Apple completes $1bn amputation of Intel's 5G modem biz, Chipzilla out of mobiles for good Apple's acquisition of the bulk of Intel's smartphone business, valued at $1bn, was completed today against the backdrop of of a complaint from Chipzilla that the deal was done "at a multi-billion dollar loss". The completion of the deal, first announced in July, came three days after Intel filed an amicus brief (PDF) bitterly …
"Apple's acquisition of the bulk of Intel's smartphone business, valued at $1bn, was completed today against the backdrop of of a complaint from Chipzilla that the deal was done "at a multi-billion dollar loss"."
Well Intel, show us working 5G modems and we might believe you. The 5G group were producing barely functional products and blaming the low yields on 10nm.
Only as yields on 10nm went up, functional 5G parts continued to be rarer than the lesser speckled unicorn - the rarest of the unicorn species...
Apple paid for patents, not for products.
Qualcomm tried to get in on the server CPU market only to throw in the towel shortly after making a big splash. Intel competitive pressures? Then Intel tries to get in on the modem market and I guess Qualcomm beat them back. What is sort of strange though is both companies have vast resources, they both gave up too easily. Both sides seemed to have fine working products even if they were not best in class.
"Both sides seemed to have fine working products even if they were not best in class."
In the world of tech greed, if you are not best in class, you won't make ultimate bucks and control *all* consumers.
They certainly don't want to be the ol' reliable fallback guy. No money in simply making our lives better. ;)
If you look through Intel's modem product range (xDSL and 4G/5G) they had a lot of issues or straight out didn't work. Intel didn't walk away, they couldn't compete...
As for Qualcomm on the server CPU side, their challenge was entering a reasonably competitive market (high end multicore ARM/low end multicore x86/others) where there was significant investment required to get high core count ARM chips working with decent IO/memory bandwidth while also making money. That risk was compounded by being dependent on a fab partner (Samsung) and Intel appearing to have a significant advantage in the period Qualcomm tried (2013-2018). The problem was Qualcomm likely spent $0.5bn to get to that point and needed another $1bn to get the next generation done and I'm not sure how much they made on server chips. I'm not saying Qualcomm couldn't have made competitive ARM server chips and their products worked without any significant issues, it's just that the costs were significant and the risks made it hard to justify continuing. I'm unsure if Intel floundering would have helped in any significant way given AMD/TSMC's rise.
I'm kind of wondering if the brouhaha Intel made about 5G modems was aimed to get Qualcomm to step down on it's server CPU ambitions. Sort of a: "Fine, you want to step into our market? Fine looking mobile applications modem market you got there. Would be a shame if something were to 'happen' to it, wouldn't it?"
While Qualcomm may have been a consideration, the stated aim was to increase utilisation of existing fabs which made sense based on the costs of 22nm/14nm/10nm fabs and started with WiMax, so before ARM server chips were even considered.
WiMAX/3G/4G/5G/xDSL required high component counts to achieve the required performance levels and it fitted in nicely with Intel's existing fab strategy up until the point where tick tock became tick tock tock tock tock.
Because Apple told them that they weren’t paying any more, that’s why. After all, who else could they sell to? Microsoft is no longer in the cell phone business, and the likes of Samsung are too deep in Qualcomm’s pockets. There was really only one possible customer, and Apple was in a position to just live with Qualcomm until they built their own stuff. It was sell to Apple, at Apple’s price, or lose all the investment.
Apple mostly wanted the tech for the patents and was not going to overpay. Intel had zero leverage.
Once Intel was forced to throw in the towel after years of trying to get people to use x86 chips in phones and tablets, the cellular business was no longer of any use to Intel itself. Apple was the only outside customer/user for their cellular modems, but they were forced to settle with Qualcomm because they couldn't trust Intel would be able to make a working 5G chip. Then Intel then had zero customers.
How much is a business with zero customers worth? They lost billions because they bought companies that had customers, and managed to lose them all through a combination of trying to force x86 into a market that doesn't want it and designing substandard cellular modems. They are lucky that Apple REALLY wants to be out from under Qualcomm's thumb, if not for that Intel would have had to shut down the division without a buyer and ended up losing a billion more on the whole ordeal.
No-one made Intel sell most of the ARM to Marvell (though it had come from DEC).
Their Cable modems were a problem, not just the xDSL.
The i960 was good, but that was a long time ago.
The 64 bit instructions of the successful x86-64 were an AMD idea, Intel had done the HP inspired Itanic, which had a very short lived Windows XP, a fail.
The Intel integrated graphics were a disaster for years, crippling performance of cheaper PC/laptops.
I suppose the 10/100M ethernet cards and the Intel WiFi was/are OK.
Now it seems almost all the clever performance enhancements of Intel CPUs are security risks.
Where is the persistent storage product and is it anything like as good as promised?
Why did they buy McAfee?
The shine seems to have gone off Chipzilla.
I could write about why 10nm isn't, however no-one's chip geometry now means anything like it meant at 90nm or maybe 45nm. The figure isn't the typical geometry, but the width of the smallest feature.
There are more ARM cpus shipped in less than a week than Intel cpus in a year. The majority of users use ARM now to message, browse, view video, read ebooks (52% on phones), listen to audio books etc. Intel had a full ARM licence and did do ARM development.
...they didn't go screaming to Congress that their 5G tech is "too big to fail" and "critical to national security". How I was expecting this to roll is that Intel was going to play the national security card, try to get multiple billions of taxpayer dollars in a bailout, THEN try to make a marketable product. Or spend the cash on biz jets, entertainment, and blow.
If Apple have indeed taken their 5G tech in-house but can't get it to work, there will be no 5G Apple generation, they will have to stick at 4G. So what would you replace your old iPhone with? Buy a 5G Android that is faster in the lab but has to play catchup whenever a truck goes past and loses signal completely as soon as you step indoors away from the rain and noise, or an otherwise far shinier 4G iPhone that you can at least use near a window?
Not that Apple can be planning such a thing or they would have left Intel's business well alone. H'mm, unless it's just a 1bn patent hedge so they can still make money if the lemmings do fall for 5G? My head hurts.
Apple have already agreed to use Qualcomm's 5G modems so crisis averted.
As for why Apple wants what Intel sold? They can potentially do a gen-2/3 modem that does work with sensible power usage and avoids having to buy large numbers of buckets and employ large numbers of people to carry them to Qualcomm each time they fill with cash...
As for your 5G predictions? It's coming whether you like it or need it. The complaints about poor 5G service were just as applicable to early 3G/4G rollouts so no real change.
The 5G is nearly irrelevant. Basically subscription WiFi tech as far as RF. The important aspects of 5G are infrastructure changes which allow more seamless use of 2G, 3G, 4G and WiFi with voice and data. A more unified backhaul and back office.
An actual 5G radio in a phone doesn't much matter.
Apple has a "multiyear licensing deal" with Qualcomm, and will be using their 5G chips next fall. Given that they designed by far the leading the ARM core, I'm sure they can design a working 5G modem. It doesn't need to be better than Qualcomm's, it only needs to do the job.
Might take them a couple years, but they have given themselves plenty of time thanks to the Qualcomm deal. The real fight will be when their modem is ready, Qualcomm will no doubt sue immediately so the cellular IP they acquired from Intel (and in the past from Nortel) will come in handy.
If they don't think that will be enough to take them on, don't be shocked if there isn't some sort of deal announced where Apple buys some of Huawei's 5G IP, so they're in a position to force Qualcomm into a cross licensing deal.
Producing modems is a rather different and more difficult art than producing CPUs. Even the basic spec in 4G is about 3 or 4 times the number of pages as all of the ARM v8 instruction set for the EUTRAN protocol stack (which is more mathematically sophisticated than ARM in certain layers), and the actual implementation is wholly different from CPUs requiring special process steps with design being very functionally (rather than performance) sensitive to transistor layout. Much of the difficulty is in the intricate realtime software running on top of the silicon, and a load of lengthy validation that needs to be done on a country by country basis.
Apple probably has bought itself an affordable discount off of whatever royalties it owes Qualcomm, but it will probably take it some 4 to 5 years to get to a working 5G modem. Intel's efforts were well resourced but mismanaged; Apple probably will be better managed but the problem is still very hard and the rest of the industry is not standing still. Apple has demonstrated a willingness to cut their product and divisional complexity in exchange for better performance such as cutting compatibility with legacy apps or folding their Mac division into their iPhone division. This may be acceptable for user facing apps so long as users are still attracted to the brand and don't miss the compatibility in exchange for hot new performance, but the cellular connectivity layer is just expected to work and be abstracted away beneath the apps. They have enough engineers to integrate Qualcomm's 5G solutions, but developing their own modem stack doesn't play to their strengths which are really app facing and seems like a big technical distraction to their bottom line when they can buy an excellent solution for a couple bucks per phone from a third party.
Apple believes in vertical integration.
They design their own CPUs, now they probably want to design their own modems too. In the distant future, they can integrate the CPU and modem onto the same chip, or perhaps just the same multi-chip-package, to make their phones a tiny bit smaller, lighter and cheaper to manufacture.
They have plenty of money to throw at R&D, so they'll get it working eventually.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
