Apple's Acquisition of Intel's Smartphone Modem Business Completed, Intel Admits 'Multi-Billion Dollar Loss'

[ElectricPotato]

You can walk in and buy most Ryzen chips. Threadripper may be constrained now but I'd expect that to level out soon.

By your own admission, AMD is capacity constrained. The quantities required by a major manufacturer are a different matter than individual purchases from newegg. In Q3 2019 Intel did $10B in processors, AMD did $1.3B in both processors and graphics. It is safe to assume a constrained AMD will not be displacing Intel's volume any time soon.

I'm sure Apple doesn't care about their Special Intel treatment.

I'm equally sure they do care about early access to design information, allocation of parts and discounts for being exclusive to Intel.

AMD getting out of the foundry business (unlike Intel) means they can add capacity easier than Intel.

I don't follow this. Both AMD and Intel can buy capacity from a foundry (if available at the desired node), only Intel has their own production they can unilaterally decide to expand.

 

[Brien]

At least Intel makes better processors than AMD so this mistake should not hurt them that much (i.e. their computer chip business will make up the loss).

AMD is beating Intel, more importantly seems the industry is shifting to ARM (Win 10X for example), if I had Intel stock I’d be concerned.

 

[laz232]

I don't follow this. Both AMD and Intel can buy capacity from a foundry (if available at the desired node), only Intel has their own production they can unilaterally decide to expand.

CPU processes are different to modem processes and cannot be transferred even if they may be on the same "node". A 28nm process that is suitable for logic is not something that can just be used for RF designs (AFAIK).

Intel CPUs are on Intel's own fab process.
Intel modems are (were?) on TMSC's 28nm process.
AMD produces its CPU/GPUs at Global Foundries AFAIK.

NB am not a chip designer, but have worked at Intel Mobile Communications division a while back.

 

[konqerror]

CPU processes are different to modem processes and cannot be transferred even if they may be on the same "node". A 28nm process that is suitable for logic is not something that can just be used for RF designs (AFAIK).

Intel CPUs are on Intel's own fab process.
Intel modems are (were?) on TMSC's 28nm process.
AMD produces its CPU/GPUs at Global Foundries AFAIK.

There's two parts to a modem. One is the baseband. It is all digital, contains a CPU for managing data flow, a number of DSPs and digital accelerators. Intel's basebands are currently on their own 14 nm node. In modern wireless standards, signal processing takes a huge amount of computation, and therefore battery power, so CPU technology is directly applicable.

The second piece is the RFIC, which is a separate package or at least die. This is a mixed signal design where all the RF parts lie. Intel has a separate process based on 14 nm but with different parameters, called 22FFL.

So yes, Intel is transferring their CPU technology to both parts of their modems. Recall that Intel has business in Wi-Fi that need RF chips too.

 

[Kabeyun]

I see them as 2 separate matters.

Qualcomm knows very well that Apple doesn't want to remain as Qualcomm's customer forever, so they likely wanted to lock in Apple's business for as long a period of time as possible. Meanwhile, Apple is definitely exploring the development of their own 5g modems, but this takes time, and until that day comes, they still need Qualcomm.

The deal has bought Apple some time to develop their own modems. I don't think Apple will actually need the full 6 years to come up with a viable alternative, but based on how the lawsuit earlier this year turned out, Apple seemed to be in a hurry to settle as the odds were not in their favour (plus Apple most certainly needs a 5g modem in time for the 2020 iPhone), so I suspect Apple simply didn't have the leverage to negotiate a better deal. So 6 years it was.

As for whether Apple is able to deploy their own 5g modems, I have no idea, but it's also possible that they may decide to just sit on it and continue refining it until their contract with Qualcomm expires. In all fairness, Qualcomm's modems are very good; the key reason why Apple went with Intel ones was due to lower licensing costs, not so much performance.

Thanks for the post but you didn’t answer the question. If Apple & QC have a 6-year deal how will Apple release a phone with an in-house modem in 3-4 years? As I said, the only stipulation I can think of is if the deal would allow Apple to deploy their own modem alongside QCs. First, it’d have been weird for QC to agree to that. Second, Apple certainly anticipated Intel’s cellphone modem collapse after the QC deal, which means it’s an exclusivity deal.

 

[Abazigal]

Thanks for the post but you didn’t answer the question. If Apple & QC have a 6-year deal how will Apple release a phone with an in-house modem in 3-4 years? As I said, the only stipulation I can think of is if the deal would allow Apple to deploy their own modem alongside QCs. First, it’d have been weird for QC to agree to that. Second, Apple certainly anticipated Intel’s cellphone modem collapse after the QC deal, which means it’s an exclusivity deal.

I guess my answer is that Apple won’t. 3-4 years might be their own internal deadline, but they may not actually use their own modems until their contract with Qualcomm expires, even if it is ready before that.

I also suspect that we might see an Apple-branded modem in another product first. Apple will likely want to test it on a lower volume product that carries with it a lower penalty for failure (like a wearable perhaps?) to make sure it works properly before deploying it en-mass on their most lucrative product - the iPhone.

Either way, I don’t think Apple is actually in any rush now to use their own modems. The key reason for their lawsuit in the first place was because Apple felt they were paying too much, and I am willing to bet that the new contract terms includes cheaper modem prices for Apple. So in a way, Apple likely still got what it wanted out of the lawsuit.

 

[Kabeyun]

I guess my answer is that Apple won’t. 3-4 years might be their own internal deadline, but they may not actually use their own modems until their contract with Qualcomm expires, even if it is ready before that.

So your solution to [how can A and B both be true if A is incompatible with B] is that B is not true. Fair enough. According to this article the 2022-2023 year came from Fast Company. No idea how reliable their predictions are.

 

[Ronlap]

Qualcomm is basically killing the market...
I’m very intrigued by Apple solution in 2022. When they entered the market for CPU and GPU in the past years, they literally annihilated competitors.

Indeed. Top designers seem to love working for Apple since they can really reach outside the box and combine multiple technologies onto the same chips.

Still want Qualcomm chips though.

Why? Right now Qualcomm > Intel, but Apply may be > Qualcomm just as their purchase agreement runs out.

With the inclusion of Intel’s mobile chip IP in the purchase, does this mean there could be potential improvements with existing/older iPhones powered by Intel chips?

Only if the Delorean hits 88 MPH. They cannot go back in time to replace existing hardware -- and it IS the hardware that is inferior.

So now, starting with 2021 probably we will have macs with 5G built in

I've never grasped why anyone would want to pay for an additional line of service for their tablet or laptop when the phone strapped to their hip has a hotspot built into it and you're probably already paying for unlimited data. Yeah, I usually buy the cellular iPad, but that's to get the GPS.

 

[Rigby]

these numbers don’t add up unless the licensing agreement allows Apple to deploy their own modems alongside QC’s. Which would be weird. Any other interpretations?

You're mixing up two different things. The 6-year licensing agreement just settles the patent dispute. In addition a "multiyear supply agreement" was announced. We don't know the terms obviously, but I doubt that Apple agreed to using QCOM chips exclusively. My guess is that we'll see Intel-derived modems in some of their products (e.g. cellular iPads) fairly soon.
[automerge]1575388019[/automerge]

AMD is beating Intel, more importantly seems the industry is shifting to ARM (Win 10X for example), if I had Intel stock I’d be concerned.

The industry has been "shifting to ARM" for a decade now. Intel's market share doesn't seem impressed.

 

[ElectricPotato]

CPU processes are different to modem processes

OK. I still don't see why that means AMD's reliance on third party production gives them a greater ability to expand capacity than Intel (with both foundry and internal production), which is the topic of the post you are responding to.

AMD produces its CPU/GPUs at Global Foundries AFAIK.

AMD moved 7nm production to TSMC. This includes Ryzen and Epyc.

 

[laz232]

There's two parts to a modem. One is the baseband. It is all digital, contains a CPU for managing data flow, a number of DSPs and digital accelerators. Intel's basebands are currently on their own 14 nm node. In modern wireless standards, signal processing takes a huge amount of computation, and therefore battery power, so CPU technology is directly applicable.

The second piece is the RFIC, which is a separate package or at least die. This is a mixed signal design where all the RF parts lie. Intel has a separate process based on 14 nm but with different parameters, called 22FFL.

So yes, Intel is transferring their CPU technology to both parts of their modems. Recall that Intel has business in Wi-Fi that need RF chips too.

Interesting - I see that's a generation or two after I left the company... I hadn't realised that Intel was now in the RFIC game (usually I just peripherally follow the GaAs fabs, none of the high-volume stuff)

 

[mi7chy]

What goes around comes around. No sympathy.

 

[techwhiz]

Conservative? Like when they surprised the whole industry and put 64-bit chips in all their mobile products?

Or conservative like when they flipped a switch and put a completely new file system on millions of devices all at once?
[automerge]1575317732[/automerge]

There’s always triquint

https://casetext.com/case/triquint-semiconductor-2

FBARs are fascinating to me.

Actually the name is Qorvo now. It looks like Apple is using their LTE filters already.
Sometime around 2018 they started using them and dumped Broadcom.
I haven't been paying that much attention.

But FBARs are interesting. I did some signal processing stuff in grad school but nothing lie this stuff.

 

[konqerror]

I've never grasped why anyone would want to pay for an additional line of service for their tablet or laptop when the phone strapped to their hip has a hotspot built into it and you're probably already paying for unlimited data. Yeah, I usually buy the cellular iPad, but that's to get the GPS.

Less fiddly, especially for line of business use where somebody needs data access but you don't want to give them a phone.

You only need to manage one battery. If the device is shared, you don't need to keep track of who has what or the case where somebody hands over the laptop and forgets to give the phone. More manageable by not having a phone to update and not having to manage a hotspot SSID and password.

More secure, and less abuse potential, like people making phone calls and using Snapchat on the phone. For example, a company can sign up for a special carrier plan on a laptop that directly connects it to the company network where the connection is restricted to internal services.