Qualcomm CEO says Intel ‘not an option’ for chip production — yet

Qualcomm
(Image credit: Qualcomm)

Qualcomm CEO Cristiano Amon says that Intel’s chipmaking tech still isn’t up to snuff — at least not for the Snapdragon X. In an interview with Bloomberg published September 5, Amon said that Intel “is not an option today,” but left open the possibility for a future partnership, adding “we would like Intel to be an option.”

It’s a short but pointed comment that lands hard in the middle of Intel’s planned foundry turnaround. The company has staked its future on becoming a contract manufacturer for other chip designers and has repeatedly said that its roadmap depends on securing a large external customer. Unfortunately for Intel, Amon’s remarks obliterate one of Intel’s most realistic prospects for building advanced client silicon for an outside firm, at least in the near term.

Performance is improving fast enough that Qualcomm is now a direct competitor to Intel in thin-and-light notebooks. This gives Amon’s statement some significant weight: One of the most promising companies in the PC space just publicly said that Intel isn't ready to deliver on its needs.

Intel said in July that it might pause or abandon 14A development if it can’t win significant external business or achieve critical progress targets. Since then, questions have been raised about execution risk on 18A, the node the company has pitched as its return to industry leadership, due to yield issues. Amon’s comment doesn’t help.

Still, Qualcomm hasn’t slammed the door on Intel entirely. Amon said that his company would consider Intel if it could deliver on efficiency, and the two companies have previously signaled interest in working together. But for now at least, the Snapdragon X is staying with TSMC.

Luke James
Contributor

Luke James is a freelance writer and journalist.  Although his background is in legal, he has a personal interest in all things tech, especially hardware and microelectronics, and anything regulatory. 

  • thestryker
    I've seen multiple reports taking the quote to mean their chip manufacturing isn't up to par. Without any context to the conversation it's impossible to make that leap (even Bloomberg's own reporting contains nothing but his quote with regards to the conversation). It's entirely possible that's what he's saying of course it's also possible there are capacity and timeline issues to contend with.

    Intel has a singular fab for 18A because the second one isn't up and running yet and I'm fairly certain Intel wouldn't be running external customer's chips through their development fab.

    From the outside it also seems fair to say that 18A yields are likely behind original projections. Gelsinger was all in on being able to do external volume at the same time as internal. Whether or not the external issues have to do with all the capacity slow downs or node issues nobody external can say, but my bet is a combination of the two.
    Reply
  • acadia11
    Intel needs to spin off foundry so it can properly address the market, but still retain a controlling ownership otherwise it will end up GF, which after its spin off from AmD for a time remained a viable leading edge provider but exited that market in 2017-18 focusing on trailing edge and near leading edg capability. And no longer could really meet AMD most advanced needs.

    There would be a chance a completely independent intel foundry would make the same decision. Leaving just two leading edge players Samsung and TSMC. Politics aside Intel needs to commit to foundry not just throwing money but addressing this perception and ease of adoption as a contract manufacturer. What would lead Qualcomm to say they aren’t there yet? Tech they have it but there ability to support clients is suspect because they also compete with Qualcomm.
    Reply
  • Notton
    If Qualcomm doesn't want it with Snapdragon X, which is a smaller monolithic, then that must mean yields are really bad.

    To put the size into perspective...
    SDX: N4P, 169.6mm²
    R7 350: N4, 195mm²
    RZ1E: N4, 178mm²
    9060XT: N4P, 199mm²
    5060Ti: 4N, 181mm²
    5050: 4N, 149mm²
    Reply
  • wussupi83
    acadia11 said:
    Intel needs to spin off foundry so it can properly address the market, but still retain a controlling ownership otherwise it will end up GF, which after its spin off from AmD for a time remained a viable leading edge provider but exited that market in 2017-18 focusing on trailing edge and near leading edg capability. And no longer could really meet AMD most advanced needs.

    There would be a chance a completely independent intel foundry would make the same decision. Leaving just two leading edge players Samsung and TSMC. Politics aside Intel needs to commit to foundry not just throwing money but addressing this perception and ease of adoption as a contract manufacturer. What would lead Qualcomm to say they aren’t there yet? Tech they have it but there ability to support clients is suspect because they also compete with Qualcomm.
    It's always easier to say these things in hindsight but I agree. I think Gelsinger was right to identify the world was going to need more fabs in coming years. But the idea that competitors were going to help pay for Intel's build out seems half baked.
    Reply
  • acadia11
    wussupi83 said:
    It's always easier to say these things in hindsight but I agree. I think Gelsinger was right to identify the world was going to need more fabs in coming years. But the idea that competitors were going to help pay for Intel's build out seems half baked.
    I think it’s less about them being a competitor and more they aren’t able to operate as a outsourcing manufacturer should. They are getting better as a partner but still had to redesign their tooling and design capabilities to be more flexible to build and accommodate non-Intel designs. Gelsinger was right … the problem has always been that they took an Intel first attitude vs TSMC which by default is customer need first attitude since they aren’t competing in producing convenience for themselves.
    Reply
  • hotaru251
    thestryker said:
    Intel has a singular fab for 18A because the second one isn't up and running yet and I'm fairly certain Intel wouldn't be running external customer's chips through their development fab.
    Intel doesnt even use their own fab for their best chips...the only reason you do that is if you know youre not as good as other option and paying others is better than using your own lesser fab.

    when company who owns it doesnt use it much why would anyone else want to use it?
    Reply
  • TerryLaze
    hotaru251 said:
    the only reason you do that
    Intel had the option to do yet another (a 3rd) gen on intel7 or wait for 2-3 years to release the next gen until development and fab building of 20A or 18A would be ready.
    Or just use tsmc.
    hotaru251 said:
    if you know youre not as good as other option and paying others is better than using your own lesser fab.
    The problem with that is that 15th gen is worse in every single way other than power efficiency, and that's only because tsmc can't reach the same clocks as intel can....
    hotaru251 said:
    when company who owns it doesnt use it much why would anyone else want to use it?
    This is about 18A and not previous nodes, intel is already making xeons on 18A and is preparing a whole lineup for desktop starting end of year.
    https://www.tomshardware.com/desktops/servers/intel-reveals-288-core-xeon
    Reply
  • thestryker
    hotaru251 said:
    Intel doesnt even use their own fab for their best chips...the only reason you do that is if you know youre not as good as other option and paying others is better than using your own lesser fab.

    when company who owns it doesnt use it much why would anyone else want to use it?
    Last time I checked 18A is currently in ramp for volume manufacturing. How exactly was Intel supposed to use a node that wasn't in volume yet?
    Reply
  • watzupken
    thestryker said:
    Last time I checked 18A is currently in ramp for volume manufacturing. How exactly was Intel supposed to use a node that wasn't in volume yet?
    Volume manufacturing/ production does not always mean that the yield is good. You can always look back on reports on Samsung's volume production followed by reports of "abysmal yields. While the CEO of Qualcomm did not specifically mention what is not ready/ up to his standard, if you connect back to what Broadcom reported on poor Intel 20A yield, you should kind of get an idea. There is no magic changing the number from 20A to 18A, such that the yield will improve without material change in their fabrication process. Samsung is a great example in this case.
    Reply
  • rluker5
    Amon could have been referring to capacity.
    Does Intel currently have enough 18A capacity for Qualcomm? Simple question. If they don't then they are not a viable option. Regardless of any conjecture on yields or performance.
    Reply