Gururu's comment and the quoted text from the article seems like a reasonable possiblity for what's going on.
"we’re developing Intel14A as a foundry node from the ground up in close partnership with large external customers. This is essential to designing a process that meets specific customer requirements and enables us to address a broader segment of the market. Going forward, our investment in Intel 14A will be based on confirmed customer commitments."
That could definitely read that Intel has some customers, and they just don't want to talk about them openly all that much. Lots of possible reasons, depending on who the customers are.
Other thing, personally, is that there's very little ability for "normal" people to request something being manufactured. And at least in part, that probably disincentivizes large customers. It's 1000's of millions of dollars, or it's nothing.
I can't really request a "couple" wafers, or 1000's of chips. Even if I know what I'm doing, and request "just lithography me something" because they require significant upfront Non-Recurring Engineering (NRE) & IP integration costs to use Intel's RibbonFET transistors and PowerVia backside power delivery.
Except, even for big people, that's probably still daunting. You can't "try" something. Sign the $1000 million order, or go talk to somebody else. You can't do a "lets try a $million order, or $10 million order, and if we like it, maybe a $1000 million order."
I get that they don't want people to just steal their tech, with some cheapo order, and that below some amount, it's probably more work just trying to even load their files in the machine and print something (even if they know what they're doing). Yet it's probably still a disincentive.
Rumor mill places the wafer cost at $25,000-$30,000 / wafer [1] (which is a whole nother issue with the lack of transparent pricing, ref is a Twitter post...). However, being able to pay double(?, quad?, probably way more, 10x?) that and get a short run wafer would probably have a larger number willing to "try something". (even if it was, "we'll print what you send, and if it's broken and doesn't work, it's your fault").
To be fair to Intel, TSMC is basically almost the same. If you're not Samsung / Nvidia scale, they probably don't want to talk with you.
Edit: Probably for my own reference. The short run stuff is usually referred to as Multi-Project Wafer (MPW) Shuttle Programs and Tower Semiconductor has a relatively well written site [2] describing the basic idea and requirements.
Looking through GlobalFoundries, Advanced Micro Foundry, Microchip Technology, and Tower (people that offer MWP), it's really difficult to find a site that lays out "what we expect from you", "what the base costs are", "what additional fees are", "what you can expect to receive".
Edit2: Actually, the MOSIS people [3] (out of University of Southern California’s Viterbi School of Engineering [4]) have a really cool site that shows all the various process design steps and lets you add and remove and create your own). [5]
I hope Intel succeeds at 1.4 and manages to provide sufficient capacity. AFAIK, they are doing 1.8 at their research fab in Oregon, not sure how many wafers it can crank out per month but hopefully they'll manage to transfer the tech to a real fab when demand increases.
I've lost track of Intel's fab cancellations and restarts, I'm not even sure if they have a (near) completed production fab with EUV in place.
Intel CEO confirms that he doesn’t know what he’s doing as well as previous Intel CEOs, lest we worry.
Intel CEO assures the public the check is in the mail
Gururu's comment and the quoted text from the article seems like a reasonable possiblity for what's going on.
That could definitely read that Intel has some customers, and they just don't want to talk about them openly all that much. Lots of possible reasons, depending on who the customers are.Other thing, personally, is that there's very little ability for "normal" people to request something being manufactured. And at least in part, that probably disincentivizes large customers. It's 1000's of millions of dollars, or it's nothing.
I can't really request a "couple" wafers, or 1000's of chips. Even if I know what I'm doing, and request "just lithography me something" because they require significant upfront Non-Recurring Engineering (NRE) & IP integration costs to use Intel's RibbonFET transistors and PowerVia backside power delivery.
Except, even for big people, that's probably still daunting. You can't "try" something. Sign the $1000 million order, or go talk to somebody else. You can't do a "lets try a $million order, or $10 million order, and if we like it, maybe a $1000 million order."
I get that they don't want people to just steal their tech, with some cheapo order, and that below some amount, it's probably more work just trying to even load their files in the machine and print something (even if they know what they're doing). Yet it's probably still a disincentive.
Rumor mill places the wafer cost at $25,000-$30,000 / wafer [1] (which is a whole nother issue with the lack of transparent pricing, ref is a Twitter post...). However, being able to pay double(?, quad?, probably way more, 10x?) that and get a short run wafer would probably have a larger number willing to "try something". (even if it was, "we'll print what you send, and if it's broken and doesn't work, it's your fault").
[1] https://x.com/zyl19911/status/1876466680169414861
To be fair to Intel, TSMC is basically almost the same. If you're not Samsung / Nvidia scale, they probably don't want to talk with you.
Edit: Probably for my own reference. The short run stuff is usually referred to as Multi-Project Wafer (MPW) Shuttle Programs and Tower Semiconductor has a relatively well written site [2] describing the basic idea and requirements.
[2] https://towersemi.com/manufacturing/mpw-shuttle-program/
Looking through GlobalFoundries, Advanced Micro Foundry, Microchip Technology, and Tower (people that offer MWP), it's really difficult to find a site that lays out "what we expect from you", "what the base costs are", "what additional fees are", "what you can expect to receive".
Edit2: Actually, the MOSIS people [3] (out of University of Southern California’s Viterbi School of Engineering [4]) have a really cool site that shows all the various process design steps and lets you add and remove and create your own). [5]
[3] https://www.mosis2.com/
[4] https://www.isi.edu/
[5] https://www.mosis2.com/fab-service-explorer
I hope Intel succeeds at 1.4 and manages to provide sufficient capacity. AFAIK, they are doing 1.8 at their research fab in Oregon, not sure how many wafers it can crank out per month but hopefully they'll manage to transfer the tech to a real fab when demand increases.
I've lost track of Intel's fab cancellations and restarts, I'm not even sure if they have a (near) completed production fab with EUV in place.