Customers ( as in plural)? In terms of Smartphone modems, Intel pragmatically only had one customer; Apple. That was actually a contributing factor in it being a problem. As soon as Apple starting making big moves to jump into the modem space themselves they were going to loose leverage with Intel on the modem. Do this , do that, jump and do a triple back flip. ..... and oh by the way we're going to the rug out from under you completely in 2-3 years. If Intel could transition to some other substantive smartphone vendors in the future, but almost none are on discrete modems now and even fewer will be on discrete modems in the future. [ Even Apple about 1-3 years from now would have probably been looking for an "on package" modem chip , if not just a straight out license to weave into a single die/package. Apple is on track to drop out of the discrete modem package also sometime in the intermediate future. ]
If Apple highly ramps down and/or 'quits' on Qualcomm 3-4 years from now Qualcomm actually has more than several other customers. Qualcomm did OK when Apple stopped buying for a two years ( not super great, but far from 'dooms day" status. ). They'd probably do OK in the future if Apple 'quit' again.
The main market for discrete modem is more so going to be in PC/Tablets/Routers/ USB-'cards' . Also mini stations. These don't have the same lowest power consumption possible constraints Apple was probably asking for. ( for stuff plugged in and running 24/7 not even close. Low is better but not paramount. ). Note that Intel said they were quitting the smartphone 5G model ; not all of 5G modems. The non smartphone space is at least as big, if not potentially bigger, than the smartphone space. Intel could go get multiple customers there for their modems.
Intel merging a CPU+GPU+Modem in some EMIB/MCM package for future laptops is a far brigther market for them. If the modem is hard bundled to the core CPU package of the PC then margins won't be as big of a problem for Intel. Same for a package "mash up" of CPU+Wi-Fi-Modem for a mid-high end consumer router chipset if Intel wanted to go that way. Those two would also be a path out of discrete 5G modems for Intel that Apple has zero interest in. There is no "free lunch" there for Intel ( as Qualcomm, Samsung, Hauwei , etc. ) will have modems in this space too but the number of customers is higher ( so more possible designs to get a win on. )
That would have been in cooperation with Qualcomm. Apple ordering reference boards.
The 'tilt' aspect is probably over talked about in these forums. Apple got some stuff and Qualcomm got some stuff. Nitpicking on the stack on one side being incrementally higher than the other misses the forest for a tree. Neither one of these two companies are going to the 'poor house' over this deal. [ Apple wailing over how Qualcomm is over charging on their modem component is beyond lame when look at what Apple does to SSD/NAND cost markups with about zero added direct value. ]
Eh.... Intel has a major role but Apple's Scrooge McDuck skimping on their component suppliers is also part of the problem. Since they are about the only discrete smartphone modem buyer if they wanted to dump Qualcomm and go to someone with only one customer, Apple should have paid more ( not less) than what they were paying Qualcomm. Qualcomm has/had a order of magnitude more customers to spread R&D costs over. Intel doesn't. The complete 5G solution covers even more wider variety of radio frequencies/band and technologies than 3G-4G did. And Apple want's that all done on Scrooge McDuck pricing. Intel has money, but they also have about two order of magnitude more products to work on than Apple does.
Intel jumped in and bought Infineon ( https://newsroom.intel.com/news-rel...ion-of-infineons-wireless-solutions-business/ about the same time Apple dumped Infineon for Qualcomm. (and Infineon had underestimated LTE demand so lost even more customers pretty quickly). As long as Apple's primary strategy is to dump all of their cellular modem buys into one and only one vendor, they were going to have a long term problem finding an ecosystem of vendors to buy from. if starve off all the other options, eventually they won't be there. ( Qualcomm having a one modem that does pragmatically just about every cell service worldwide is/was attractive to Apple's view of simplifying supply chain. But that becomes a dual edge sword if taken to the extreme for long periods of time. Minding that other sword edge is Apple job. )
Bringing modems largely in house may not be much better in managing the complexity of the product evolution. Probably not the "Emperor's New Clothes" fab production problem Intel goofed on pretty badly 6+2 years contract is indicative that Apple isn't going to easily slam dunk this in a year or so either. This may take some substantial time.
On modems yes. That is more so driven by being Intel's one major customer in that space. ( Similar to collapse of Global Foundries 7nm business when AMD largely wanted to walk away. ) Off every Intel product possible? Maybe not. In part that depends upon AMD ( who doesn't have a completely spotless record either over last 5 years. Of recent better, but have stumbled also. ).
There isn't Mac volume to support complete detachment from the broader market R&D.
[doublepost=1555530196][/doublepost]
Intel is one of CEVA's license buyers. The core baseline of the cellular DSP could be here. ( versus custom in-house).
https://www.ceva-dsp.com/ceva-licensees/
If so that would be another reason for Apple to be a bit more than agitated to jump into doing it themselves.
"..CEVA expects revenues from Apple starting Q3.."
https://techtime.news/2018/09/04/ceva-19/
If they can license large chunks of a core baseline design and substantively "improve it" while at the same time sharing R&D costs like they do with ARM architecture license, then they'd be more prone to 'cut out the middle man'.
"...
The first companies that come to mind are the traditional players such as Qualcomm, Intel and various other SoC vendors. Between all of these vendors there is one distinction to be made: those who use CEVA IP for their modem designs and those who do not. ..."
https://www.anandtech.com/show/12450/ceva-announces-pentag-5g-ip
There is far more that goes into a pragmatically working system across a scores of cellular network set ups. Some aspects of wrapping the rest of the system around this is where Intel came up a bit short, ( And also major hiccups on fab process rollout and expectation management. )
Intel is just getting out of the 5G smartphone modem business. They didn't say they were getting all the way out of the modem business completely. There are other devices than just smartphones that will use 5G. Getting 5G into an "even thinner" iPhone is not the only problem market to be addressed.
Last edited:
