0:23:22.760 --> 0:23:26.040 designs would emerge from ARM. ARM eleven would launch in 0:23:26.160 --> 0:23:30.600 April two thousand two. This family of processors didn't come 0:23:30.600 --> 0:23:33.919 out of the Cambridge office. This one actually originated in 0:23:34.600 --> 0:23:39.920 a French office at Sophia Antipolis, where ARM had expanded. 0:23:40.320 --> 0:23:42.800 At this stage, a lot of the optimization was centering 0:23:42.800 --> 0:23:47.119 around playing media files efficiently as MP three players. You know, 0:23:47.200 --> 0:23:52.320 most notably the iPod were becoming popular. These devices needed small, 0:23:52.520 --> 0:23:57.000 powerful and efficient processors to avoid problems like overheating or delays. 0:23:57.600 --> 0:23:59.960 You know, no one wants a super hot electronic gadget 0:24:00.000 --> 0:24:02.760 burning a hole in their pocket or have the experience 0:24:02.760 --> 0:24:06.280 of navigating a menu in an iPod or similar device 0:24:06.280 --> 0:24:08.440 and then having to wait for the device to catch up. 0:24:09.280 --> 0:24:12.120 In two thousand five, ARM would make a really big change. 0:24:12.160 --> 0:24:15.040 The company recognized that in order to continue to grow 0:24:15.480 --> 0:24:19.840 and to succeed, it needed to diversify beyond just making 0:24:19.920 --> 0:24:24.000 smaller and more powerful processors. The ARM eleven family would 0:24:24.040 --> 0:24:27.879 be the last group of ARM microprocessors using that particular 0:24:28.440 --> 0:24:33.280 simple numbering system. The company changed gears and introduced a 0:24:33.320 --> 0:24:38.160 new classification, several new classifications of processors under different lines 0:24:38.200 --> 0:24:42.800 called Cortex. So you had Cortex A, and Cortex A 0:24:43.160 --> 0:24:47.200 was a continuation of the chain that had left off 0:24:47.240 --> 0:24:49.200 with ARM eleven. So you could think of the first 0:24:49.240 --> 0:24:53.560 generation of Cortex A almost like it's ARMED twelve, because 0:24:53.560 --> 0:24:55.960 that's it was a continuation of that process. But there 0:24:56.000 --> 0:24:59.439 was also Cortex ARE. These were processors that ARM had 0:24:59.480 --> 0:25:02.919 optimized for real time applications, so in other words, for 0:25:02.960 --> 0:25:07.680 applications that needed super fast responsiveness and really low latency 0:25:07.720 --> 0:25:11.960 but not necessarily super powerful processing power. Then there was 0:25:12.119 --> 0:25:15.760 Cortex M still is actually there is Cortex M. That's 0:25:15.760 --> 0:25:18.800 a family of processors that use very little power and 0:25:18.840 --> 0:25:23.880 are inexpensive, and those are mostly meant for embedded technologies. 0:25:24.200 --> 0:25:26.560 And this was right around two thousand five, and to me, 0:25:26.960 --> 0:25:31.359