Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/the-zombie-fabs-of-ti.17135/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

The Zombie Fabs of TI

benb

Well-known member
BCG (reference: https://www.bcg.com/publications/2022/tracking-next-phase-automotive-semiconductor-shortage) put out a slide deck that suggested, in my words, that zombie fabs are the cause of automotive chip shortage, in analog and MEMS. The main analog players are TI and ADI, and it wasn’t immediately obvious that TI or ADI have a zombie fab problem. Only two of TI’s current fabs, DMOS6 and RFAB1, are non-zombie; the rest are zombies.

KBh7TTm5NAnw_OCy5sRMXZ8VmEx5GoVyGNQivlcvMtXvLwQyBFnyKV2-PUD7slfZfR30X_R2udSniZpEFODmsmJUsTtPKAghf0EteAtZjOcjvZAyjX9LScSUU2snIzETyCLC0NDJr1WwhPh3dTTSIMp5MWCOnWaOZrd9LZ-UYc0DaXL2xK-DN__mfvgQ0A


https://www.ti.com/about-ti/company/ti-at-a-glance/manufacturing.html

Zombie fab: A fab that first started operating in the 1980s or 1990s, that has 200mm or older equipment. 200mm equipment is not available for sale anymore, so these fabs cannot be expanded, and tool repairs require cannibalization, which slowly reduces the number of functioning tools available. The term zombie refers to an outlier situation in which a fab continues operating with extremely old technology, with steadily diminishing capacity, typically producing a few parts with a long life at a rock bottom price.

TI non-zombie fabs: DMOS6 (Dallas), RFAB(Richardson).
TI non-zombie fabs coming online: LFAB (Utah), RFAB2 (Richardson).

TI is in a transition phase currently. Two new fabs are coming online in 2022 and early 2023. The 150mm fab in Sherman will be shut down in 2024-2025 (about the same time the new fab in Sherman comes online). This suggests other zombie fabs will similarly stop operating as the newer, more sustainable and efficient RFAB2, LFAB and eventually SFAB come online.
Reference: https://techxplore.com/news/2022-07-texas-instruments-phasing-chip-modernizing.html


w0LzuHzjY1BpTEEv8n1HgshXYCaMlLeyH7DoO9jo6XStHkf6YP0-TCx7D1GLnhW7f3WTuYdH46alvGHafq6NjS5bWYUagbcFZE0_kA2ICKXc9b3r0h9A-4b8bmJE9E5L5plqB5xt1X-zU77cDavPb2PaPXrPntEfO9OUxZMOqM7_dmtOqJId8IR7s56tHg

TI zombie fabs: Dallas (200mm), Sherman (150mm), So. Portland, Freising, Chengdu, Aizu, Miho. Will some of these sites continue on, perhaps gaining strength as new cannibalization opportunities occur from the fab shutdown in Sherman? Or will the efficiency of the new 300mm fabs make these sites uneconomic to continue operating?
 
Last edited:
Microchip's fabs might be considered "zombies" at first glance, since all use 200mm wafers. (or smaller? not sure) I'm puzzled hearing that 200mm equipment is not available for sale... sounds believable, but then I have this data point:

Microchip's Fab 2 (Tempe, Arizona) and Fab 4 (Gresham, Oregon) has been expanding; this year's 10K filing doesn't give any numbers, but mentions:

Fab 2 currently produces 8-inch wafers and supports various manufacturing process technologies, but predominantly utilizes our 0.25 microns to 1.0 microns processes. During fiscal 2022, we increased Fab 2's capacity to support more advanced technologies by making process improvements, upgrading existing equipment, and adding equipment.

Fab 4 currently produces 8-inch wafers using predominantly 0.13 microns to 0.5 microns manufacturing processes. During fiscal 2022, we increased Fab 4's capacity to support more advanced technologies by making process improvements, upgrading existing equipment, and adding equipment. A significant amount of additional clean room capacity in Fab 4 is being brought on line to support incremental wafer fabrication capacity needs.

Fab 5 currently manufactures discrete and specialty products in addition to a lower volume of a diversified set of standard products.

We believe the combined capacity of Fab 2, Fab 4, and Fab 5 will allow us to respond to future demand of internally fabricated products with incremental capital expenditures.

As a result of our acquisition of Microsemi, we acquired several smaller wafer fabrication facilities, which utilize older technologies that are appropriate for the discrete products they manufacture. We currently plan to continue to operate these fabrication facilities with modest investment to keep them operational with the exception of the facility in Santa Clara, California, which we closed in fiscal 2022.

We continue to transition products to more advanced process technologies to reduce future manufacturing costs. We believe that our ability to successfully transition to more advanced process technologies is important for us to remain competitive.

We augment our internal manufacturing capabilities by outsourcing a significant portion of our wafer production requirements to third-party wafer foundries. As a result of our acquisitions, we have become more reliant on outside wafer foundries for our wafer fabrication requirements. In fiscal 2022, approximately 60% of our sales came from products that were produced at outside wafer foundries.


...but this part of the "zombie" definition doesn't seem to apply to the 40% of MCHP sales produced internally, except for the "long life" issue.

with steadily diminishing capacity, typically producing a few parts with a long life at a rock bottom price.

- steadily diminishing capacity.... no.

- producing a few parts.... no.

- rock bottom price... well I suppose some are less expensive than other sources....

Are you sure these apply to TI's 200mm fabs, or even 150mm?

Maybe TI will sell some of the older fabs to someone lower on the totem pole. (see for example ATREG press releases like the ON Semi fab sales to Diodes Inc and LA Semiconductor) Trickle-down economics in action?

(Disclaimer: I work for MCHP, but my statements on this forum are my own and do not represent my employer, etc. etc. blah blah blah.)
 
Last edited:
Microchip's fabs might be considered "zombies" at first glance, since all use 200mm wafers. (or smaller? not sure) I'm puzzled hearing that 200mm equipment is not available for sale... sounds believable, but then I have this data point:

Microchip's Fab 2 (Tempe, Arizona) and Fab 4 (Gresham, Oregon) has been expanding; this year's 10K filing doesn't give any numbers, but mentions:




...but this part of the "zombie" definition doesn't seem to apply to the 40% of MCHP sales produced internally, except for the "long life" issue.



- steadily diminishing capacity.... no.

- producing a few parts.... no.

- rock bottom price... well I suppose some are less expensive than other sources....

Are you sure these apply to TI's 200mm fabs, or even 150mm?

Maybe TI will sell some of the older fabs to someone lower on the totem pole. (see for example ATREG press releases like the ON Semi fab sales to Diodes Inc and LA Semiconductor) Trickle-down economics in action?

(Disclaimer: I work for MCHP, but my statements on this forum are my own and do not represent my employer, etc. etc. blah blah blah.)

I still see us sending product to TI Clark and Freising, not a lot but some

Microchip solely using GF now or still use Silterra for some products
 
JMS, why not use your fabs for interposers? Even a 2u process is perfect for that. Bump pads are typically at 130--110um nowadays. HBM3 memory is at a exotic 40um pitch. Skywater seems to be focused on 130-90nm interposers, which is probably overkill (any opinions out there?). Can you guys add TSVs (through silicon vias)?
Amkor cannot handle packaging of die below 90um. I called them 2 months ago. Can you guys flipchip ASICs with a
55um pitch and HBMs with a 40um pitch onto your 1u process? If yes, you have a winner, even if we need to wirebond from there (TSVs would be a bonus).
 
(Disclaimer: I work for MCHP, but my statements on this forum are my own and do not represent my employer, etc. etc. blah blah blah.)
I might have to steal that :D
In all seriousness though I wonder how they even get tools and parts to expand their 200mm fabs. At my firm we can’t even get spares for some of the early 300mm dry etchers we have laying around. And I don’t think the vendor will even help us if we ask about tool issues on those tools anymore. Which is a real shame because from what I’m told those tools are falling apart.
 
I'll have to look at MCHP in more detail but in general, yes, it's a zombie operation that is at the end of useful life.

"making process improvements, upgrading existing equipment, and adding equipment."

There are extreme, uneconomic ways of adding equipment to 200mm fabs, which illustrates why no 200mm fab is ever expanded: You have to buy 300mm equipment and then modify the wafer handler robots and factory interface for 200mm wafers, for a start. The costs are outlandish and the level of work is outlandish; several times what it costs just to buy stock 300mm tools. Everything is custom, nothing is off the shelf.

I explored the outlandish steps in 2008 at Qimonda. The word "upgrade" and the word "add" can be stretched to the breaking point to sound optimistic; purchasing a commercial off the shelf valve could be considered an upgrade, while "add" equipment could mean moving a piece of equipment that was stored for parts back into the cleanroom.
 
Back
Top