Intel's "can we afford fabs" moment

Just thinking… there was the moment where AMD decided to go “asset light” and ditch the fabs because they couldn’t afford to continue to run bleeding edge fabs. And obviously they made it work.

For all that Intel seems committed to continue to wrestle with fabs at this point, it’s horribly expensive and maybe their current pain can be seen as asking “Do we have deep enough pockets and big/steady enough demand and volume to go toe to toe with the foundries like Samsung, TSMC etc.?”

I’ve heard people say in the past that Intel should sell its fabs. Maybe they’re too big to divest but perhaps these financials and the coming years of austerity suggest that Intel has in the past hung in there too long.

I also suppose Intel is simply so big that no foundry/fab other than their own can handle the sheer volumes of parts they’d need… so in their shoes you can sell it and then turn around and be the new company’s biggest customer (and possibly still dissatisfied with the parts you can get) or you can hold on to it, operate it as Intel Foundry Services, and deal with whatever external customers you can attract to the processes you have. That whole process of divesting would probably just add moving parts and costs and waste time.

Anyone have thoughts on the possible parallels between AMD deciding to divest GloFo and the position Intel is in now?


I think Intel drank their own Kool-Aid. They lost their lead in their most important product line (CPUs) and took far too long to recognize that problem. They moved away from their sound engineering principles.

The problem with foundries is that in poor economic times, you’re paying too much to keep them afloat when you can’t sell all you can make. The problem with fabless is that in good economic times you are unlikely to make enough to fully take advantage of opportunities.

I don’t think Intel will stay behind AMD forever. Their R&D budget says they will catch AMD at some point. But bad economic times plus inferior products will make the next year or two likely the most painful in Intel’s history. Without all the information Gelsinger has on their reality, but also none of the pressures Intel faces from stockholders, their board, and the world, I think it’s time to close some fabs and take the downturn as a time to catch up to TSMC in process by upgrading those fabs. Yes, it will be very expensive not to get full amortization out of their existing machinery. But to me they are behind in process and in architecture, and process is the easier problem to manage. My assumption is that since Gelsinger became CEO they have got real architecture improvements in the pipeline, and despite their marketing fluff, those products won’t be arriving until 1-3 years off.

If there is real uncertainty about those products actually passing wherever AMD will be, then close some fabs (without upgrades) just to reduce costs will free up money for multiple parallel development teams. If Intel is not at least competitive in architecture, they are doomed. The problem with this approach is that it will cost a lot more to finally rebuild production once they have competitive products.

I don’t think Intel will benefit from spinning off their fabs. They have too many and they are behind TSMC in process. The logical move would be to sell them to TSMC, which will give TSMC even more non-Taiwan capability to survive. But I expect antitrust would not approve that spinoff.

As always in this industry, if you can accurately forecast demand for your products, you can make a lot of money. As time goes on, it looks more like we’ll manage a soft landing instead of a deep recession. But with the Fed split between larger and smaller interest hikes, forecasting accurately seems difficult.

I do not believe Intel will be able to maintain their dividend for all of the time between now and when the new products arrive. The CHIPS act will certainly help Intel, and if money from it will allow them to upgrade existing fabs instead of building new ones, even better. But politics will be a problem if Intel is closing fabs and losing jobs while taking CHIPS money to build new fabs. I’d sure like Gelsinger’s salary and stock options, but I believe in the short run, whatever he does is going to cause headaches.

Maybe I should just have written “I dunno” and saved myself a lot of typing ':wink:

Fool on!


Intel made this decision about two years ago. At the end of Swan’s time, he greatly reduced spending and capex toward new factories and started to make deals with TSMC for wafers. Intel replaced Swan with Gelsinger and the decision was made to go full swing toward factory expansions, rather than getting wafers from TSMC. Gelsinger doubled the budget and headcount for process development, and has since started construction on four new factories with two more in the works.

Old DUV factories can not be upgraded to EUV, so they live their life out on the older process up to Intel_7. Intel_4 and newer require a brand new fab. Intel currently has about $100B in capital targeted toward this expansion with 8 new fabs in various stages of construction. Due to the cutbacks in the Swan timeframe, Intel has very little EUV production capacity now, but starts adding that capacity later this year. I felt like they were very late with EUV capacity, but with the current market slowdown their timing seems very good.

In terms of capacity, I have been thinking about it like this:
Two new EUV factories for Intel X86 (Ireland and Israel)
one “chips act” factory for the NSA/US military (Arizona)
one “Brookfield” factory for contracting to QUALCOMM, apple, MediaTek, etc… (Arizona)
four more for future growth, moving some TSMC products back to Intel factories, etc… (Ohio x 2, and Germany x 2)

1 Like

I just recently posted my view on Intel as a manufacturer over at the AnandTech Forums:

“If Intel’s turnaround fails, the fabs may be sold or spun off into a joint venture, e.g. with GlobalFoundries, I guess. If Intel’s turnaround is successful, I picture they will end up like Samsung — a conglomerate with various subsidiaries, where in particular, product design and manufacturing are separate businesses insulated from each other.”

Intel Q4 Results | AnandTech Forums

I have for a long time been convinced that Intel’s IDM business model is unsustainable in the long run. I started a discussion thread about it a while back over at Anandtech Forums:

Speculation: Intel will become fabless | AnandTech Forums

In the end, the best chips in the industry may be manufactured by Intel, but not designed by Intel.


Do not overlook Pat Gelsinger’s ego. He has committed everything to IDM 2.0. Intel is firmly set on that course no matter what. It is now just a race between running out of money or becoming IDM 2.0.

Intel at this point is starting to throw everything off the ship they can in order to lighten it up and stretch it to the finish line. We will see how the employees react to this.

1 Like