Can someone explain why Intel was having such a hard time competing with Qualcomm? What are the challenges in developing modems?
Having worked in the semiconductor business in Germany and having friends who worked at Infineon there, the cause would be the corporate culture that rewards incompetent management riding the gravy train(old boys club) instead of engineering effort and playing the politics game is the only way to move up, even though lots of engineers there are very talented people.
This, coupled with Intel's own innovation culture that fails at anything that doesn't involve milking the X86 resulted in a dumpster fire.
Under Apple, these same engineers could probably ship something competitive if Apple plays their cards right.
How Qualcomm shook down the cell phone industry for almost 20 years - We did a deep-dive into the 233-page ruling declaring Qualcomm a monopolist.
https://arstechnica.com/tech-policy/2019/05/how-qualcomm-sho...
[1] https://www.forbes.com/sites/patrickmoorhead/2019/01/16/if-t...
What I mean is that even today, if you buy a 4G enabled phone not powered by Qualcomm, your battery will drain in a few hours and you'll be left with a dead brick for the rest of the day.
Source: I spent years building 3G and 4G PHY chips at various QC competitors that have all since folded up or have been bought and merged into other teams.
I had an iPhone 7 with an intel modem (AT&T iPhone) for 3 years. The battery did not die in 4 hours.
Miniaturized, high frequency, low-power, high bandwidth, radio modems to be exact - supporting multiple complex protocols. In real-time. Meaning it is all done in hardware, although firmware does some management stuff. It's about as hard as silicon engineering gets.
( Disclaimer: I used to work there, could give more details, but NDA )
then they can‘t say anything because of nda. or the gsm standard manual costs hundreds of dollars.
my tin foil hat is glowing, i‘m telling you. we already know that gsm encryption is a joke. i don‘t want to imagine what‘s going on inside all those closed-source baseband firmwares. it creeps me out.
Then, the protocol stacks on top of that are enormous and complex. The software is very large.
Broadcom also tried and failed because they underestimated the task, IMO.
Those who can/could make it happen are those who are willing to invest massively over many years and, ideally who have a sure customer. E.g. Huawei who is willing to take the long view and who ships hundreds of millions of phones.
The only real product success they had was getting Apple for 4G and we can quite clearly see that as more of a strategic move by Apple against Qualcomm. Over that period there were numerous issues about Intel chips not being as high performance or low power as Qualcomm - which we'd hardly find surprising.
So there's that element to it, but the second element is that Intel has a fantastic reputation for destroying the businesses they acquire. Essentially what happens is that once they acquire a business they do a number of things:
1. They take a long time to perform these acquisitions. The timeline from first hints of acquisition to close can be years. So the engineering organisation of the company being acquired starts to hide problems because they fear it will endanger the acquisition. So you have a steady build up of issues that will overflow on day 1 after the deal closes.
2. They massively invest. This means massively increasing the cost structure of the acquired company, they do this whilst setting much higher targets to justify the investment. But there's two problems: The new employees take a long time to bring up to speed, and because Intel is constantly re-prioritizing you have existing engineers in Intel who are trying to move across into any role available. Suddenly you have a team of engineers in Folsom who basically have no work to do and so they're dumped into the new growth area (because as we all know, engineers are fungible commodities). The acquired company needs to figure how the hell these engineers are going to contribute.
3. They massively increase expectations. That investment has to pay off, so they set way higher new goals - and not "In 5 years time you need X revenue" more "This year you need 30% revenue increase". This immediately puts the new business in panic mode - EOLing products and doing sales tricks to invent revenue to hit the target. Year 2, all the sales guys know they pumped year 1's numbers to hit their targets so the good sales guys jump ship -either into different parts of Intel or entirely out of the company.
4. They find SYNERGY. What that means is every other branch of Intel will turn up and start either insisting you use their technology (guess what: You're using the Intel Fab now despite the fact we can't deliver on 10nm). Because Intel is so much larger than the companies they acquire your little business group suddenly has thousands of people coming to you saying "How about you work on this" or "Our client needs this let's bundle these products". That creates massive attrition on your core business.
5. They scale up: Intel is such a big company they literally can't chase small revenue - it would eat them up in COGS to sell $1m at a time. So very quickly the acquired business starts to lose its small customers that make up the revenue that made the business attractive in the first place.
So in the end: You've alienated all your customers, you've thrown your engineering organisation into dis-array, you've set ridiculous targets, and you're now owned by a company that's perfectly willing 5-10 years down the road to pull the plug on the entire sector you're working in. Hey presto: They 10 year cycle from Acquisition to Spin-off. Say hello to the boys at McAfee!