Q&A with GloFo CEO: ‘IoT is no mystical animal’

Q&A with GloFo CEO: ‘IoT is no mystical animal’

Business news |
By eeNews Europe

Globalfoundries’ CEO Sanjay Jha came to Shanghai this week and said that mobility and pervasive computing will continue to drive the industry. But he emphasized that many applications expected to drive the industry’s growth – which includes mass market smartphones, M2M, IoT and automotive – "don’t require the cost and complexity of FinFET."

Lu JunChina's Big Fund manager(Photo: EE Times)
Lu Jun
China’s Big Fund manager
(Photo: EE Times)

Instead, Jha pitched FD-SOI and RF-SOI as "the right technology at the right time," in his keynote speech at the Shanghai FD-SOI Forum. The Forum audience was invitation-only, ranging from chip company executives who had traveled from Silicon Valley, Europe and Japan, to local fabless, research institutes and investors in China. Most notably, Lu Jun, a man who manages China’s Big Fund, was sitting in the front row.

As he queued up a slide illustrating the downward growth trajectory in smartphone shipments, Jha cautioned the audience: "Those days of 13 to 14 percent annual growth may be over."

(Source: Globalfoundries)

(Source: Globalfoundries)

In his opinion, what will drive the chip industry forward are "emerging markets – the next 2 billion subscribers, 5G, social, machine-to-machine interaction, and content consumption." The question, then, is what semiconductor companies can do about all this.

Jha said "cost, performance equivalent to today’s high-end smartphones and power consumption" are the three things semiconductors need to deliver. "I am not talking about sub-spec smartphones. I am talking about smartphones whose performance is equivalent to today’s high-end smartphones, which are, however, priced at less than $100 or $80. For that, you can’t afford FinFET," he added.

IOT ‘no mystical animal’
The Globalfoundries’ CEO also touched upon IoT. Although everyone in the industry today talks about the Internet of Things as if it’s a mystical animal, Jha stressed "it is not. But first, we need to define it."

He defined IoT as all devices "used in a sensorial environment, connected and share their sensor state with the Internet to optimize computing."

For IoT, "Ultra-low power consumption isn’t an incidental fact," said Jha. It’s the first and foremost requirement for semiconductors to make IoT happen.

Comparing IoT devices to a PC "which typically needs the power of 20kWh, running 3 to 4 hours" and a smartphone "that demands 2kWh, operating 24 hours," Jha said, "IoT requires energy of 200 milliwatt-hours and it must last 2 years." Equally important is its cost. "We are talking about an average selling price equivalent to $1."

So, why FD-SOI and RF-SOI are the right technologies at the right time?

Jha, in his keynote, unequivocally stated that Globalfoundries’ 22nm FDSOI platform provides "14-nm FinFET-like performance at 28nm equivalent gate cost." It offers ultra-low power, high performance "at 0.4V operating voltage." Designers can also use software-controlled transistor body-biasing for further optimization, he added.

Growing number of switches
Globalfoundries which completed the acquisition of IBM Microelectronics’ business in July is also banking on RF-SOI technology. RF-SOI was one of the strongholds of IBM’s chip businesses. 

Jha, who pitched RF-SOI as the best way to deliver cost-effective front-end modules, said, "Look no further than Apple’s new iPhone 6s."

As smartphones are designed to connect to multiple cellular modems, the number of bands a smartphone’s front-end module must support are growing exponentially. The iPhone 6s, for example, supports not 12, but "24 bands," Jha noted. More bands means more switches in the front-end module. The good news for Globalfoundries is that a majority of switches today are built on RF-SOI.

Jha said, "At a time when the average selling price (ASP) of an application processor is declining, the front-end module RF is the only ASP that’s increasing in a smartphone."

After the keynote, Jha sat down with EE Times to answer questions on Globalfoundries’ business including both FD-SOI and FinFET. Following are excerpts.

EE Times: Roughly, how big a percentage of Globalfoundries’ business will come from FD-SOI?

Jha: We haven’t gone that far to predict. But we already have multiple customers working on our 22nm FD-SOI platform, and we are expecting the first test chips out of Dresden in a couple of days. 

Our customers already have our process design kits (PDK). The data we took a look at last night matches with what’s offered in our simulated PDK. The tool set is ready, and our silicon-based PDK will be made available soon. 

Sanjay Jha, Globalfoundries CEO, in Shanghai

Sanjay Jha, Globalfoundries CEO, in Shanghai

EE Times: When will the first chip (based on Globalfoundries’ 22nm FD-SOI) become commercially available?

Jha: We are expecting the production tape-out to be in the second half of 2016.

EE Times: Are you counting on FD-SOI/RF-SOI to differentiate your foundry business from your competitors?

Jha: They are one of the differentiated factors. I would also like to remind you that TSMC and Globalfoundries are the only two foundries in the world offering an entire suite of process nodes ranging from 7nm to 180nm.

EE Times: And I suppose you also do FD-SOI, while TSMC doesn’t. But with all due respect, when it comes to leading-edge process nodes, it strikes us that Globalfoundries is always a quarter or two late ramping up, compared to your rivals. That doesn’t seem to help you win high-profile business from leading-edge customers like Qualcomm or Apple.

Jha: You’re right. We’ve been a quarter or two late on the leading-edge nodes. That’s one of the issues we’ve been trying to address. Today, we have one 14nm customer ramping to volume right now, followed by two others who will be ramping later this year and early next year. That’s not too bad. When it comes to 14nm volume production, I think we are roughly on par with our competitors.

EE Times: For 14nm node, Globalfoundries depended on the technology licensed from Samsung. What’s your plan for 10nm and 7nm? Will you develop your own technologies?

Jha: We’re developing our own technologies for the next nodes. The whole point of the IBM Microelectronics business acquisition is to leverage IBM’s technologists and technology to accelerate our own development of leading-edge process technologies.

EE Times: What do you see as your strength?

Jha: Intel offers lithography scale. As long as customers can pay for it, they can get that. In contrast, we offer our customers the best transistor design engineering – both in terms of performance and material.

EE Times: When do you think EUV will insert – 10nm or 7nm?

Jha: We’re not expecting EUV before 2018 or 2019. We’re focused on optical tools for 10nm and 7nm. As EUV stabilizes, we may use EUV for some layers. We’re also using EUV to accelerate prototyping.  

EE Times: How much of your business is coming from China today?

Jha: We are not so strong yet. Currently, we have five reasonable-size Chinese customers. We’re extremely keen on serving customers in China. We think FD-SOI will play a key role to trigger innovations among Chinese fabless companies. FD-SOI excels in such applications as mass market smartphones, power-consumption critical IoT devices and automotive. FD-SOI is inherently radiation hardened, making it ideal for automotive. Analog also benefits from FD-SOI.

FD-SOI Roadmap beyond 22nm

EE Times: FinFET promoters already have a roadmap for 10nm and 7nm. What’s the FD-SOI roadmap beyond 22nm?

Jha: We have an FD-SOI roadmap that also scales. We’re currently exploring with our partners if we should use a different tool set for the next node. If it turns out to be different, we’ll do the next-node FD-SOI production in Malta, NY, while Dresden remains focused on 22nm FDX.

EE Times: Have you made the decision?

Jha: No, we have not made the decision yet.

EE Times: Is the market for RF-SOI tight today?

Jha: It has been the case. Now that we bought IBM Microelectronics, we’re committed to investing to create more capacity. RF-SOI production will be no longer limited to our facility in Burlington, Vermont, but will be expanded to our Singapore operations. We’ll be doing both 200nm and 300nm RF-SOI production there. The technology transfer from Burlington to Singapore is happening now. 

About the author:
Junko Yoshida is Chief International Correspondent at EE Times

Related articles:
GlobalFoundries’ FD-SOI revolution
FD-SOI development as a service
Can GloFo and Europe’s chip firms unite?

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles