【EE|Times全球联播】安森美收购GF晶圆厂、台积电7/6/5nm工艺线路图、特斯拉AI芯片和中国IC设计调查

DAVID FINCH: This is your EETimes Weekly Briefing. Today is Friday, April 26th, and among the top stories this week:

DAVID FINCH: 又到了EETimes全球联播时间，今天是4月26号，星期五。

本周新闻热点包括：

* ON Semi buys Globalfoundries’ ex IBM fab;

* 安森美收购格芯(GF)的前IBM晶圆厂;

* TSMC’s CMOS process node shrinks;

* 台积电的CMOS工艺节点继续缩小;

*and Tesla’s Kitchen Sink Approach to autonomous vehicles.

* 特斯拉在自动驾驶上使出浑身解数。

Later in the show, Echo Zhao will be here to discuss the 2019 China fabless survey results. And, at the bottom of the program, Bolaji Ojo, joins us to discuss what prompted him to start at EETimes a revealing new column, “My 35 years of journalism.”

接着，赵娟将讨论2019年中国IC设计调查结果。最后，Bolaji Ojo和我们一起讨论是什么促使他在EETimes开始一个新的专栏：“我的35年新闻生涯”。

All that to come, but first, Rick Merritt, EE Times’ Silicon Valley Bureau chief, explains why ON Semi had to buy Globalfoundries’ exIBM fab, and Globalfoundries’ next slimdown move.

所有这一切都会逐一呈献，但首先请EE Times驻硅谷记者站主任Rick Merritt解释为什么安森美要购买格芯(GF)的前IBM晶圆厂，以及格芯的下一步瘦身计划。

以下只有英文原文，Enjoy!

RICK MERRITT: Globalfoundries sold ON Semi its 300mm fab in East Fishkill, NY for $430 million in what seems like an unusual deal, but actually could be a trend of things to come.

For ON, this was an inexpensive way to make an important transition from its work making chips, largely discretes and power semiconductors... a whole range of things that ON makes in 200mm wafer fabs it owns, and it also sends a lot out to foundries for work in 300mm. So here it gets a fab up and running with some of the gear it needs, all the people it needs, and it's able to make that transition for maybe a third of the cost of building a fab from the ground up.

It need to do this because its rival Infineon is already well along on its transition to 300mm wafers. And as we understand it from talking to TSMC yesterday, it also has been transitioning its RFSOI chips off from 200mm to 300mm fabs.

So analysts tell us that this could be a trend because there's a lot of aging 300mm fabs out there that can’t readily be expanded for high-end digital production and could be great for analog folks. So look for more.

For Globalfoundries, this was just a continuation of their reorg announced last year when it said they were going to stop competing with TSMC and Samsung in leading-edge foundry processes, get rid of some assets, slim down and aim to be a profitable specialty semiconductor foundry. So they've still got some work to do. They've got one more sale that they haven't announced yet that they've got pending. And while they hope to be free cashflow positive this year, they say they've got a long way to go to be profitable because they got a lot of debt overhang.

DAVID FINCH: After Samsung’s 5 nanometer process announcement last week, it was TSMC’s turn this week to talk about its progress on the new process node. We asked Rick what this means to chip designers.

RICK MERRITT: TSMC gave an update on its roadmap at an annual Silicon Valley event last week. And what’s clear is that it's slicing and dicing new nodes, basically one new node a year-- 7, 7+, 6, 5, 5+. And it's stepwise progress. It's worthwhile shrinks, and not too bad in power consumption reductions, and the speed gains are pretty modest.

It’s a very similar strategy as we're seeing with its closest rival, Samsung. The good news is that they're both able to make some gains. They're both ramping extreme ultraviolet lithography steppers at a pretty good pace, so that technology is finally working after a 10-year development cycle.

For chip designers, it really means a couple of things. Analysts suggest you should really just focus on the long-lived major nodes-- which are 7 and 5-- and wait until they are mature. Maybe the first 100,000 wafers have gone through those processes, so they've been thoroughly debugged before jumping in. Of course, if you need the latest, greatest, bleeding-edge process at any given moment, there is the flavor of the year, and that might be one of those interim processes like a 7+ or a 6 or a 5+. So you take what you can get.

Perhaps more interesting is, TSMC was able to give more details on a couple of its next-generation packaging technologies. And different kinds of 2.5D and 3D stacks are becoming increasingly important to deliver gains beyond what CMOS scaling can do these days. So there's a couple of options coming up that won’t be here until 2021 in commercial products, but they are definitely worth looking at.

And then, the 3nm node. That's going to be a big change, because it's going to need a new transistor. And we didn’t hear much about that really. Some people suspect that’s where Samsung could pull into the lead and gain an advantage on TSMC, but right now the two are neck and neck. So we'll have to wait and see.

This is Rick Merritt in Silicon Valley for EETimes.

DAVID FINCH: Earlier this week, Tesla unveiled its full self-driving deep learning accelerator and new applications. Junko Yoshida, EE Times’ chief international correspondent, breaks down this apparent breakthrough.

JUNKO YOSHIDA: Tesla’s Autonomy Investor Day earlier this week was sort of a throwback to the 1990s. Remember? In those days, the launch of a new PC model was such a big deal? Every PC company raved endlessly about stuff like the latest Pentium processor, its product specs, new graphics cards and the cool PC games you can play on the new model.

Tesla just did that. Elon Musk and Director Peter Bannon detailed Tesla’s home-grown, deep learning accelerator, its architecture and its specifications, along with the new applications it will enable.

While listening to their presentation, I did an “I Love Lucy” double-take. Is this really a carmaker’s announcement? The future of a new car will be all about teraflops and wattage?

I’m OK with car OEMs eliminating the bit where a female model in a very little black dress, mounted on a rotating stage, lifts a giant sheet off a shiny new car, reinforcing the eternal link between gearshifts and sex. Good riddance to all that.

But remember, Tesla is literally rewriting the history of the autonomous industry by leading the EV and AV market. Forget about sex, but shouldn’t somebody at Tesla say something about safety?

Tesla’s two-chip module-- called Full Self-Driving computer-- delivers a total of 144 trillion operations per second (in other words, TOPS), drawing 72 watts. Great. Everybody's impressed. But I can’t stop wondering how safe and secure that full self-driving computer can actually make a Tesla on the deadly Nimitz Freeway, or the even deadlier Jersey Pike.

Speaking of Full Self-Driving--or “FSD” as Elon calls it-- we are looking at another unfortunate case of Tesla’s marketing deception. This-- just like Tesla’s Autopilot-- is a blatant misnomer in my opinion.

Let’s make it clear: FSD does not match the auto industry's definition of Level 4 or Level 5 autonomous vehicles. It does not drive itself. Somebody else has to do it. A person. FSD is the equivalent of Level 2-plus at best.

Elon also talked about Tesla’s Robo-taxi plan, promising the launch of this service by the end of 2020.

Hold your horses, Elon. Here’s another case of Tesla’s redefining terms to confuse the issue. Tesla is not proposing robo-taxi services as we’ve come to know them.

Under Tesla’s plan, you buy Tesla. Then you can lend it back to Tesla, to earn some extra cash. More specifically, a new app Tesla has developed will allow you to share your car with others for a fee while Tesla earns a commission of 25-30%. 

Stated accurately, Tesla’s proposed plan is a ride share service, not a fleet of robo-taxis.

Tesla has taken vast liberties in naming its features and applications, tailoring them to the company’s own marketing needs.

But why? Given that so many owners love their Tesla because it’s a good car, why does Tesla keep presenting new features and applications not by what they do, but by what they’re not?

A little bit of honesty, not bravado, can go a long way toward building public trust in an autonomous vehicle future, which to most people still doesn’t look very believable.

This is Junko Yoshida, chief international correspondent for EETimes.

DAVID FINCH: Recently, EE Times China published the 2019 China Fabless Report based on our own regional survey.

I'm joined now by Echo Zhao, chief analyst at EE Times China, to discuss the growing number of Chinese fabless companies.

Echo, welcome to the program.

In reviewing this report, about a third of the survey respondents indicated that they expect their company’s sales to increase by more than 20%. And this figure is higher than the growth projected by analysts as well as the China Semiconductor Association. I wonder: Why do so many respondents expect such robust growth, despite a slowdown in the forecasted growth rate for China’s chip industry as a whole?

ECHO ZHAO: I had the same question. When I initially got the results, I wondered: Why do they have such strong confidence? Are they too optimistic?

But when you see all these respondents’ profiles, most of them are small and mid-sized companies, and I think their results have greater variability. Certainly, the growth rate of the big companies is more strongly influenced by the market environment. But for small companies, the growth rate is more related to their market strategies, directions and opportunities.

Another interesting finding is that the profits of some respondents are higher than the Top 10 fabless in China.   

DAVID FINSH: Moving on to product promotion, the focus for 2019 seems to mirror what’s happening in the rest of the world, where we're seeing a shift from consumer products to things like automotive, IoT and artificial intelligence. Are you seeing changes in 2019 that would be maybe more specific to the market trends in China?

ECHO ZHAO: I’m not sure what products would be more specific to China in 2019, but I assume it would be IoT or internet-related products. Like what happened in bicycle-sharing. Although now we know it’s not a successful business, but it was so hot the last two years, and it did bring large demands of NB-IoT.

DAVID FINCH: You mentioned that other applications that are hot in China at the moment include intelligent door locks, power-bank sharing, and...

ECHO ZHAO: And even mini-speakers due to Dama square dance culture. These applications are not that “high tech,” but all have a broad market space and also effect IC supply chain. But I'm not sure if it is unique to China. All that happened in US, too?

DAVID FINCH: I see. Now, many of the respondents in this report indicated that they have fewer than 40 design engineers and planned to release fewer than four products last year. Do you think that the size of these responding companies poses a concern to the Chinese IC industry?

ECHO ZHAO: Yes, I do. And most people have the same thought and call it “savage growth.” In 2018, there are totally 1,698 fabless chip companies in China, and there were only 736 fabless in 2015.

And last year, nearly half of the of these fabless companies had revenue that was less than 10 million RMB.

DAVID FINCH: I see.

ECHO ZHAO: The industry is worried that too many startups are blocking the concentration of the IC design industry. Of course, some people would say that the IoT market is a fragmented market that needs many startup’s services. But personally I believe that “fragmentation” is more focused on niche solutions, not on IC products. So a large number of startups that do not have the ability to provide strong overall solutions are doomed to be eliminated from the market or acquired by big players.

DAVID FINCH: Echo, thank you for joining me.

ECHO ZHAO: Thank you. It's my pleasure.

DAVID FINCH: Bolaji Ojo has launched a new column in which he tells the stories and shares lessons he’s learned over 35 years in journalism on three continents. As much as EETimes wants to learn more about your career as an engineer, we want to share with you, through Bolaji's observations, the people behind EETimes.

We think it's a good idea to explain what it takes for any reporter, regardless of his or her beat, to deliver the news to you. Getting to know what life is like on the other side of the news can only help us to understand one another a little better.

Here's Bolaji.

BOLAJI OJO: What motivated me to start this column? I'm not even sure that I know what motivates anyone of us to do anything. Maybe a series of factors. In my case: first, I'm older; second, my dad died last year; and I... You know, he was such a storehouse of knowledge and information about our culture, about his field, and all that. And all of that just died with him.

You know how much information those of us who are professionals, we long experience in our field are carrying around. All of us are carrying around this mammoth load of information, and we're not passing that on to the next generation. I think that was, for me, probably one of the first things that I began to think about. And I said: You know what? I did make a promise with myself years ago that I would write something about my experience. So it was the right time to get started.

Journalism for me is about life. It's about bringing together all of the components of life-- you know, the sights and the sounds-- and then bringing it to an audience that really thinks: You know what? This is valuable.

Now, imagine you are the person who gets to witness, who gets to record, who gets to write it out, and then gets to tell the world. You know, you're like that person who brings together all of these different strands of life, and then you weave something marvelous out of it. That's what journalism has been for me. Also, I visit this country, that country. I look at the people, I look at the places, I look at the culture, the food. When it's about companies, I look at the corporate experience, you know, the financial moves, the strategic management moves. All of those little, little things, they're supposed to come to a particular point. That point is what I want to narrate. That point is what I want to bring to my audience so that, even without leaving their homes, they can see and they can hear, they can comprehend, understand and appreciate what's happening all around them.

In the rest of this series, I'm going to be talking about some of the challenges, the difficulties that journalists face. But, you know, I think for this particular series of EETimes On Air, one thing that I would like to talk about is very, very simply this: that the biggest challenge that any journalist faces is the fact that whatever story you are trying tell, there is somebody somewhere that's also trying as hard as they can either to make sure you don't get to tell that story, or that you tell a different version of the story, one that's favorable to them.

So you have this tug of war that's always happening with regards to journalism. You always have this back and forth, this tozu, you know. The story that a journalist wants to tell is not the story that everybody wants you to tell. And so you are caught in between different worlds and your own ideas.

And by the way, the journalist himself or herself is not even lost in this. Is not innocent in all of this. The journalist also may have a point of view. And the one biggest lie that journalists tell themselves is: I'm so completely independent. I'm not being influenced by anything. I think that's a lie.

The best way to get over that and to tell as accurate a story as possible is for you to understand and identify your own subjective ideas and notions and all of those things. If you can do that, you can tell the kind of story that at least, you know, a greater percentage of your readership will say: You know what? This person was fair. This person was balanced and accurate.

DAVID FINCH: That was Bolaji Ojo, and this has been your weekly briefing from EETimes and the AspenCore Global Service. You can read all of these stories and more at EETimes.com. Thanks for listening.

往期回顾

【EE|Times全球联播】三星5nm工艺、卡内基梅隆大学教授谈自动系统标准、后诺基亚时代的芬兰“无线谷”