Limited functionality available
China’s advances in 5G and semiconductors give it an edge in machine vision and machine learning. Those technological advances are firing up the country's ambition of greater independence and control over its own destiny, say Deloitte’s Paul Lee and Chris Arkenberg.
“It’s kind of like deep-water drilling for oil. You can do it. There’s oil there. It’s just more and more expensive to do so. That’s another fundamental shift or limitation on the horizon of the entire industry: Essentially, the end of Moore’s Law.”
— Chris Arkenberg, research manager for technology, media, and telecommunications
Tanya Ott: I’m Tanya Ott and this is the Press Room, where we dig into some of today’s biggest issues in business.
Imagine this scenario: You’re late for work, but you need to stop by the store to pick up something—let’s say a snack for later that morning.
You wave your phone over a turnstile that reads a QR code and lets you in. There are cameras in the ceiling that track your movement. You walk to the snack aisle and there’s a weight sensor on the shelf, so, as soon as you pick up your snack, the combination of the sensor and the cameras confirms you’ve got it and then, as you walk out of the store, you’re automatically billed for your snack. No waiting in line to check out. Just in and out.
There are a few of these stores in the US right now—pilot projects. But in China, multiple companies are working on solutions like this, and they’re exporting the ideas around Southeast Asia. China has made great strides in developing its mobile and tech industries in recent years and joining me today to talk [about] it are Paul Lee, Deloitte’s head of global research for technology, media and telecommunications (also known as TMT), and Chris Arkenberg, a research manager for TMT.
We’re talking about China today and, specifically, we’re talking about connectivity and semiconductors. China currently has the largest digital user base in the world and you predict, Paul, that it will have world-leading telecommunications networks as well. Give us a sense of the scope of China in this space …
Paul Lee: I think everyone knows that China is a very large market with a very large population, but I think sometimes what people aren’t aware of is just how extensive its connectivity is, particularly for things like fiber. For fiber-to-the-premise connections, China has, as of now, over 330 million connections. So that’s about two-thirds of all the fiber connections in the world. And looking at mobile, there are about 1.2 billion 4G connections in China. Just as a general comment, when you’ve got advanced connectivity, it’s often a good foundation for disruption and for the creation of a lot of new applications.
Tanya Ott: China has a much larger population than most other countries in the world. India is slightly behind it, but I think China is at 1.27 or 1.28 billion people. Is the reach of their fiber and the reach of their mobile technology simply because of scale—they have so many people—or are they doing something maybe in a more sophisticated way or in a more advanced way than other people in the world?
Paul Lee: I think a lot of it comes down to the ambition and also the pace of execution. A few years ago, China made the decision to become world-leading when it comes to connectivity for both fixed and mobile. And the decision was taken to move, for example, the fixed network from predominately copper terminations into premises to fiber, and that’s been rolled out over the course of a few years. That’s an approach which is happening in lots of other countries, but it’s taken longer than has been the case with China. For 4G also, there was a concerted effort to deploy [it] very widely, quite extensively, and that’s happened over the course of a few years. So, the quality of the pace in terms of communications is largely down [to] the execution of the plan.
Tanya Ott: Was it just a matter of will, or are there regulatory or other conditions that made China able to more quickly deploy these kinds of technologies?
Paul Lee: One of the reasons is it’s a different class of economy and there is also the ability to have a national commitment and then to deliver upon it. If I compare the Chinese market to the UK market, one of the factors in the UK market is [that] when it comes to rolling out fiber-to-the-home, there are a lot more permissions that may need to be gathered and there’re different sets of regulations. There are about 25 different sets of regulations which vary between different markets within the UK if you want to roll out fiber. So, in one market, it may be that you have to adhere to a certain distance from a house or a certain time of working or certain permissions to close roads and then you move 20 kilometers elsewhere and the regulations vary. So, the bureaucracy load can be a lot higher in some other markets.
Tanya Ott: Fiber was critical for things like video-on-demand, and then layered on top of that, as you mentioned, we’ve got 4G mobile connectivity that’s allowed for things like live streaming from smartphones. China has the world’s largest 4G network right now, with more base stations and more subscribers than I think the rest of the world combined. And we’ve got 5G, which is on the horizon—that’s even faster. You say China is in a really good position to roll that out as well and sort of dominate this 5G space.
Paul Lee: China is certainly in a good position to migrate to 5G. One of the prerequisites for 5G generally is to have small cells. The smaller the cell, the faster the speed that you can get within a certain area. There are about two million base stations in China. When you look at a per-population or per-square-kilometer basis, that’s an unusually high degree of concentration. The migration from 4G to 5G is then relatively easy in not having to construct new base stations to be able to deliver that.
In terms of leading in 5G—when it comes to subscriber numbers—in five years’ time, China is likely to be at the forefront, if not leading. In terms of a portion of the population with 5G, it will be lower than lots of other countries simply because there are over a billion people to connect. It will have the largest user base most likely in five years’ time, [but] it won’t have the highest penetration of 5G. There’ll be other smaller markets where they can get to a higher penetration rate faster simply due to the fact that they’re smaller.
Tanya Ott: What kind of new business opportunities is 5G going to open up for China?
Paul Lee: One of the fascinating areas of change coming through [in] the next few years will be around machine learning. For machine learning, you need to have large data sets. You need to be able to gather lots of data to feed those algorithms and to tune them and then to deploy those. That’s where I think China’s got, perhaps, an advantage because there are lots of people … the data can leave quickly because the connectivity is great and then it can be deployed quickly because the connectivity is great. For something like machine vision, which will be a major factor in the next few years for business and for consumers, China has probably got a lead because of those kinds of organic capabilities and, also, the network capabilities that it’s got. Machine vision [and] machine learning—those are great components that work very well with fast connectivity.
Tanya Ott: You talk about the hardware that underpins all of this and one of the things that China has really made a priority is semiconductors. It’s something they’re very heavily invested in. Talk to us a little bit about what China is doing in the area of producing semiconductors and what that opens up for opportunities.
Chris Arkenberg: It’s a fascinating landscape because this is an effort that China has been attempting for a few decades now. They still lag a few generations on the most advanced processes, the smallest transistors, and the most dense semiconductors that some of the leaders can develop. But, as Paul alluded to earlier, this is embedded in this large push for China as a whole to advance their technological capabilities. That’s everything from the services that they can offer, the way that they can integrate that with economics and municipalities and every aspect of their society, all the way down to advancing their independence. A big part of this for China’s broader goal is to have greater independence, greater control over their own destiny—for lack of a better term—and, of course, now in the modern era, that destiny is, for all of us, very closely related to these dominant technologies. Semiconductors are kind of the base level of some ways. They’re right at the core.
Tanya Ott: China doesn’t want to be as dependent on foreign providers of semiconductor chips and so they’re really looking to ramping up their own production of it. What’s the risk there?
Chris Arkenberg: There are certainly challenges. I mean obviously, a lot of the headlines now show some of the geopolitical risks with that. It’s notable that China is the largest consumer of semiconductors. They only supply about 30 percent of their own demand, so they’re really a big driver for the global semiconductor industry in a lot of ways. If they meet more of their [own] demand, that obviously would impact the semiconductor marketplace. And then, maybe not so much a risk, but there’s a fundamental challenge which is just how difficult, how extraordinarily expensive it is to develop some of the largest industrial processes on the planet to deliver some of the tiniest components on the planet. Getting from 14 nanometers down to 7 nanometers—which is sort of the leading edge at the moment—is extraordinarily difficult. China still faces some challenges in those advanced processes.
Tanya Ott: My understanding is that they’ve got a goal of producing 70 percent of the semiconductors that they use, producing them in China by 2025. As you’re looking at predictions, what do you think their chances are of making that goal of 70 percent produced domestically?
Chris Arkenberg: Well, on the surface, it seems very ambitious. The Chinese government is very committed to this. They’re committing very large amounts of capital. And, as Paul noted, they have an economy that is much more aligned across different sectors. The government is very aligned with the leading manufacturers, with the financiers and investors, as well as China’s top digital platform companies. And they’re all marching on a similar agenda, although sometimes with different goals in mind. They do have the capacity to potentially move quickly on this.
One of the factors that introduces some interesting uncertainty and could possibly make it easier for them is that [in] semiconductors, in general, the architectures are shifting as more and more demand is being met from artificial intelligence and machine learning. These are reshaping the architectures of the semiconductor industry. That might rebalance the playing field in a way that could favor China’s development, which is to say they might not have to chase the most advanced manufacturers on the planet. They might actually be able to find successes through new architectures. There’s some really interesting stuff happening with older generation architectures that then get a software machine learning layer on top of them to optimize and get higher performance and lower energy usage. It’s a really fascinating time. The playing field is shifting a bit. And that may ultimately be in favor of China’s meeting its goals.
One thing I’d like to note for the Chinese semiconductor effort that I think is that their design capabilities are now on par with the top chip designers in the world. They can effectively design and spec semiconductors that are as good as any. They just tend to send them out of country to the largest fabricators that are able to develop these advanced processes. So, while their industrial processes for fabrication still lag, their designs now are globally competitive and, increasingly, that’s including their designs for AI and ML [machine learning]-specific chips as well.
Tanya Ott: And they’re looking then to up that fabrication game and bring that back within the country?
Chris Arkenberg: That does seem to be their stated goal. I guess it’s still notable that even if they don’t achieve that goal, soon enough they will have access to the best [fabricators] in the world. In fact, one of them is right across the strait from them in Taiwan. And so that won’t necessarily be a limitation to their ability to develop these chips. It just could take a while for them to be able to fabricate them, but they’ll still develop them and design them to their own needs.
As I understand it, their best processes right now are about 14 nanometers and that’s about three generations behind the leading edge, which is 7 nanometers. But it’s notable that this is running up against Moore’s Law. It’s getting harder and harder and harder to fabricate this stuff and more and more expensive because you’re getting so small that you’re starting to butt up against the limits of physics. You’re essentially getting near the quantum realm. In fact, one of the leading global foundries just signaled that it’s no longer going to pursue advanced processes or it’s going to limit its pursuit of these and look for ways to optimize existing processes. It’s kind of like deep-water drilling for oil. You can do it. There’s oil there. It’s just more and more expensive to do so. And so that’s another fundamental shift or limitation on the horizon of the entire industry: Essentially the end of Moore’s Law.
Tanya Ott: I would love to tie all this together with both of you—Chris and Paul—to talk about some of the technologies that come out of these advancements. And I’ve heard artificial intelligence. I’ve heard some machine learning. I’ve heard machine vision. Let’s just talk about some of the practical applications and maybe the business models or business opportunities that there are there. Machine vision—what is it, how is it being used right now, and what does the future potentially look like with machine vision?
Chris Arkenberg: All these things are essentially ways for software to understand the world without a lot of help from humans, who do it very naturally. Machine vision, in particular, is a way to train algorithms to understand how to recognize things in the world, and they do this bit by bit. There’s been a tremendous amount of advances in this field even just in the last five years or so, and China is now [capable] of the most advanced capabilities in machine vision. And, again, that gets you all sorts of ways for software to interrogate the world to understand what’s around us. It enables cars to navigate things. It enables cameras to recognize things. It enables things like facial recognition and behavior analysis and emotional awareness. It adds this richer layer to how software understands the world and then conversely the types of services that can be delivered to us.
Paul Lee: I’ll just give one recent example of machine vision in action. I’ve just been in the Mobile World Congress in Barcelona, which is not the biggest show out there but it’s over 100,000 people who are entering the halls every day. And what they need to do is secure entry—so checking that you have your pass and also checking that you have the right documentation. This year, the default way of entering was facial recognition. If you do the standard mechanical test using a human to do the assessment, it’s about 5 to 10 seconds per check. With the machine vision, it’s less than a second. You just pause, look at the camera, then walk through it and that’s verified. I don’t know the level of accuracy, but I talked to one of the companies working in that field and they were getting one in 50,000 false positives. That’s where somebody who shouldn’t be allowed in is allowed in. So that’s a degree of accuracy that you’re getting, and to Chris’s points—one thing which is really great about this technology is that it doesn’t tire, whereas humans do. And so, you got a consistency of operation which is great.
Tanya Ott: One of the more high-profile applications of machine learning over the last year, certainly for the average person that’s paying attention to the news, is something that China’s doing that’s called Social Credit. How does that work?
Chris Arkenberg: Essentially, what they are doing is building databases for their citizenry, for each member—and the databases, given the proliferation of digital systems and surveillance cameras and so on, are able to get all sorts of information about behaviors. Those are purchasing behaviors, those are things you might say on China’s social networks, things you might do in urban areas that are more surveilled than other areas—all of these things are essentially being documented per individual. Once you have that, all of that stuff is getting integrated into things like applying for a particular school for your child or trying to get a home loan or trying to travel overseas. China is essentially developing a system that evaluates these types of decisions based on previous behaviors.
Tanya Ott: Even behaviors like—did you run out traffic red lights, or, you know, are you jaywalking, things like that. It can be that granular.
Chris Arkenberg: Potentially. I don’t know the precise scale at the moment, but it is again an effort to tie behavior to opportunity, in a way. That tends to align with China’s collectiveness philosophy around the stability of its population and its country.
Paul Lee: Social Credit, at present, is being carried out by a few private companies and a few local authorities. It’s something which, at some time, may be deployed at the national level. But currently it’s experimental. It’s at the trial level. And one of the reasons for trying out something like Social Credit is simply because when you’re a financial company looking to understand credit worthiness, in the west, most people have pretty long credit histories. It might be working records, home records, the records of paying off credit card debts on time. For a lot of people in China, that data simply doesn’t exist. Now, offering people credit is an important way of making the economy work. It’s an essential lubricant. And what Social Credit is looking to do is provide a proxy for the records which don’t exist. When it comes to, for example, renting a car and assessing risk levels, then your behavior online is one of the key things to look at.
Tanya Ott: Well, thank you very much for your time. It’s an interesting part of the world to be watching these days and our listeners can get more on your predictions for China in your report online.
The guys: Thank you very much. Thanks a bunch, Tanya.
Tanya Ott: You can find Paul Lee and Chris Arkenberg’s 2019 predictions for China’s rollout of 5G and its expanding semiconductor business at deloitteinsights.com. That’s where we’ve gathered a whole bunch of predictions for technology, media and telecommunications—including smart speakers, eSports, and quantum computing.
Duncan Stewart: There is going to be a thing that occurs, maybe this year, maybe next, maybe never—but most likely in the next year or two—called quantum supremacy. When that happens, that’s going to mark the first time that a quantum computer is able to perform a specific calculation using a practical amount of time and resources better than a classical computer. And I make predictions for a living and I predict right now it’s going to be front page news and everybody’s going to be talking about it.
Again, that’s at deloitteinsights.com.
Thanks for listening to the Press Room. I’m Tanya Ott. Have a great day!
This podcast is provided by Deloitte and is intended to provide general information only. This podcast is not intended to constitute advice or services of any kind. For additional information about Deloitte, go to deloitte.com/about.