• Home
  • :
  • Community
  • :
  • Blogs
  • :
  • Breakfast Bytes
  • :
  • GLOBALFOUNDRIES After the Pivot

Breakfast Bytes Blogs

Paul McLellan
Paul McLellan
16 Jul 2019
Subscriptions

Get email delivery of the Cadence blog featured here

  • All Blog Categories
  • Breakfast Bytes
  • Cadence Academic Network
  • Cadence on the Beat
  • Cadence Support
  • Custom IC Design
  • 定制IC芯片设计
  • Digital Implementation
  • Functional Verification
  • IC Packaging and SiP Design
  • The India Circuit
  • Insights on Culture
  • Mixed-Signal Design
  • PCB Design
  • PCB、IC封装:设计与仿真分析
  • RF Design
  • Signal and Power Integrity (PCB/IC Packaging)
  • Silicon Signoff
  • Spotlight Taiwan
  • System Design and Verification
  • Tensilica, Design, and Verification IP
  • Whiteboard Wednesdays
  • Archive
    • Industry Insights
    • Logic Design
    • Low Power
    • The Design Chronicles
Paul McLellan
Paul McLellan
16 Jul 2019

GLOBALFOUNDRIES After the Pivot

 breakfast bytes logo At SEMICON West I sat down with Gary Patton, CTO of GLOBALFOUNDRIES, to get an update on what is going on since the “pivot”.

The Pivot

The pivot was the decision last year to pull back from 7nm and EUV development and focus their process development investment on less advanced nodes. The reasoning was that the leading-edge nodes represent a decreasing percentage of the semiconductor industry, about 20% today. Also, the economics of 7nm were not that compelling. Since GF didn’t plan to put in place large 7nm capacity, the business was always going to be “challenging” (Gary’s word).

GF is now focused on specific markets where their technology portfolio is a good match. In particular, they have a strong RF position with both FD-SOI and the existing IBM RF business that they acquired. Gary pointed out that their process portfolio is a match for the requirements of 5G, both in the sub-6GHz band and mmWave, and for both basestation and handset.

GF have rationalized their business portfolio since the pivot. They sold their small fab campus in Singapore to Vanguard and moved their analog-mixed-signal business onto the main campus. They are selling their East Fishkill fab in New York to ON Semiconductor. They sold their ASIC business to Marvell. Their fabs are now:

  • Malta (in New York): This was the fab planned for 7nm and beyond. Also called Fab 8.
  • Dresden (in Germany): This is the old AMD fab that GF inherited when they were spun out. It is Europe's largest 300mm fab. It is their lead fab for FDX, also called Fab 1.
  • Singapore (in…err…Singapore): There is one 300mm fab (I think it’s called Fab 7) and several 200mm specialist fabs (2, 3, 5).

22FDX

The main focus today is 22FDX, which is their name for their 22nm FD-SOI process. This is a 22nm version of the 28nm FD-SOI process that they licensed from ST Microelectronics, who developed the basic process as an alternative to FinFET. Both FD-SOI and FinFET have better control of the channel, so lower leakage. FinFET does it by making the channel a fin and wrapping the gate around it. FD-SOI does it by backing the channel with an insulator to cut off the sneak paths and ensure that the whole channel is physically close to the gate and so well controlled.

They have added MRAM to this, which Gary told me will be fully qualified for Dresden in Q4 (meaning production-ready). It is currently what they call M5 qualified, which means a sort of risk production level. They already have over $500M of MRAM design wins. They have eFlash for 40nm and above, so this means they have a non-volatile memory at all nodes. One notable feature of their MRAM is that it can go through solder reflow without a problem.

 Since FD-SOI is built on an insulating substrate, it is possible to add RF and so build single-chip solutions including the power amplifier (provided it doesn’t require too much power). It is perfect for narrowband IoT (NB-IoT), for example, meaning that many IoT products can be built on a single chip, including all the wireless networking.

A big, fast-growing market is automotive. This requires special qualification for reliability (I’m sure you’ve heard of ISO 26262). They have automotive qualified in Dresden and Singapore already, and soon in Malta.

12FDX

I asked Gary about 12FDX, which was the next step on their FD-SOI roadmap. He told me they pulled back from that for a period since there is no point in being ready before their customers. But they have restarted the work. They are pushing towards 12FDX with both RF and MRAM. Basically, what they have on 22FDX, they should have on 12FDX. “We will be ready when our customers are,” he told me.

China Fab

Another thing I asked about was their fab in China. GF announced a joint-venture fab in China running 22FDX (and some other older processes) with a roadmap to go to 12FDX. The expectation was that a lot of China business would want local (in China) fabrication. But that never really materialized, and with the pivot they have plenty of capacity in their other fabs so it makes no sense to add more unneeded wafer starts. So that fab is on hold.

GF by the Numbers

Gary told me that they are at $6B in revenue. They have enough cleanroom space in their existing fabs that they could grow revenue by 40-50% without needing to break ground on a new building. In particular, there is a lot of space in Malta since it is not going to 7nm as originally planned. The result of all of this is that they now have, for the first time, free cash flow. Gary emphasized that he meant from operations, “no cheating and counting the transactions.”

 GTC 2019

Coming up on September 24 is GTC, the GLOBALFOUNDRIES Technology Conference, at the Santa Clara Marriott. So far there has just been a save-the-date announcement and there are not yet any more details. I will be there to cover it as always.

 

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.

Tags:
  • globalfound |
  • semicon |
  • 22fdx |
  • 12fdx |
  • FD-SOI |