It's so weird that an oligopolistic industry with the largest barriers to entry (your average CPU factory will cost $5/10+ billions, and that's just the factory) have the lowest margin, by far. Why is that so?
While there are not many competitors, they still do (most times) properly compete due to various factors like pressure from their customer (some very very large with the capabilities to become a competition iff they stop innovating and/or become to expensive), but also factors like political pressure to stay ahead of mainland china.
Through TSMC for example has margin >40%. Which isn't that surprising I mean there is not much competition. Intel doesn't (yet) provide their capabilities to anyone else, and the other competition is behind enough to not be an option for many things, especially high end and/or high efficiency.
What is quite fascinating is that while TSMC is a close to monopoly(/duopoly if you count Intel) in many sectors and takes advantage of it they seem to have successfully resisted the common problem of such companies stopping/slowing innovation and similar problem. I'm not sure if it's related to the additional pressure of them being on of the main lifelines for Taiwan or if they have just pretty good management or if it's because they still competition oligopoly main customers do push them to do so.
It might be that that's not as high of a barrier to entry as it seems. It would cost much more than that to replicate the engineering quality that goes into the Apple system of products for instance.
I cannot speak to AMD, but for the case of Intel, it is hard to be a fast follower. The billion dollar factory translates to hundreds of millions of dollars of depreciation each year - a fixed cost that means declines in revenue hit the bottom line very hard.
Margins are not determined by how many companies sell certain products, but rather by willing the individual buyer is to consider different products. Both Apple and Microsoft can charge high prices for MacOS and Windows, because most Mac and Windows users will not consider another OS. But most people who currently use a laptop with an Intel chip will consider a laptop with an AMD chip, and vice versa, even speaking of an Intel user or an AMD user feels wrong.
Do you happen to know if it has always been this way, or is this the result of a gradual squeeze to a limit due to megahertz wars, increasingly shortened product cycles, and each companies product line structured in a way that tends to cannibalize their own sales before creating fresh profit.
And the decreasing return, in practical terms, for their customer base when upgrading since most basic computation tasks have long been solved and it's only exotic high-end workloads (plus constant bloat scaling factor that is largely a consequence of the added performance, rather than it's target) that benefit from added performance.
In short, was Intel making percent, or even single digit margins back in the 80s or even early '90s?
Thats a great question; I'd have to dig into their statements to see more. It's hard to get reliable data from that long ago for free. From what I remember as a consumer, MSFT + INTC (Wintel) absolutely cleaned up in the 90s until AMD started the CPU price & MHz wars, and Apple started making MacBooks.
For Intel, it looks like they kept one dollar out of every four until the recent crash. I don't know the reason for the decline, but I know I wouldn't want to compete against Apple.
Having margins because no one can legally compete with you is pretty remarkable in the chip industry. No one else has carved out such a large market like this.
The other bit that blows me away is that sometimes it feels like Qualcomm's perpetual inability to integrate chips is a vast submarine point no one using Qualcomm sees coming, but which forever works out in Qualcomm's favor. I don't know how many chips a MediaTek cellular modem takes, but iPhone 15 pro has 6+ chips dedicated to cellular, probably more. Modem, two transceivers, an envelope tracker, and probably a bunch of in this case non-qualcomm power supplies but most probably use Qualcomm pmics, maybe plural. https://www.ifixit.com/Guide/iPhone+15+Pro+Max+Chip+ID/16532... . There's also separate but non-Qualcomm solutions for wifi & bt too, where-as again I rather expect a company like MediaTek has much more working together.
I'm really interested to see chip identifications of the upcoming desktop/server class Qualcomm Oryon, supposedly by the Nuvia team (one of the strongest possible cases for anti-trust I have ever seen). https://www.semiaccurate.com/2023/09/26/whats-going-on-with-... is not partial journalism, but this cavalcade of ridiculousness is entirely the sort of poorly integrated Qualcomm solution I'd expect, that both builds terrible terrible products but which also lets Qualcomm fail upwards by requiring 5x more board components than is reasonable. Terrible for consumers, OEMs are in tears, but oh sorry, your design is late hot over priced and seemed reasonable but you're going to have to add a bunch more pmic's to get to market because we are Qualcomm and we spend more on lawyers than design.
> Qualcomm’s licensing business is best-in-class and an example many companies can learn from.
The last thing this world needs is more Qualcomm. I'd rather see the entire patent system go down in flames than this utter crap continuing.
Qualcomm has been a vampire on technology for years now, it's time for this company to end once and for all. We have to get technology under the people's control again - it is ridiculous and stifling progress that governments legally prescribe standards that are not open to everyone to implement. And especially it's ridiculous how few notable manufacturers of mobile phone SoCs remain - it's either Qualcomm or Mediatek.
> it is ridiculous and stifling progress that governments legally prescribe standards that are not open to everyone to implement
Standards-essential patents are required to be licensed this way. It's called "RAND licensing": reasonable and non-discriminatory. Everyone who participates in, say, ISO or the ITU, is required to fill out a form documenting if they own part of the standard and what terms they agree to license it on. If they refuse RAND licensing, their inventions get ripped out of the standard[0].
What you want is royalty-free RAND licensing. There are a few standards bodies that insist on royalty-free licensing, notably the W3C[1] and Khronos Group[2], but it's uncommon. Most standards bodies are in fields of endeavor that don't actually require free redistribution - in fact, they don't want it. They pay the people doing research for the standards on a "for exposure" basis[3]: they get their patented inventions in the standard, and as payment for their work, they get to charge a $2/unit royalty on every user, which they can then pour into research that yields more patents.
I'm going to be honest, royalty-bearing RAND (as in, pay up or don't implement) is probably the right fit for a lot of industries. We hate it because it steps on our (FOSS) toes, and if we can develop the software as a community, why can't we invent the underlying technology too? This is the idea behind royalty-free video codecs like AV1 - it turns out a community of amateurs can do video codec research after all[4]. But at the same time, cellular radio technology is quite possibly the worst thing to leave to amateur[5] software development. There's a lot of very expensive lab tests you need to do to make sure the technology works, to ensure hundreds of devices can communicate with the towers without jamming each other, and to ensure each one of those devices isn't pumping the user with shittons of radiation. There's a lot of boring research and regulation needed to make radio work at all, it's not something that can be hacked together in a weekend and collaborated on over time.
The real problem with RAND is that a lot of the licensing bodies that require it don't actually enforce that requirement all that well. For example, 4G and 5G cellular technology has a RAND requirement. If you want to put a 5G modem in a cellphone, there's a defined fee structure you pay that's the same deal everyone else gets. BUT, if you want to put a 5G modem in a laptop, that's not covered by the RAND promise, so Qualcomm reserves the right to charge whatever price it wants to each individual licensee, or even just not license it. This is why laptops with cellular modems in them are rare[6] and hideously expensive.
[0] For the record, most standards bodies do NOT have a means to do this to people who refuse royalty-free licensing, which is why ISO has failed to create a royalty-free video standard. Every time they try someone says "I own this, here's your royalty bearing RAND license", and the standard is dead in the water. In fact, not only did they fail to make the standard, they kicked Leonardo Chiariglione, co-founder of MPEG, out of MPEG for making too much of a fuss about it.
[1] Specifically as a reaction to the University of Minnesota trying to charge licensing fees for Gopher
[2] Which has resulted in a number of "ubiquitous" OpenGL extensions that aren't in the standard solely because they cost money to license, but can be accessed through de-facto standard extension names
[4] To be more accurate, AOM AV1 was only partially done by the community. Xiph.org contributed their work on Daala, but Google also included what would have become VP10. So there's still a lot of Google monopoly money in there. Furthermore, there's at least two companies claiming patent ownership on AV1 anyway, but I've yet to hear of any enforcement actions.
[5] Amateur as in ham radio, not amateur as in unskilled layman
[6] Apple doesn't sell a laptop with a cellular modem in it, and why the only Surface tablet Microsoft sells is the Qualcomm ARM version, not the x86 version. Qualcomm wants cellular capability to be a selling point of their ARM processors, so they (refuse to) license appropriately.
Thank you for your very thorough response to my rant, appreciate the insight - particularly the explanation why it is so rare to see mobile data modems in laptops (which to be honest leaves me even more enraged at the patent model).
However, I do disagree on some points:
> They pay the people doing research for the standards on a "for exposure" basis[3]: they get their patented inventions in the standard, and as payment for their work, they get to charge a $2/unit royalty on every user, which they can then pour into research that yields more patents.
The problem at the core IMHO is that Western nations removed themselves from R&D financing following the end of the Cold War, and instead large corporations and foreign nations moved in that had other interests in mind than the common good.
> Every time they try someone says "I own this, here's your royalty bearing RAND license", and the standard is dead in the water.
There used to be at least three competing patent pools for h264 and HEVC (until Velos Media disbanded). That even the largest commercial standardization enterprises like MPEG could not make sure that there was at least exactly one consolidated license shop just shows to me how nuts the entire patent situation has become.
> But at the same time, cellular radio technology is quite possibly the worst thing to leave to amateur[5] software development. There's a lot of very expensive lab tests you need to do to make sure the technology works, to ensure hundreds of devices can communicate with the towers without jamming each other, and to ensure each one of those devices isn't pumping the user with shittons of radiation. There's a lot of boring research and regulation needed to make radio work at all, it's not something that can be hacked together in a weekend and collaborated on over time.
Actually it can. Fabrice Bellard (yes, the same person as ffmpeg and qemu...) wrote an entire 4G/5G tower implementation that he currently sells via his own company, there's a FOSS project implementing the core services for a 5G network [2] and I think Osmocom has also made some progress in getting some sort of client side support.
Yes, the interop and jamming concerns are still valid, but those with enough brains to work in this field in the first place generally tend to know WTF they're doing - and commercial efforts definitely have the resources to properly work without endangering anyone.
The telling thing for me is that there is virtually only Qualcomm and Mediatek around who sell high performance modems, in addition to Samsung who mostly keep their stuff reserved for their own phones. Where's Apple? Where's Broadcom? Where's anyone to compete with Qualcomm's extortion tactics and Mediatek's IMHO embarassing code quality (just look at one of their Android BSP packages and the kernel patches they do)? For me, this state of the market shows that there is something seriously broken, and given that Apple has more cash on hand than entire countries' GDP and some of the most capable semiconductor engineers in the world managing to kick the arses of behemoths like Intel, the only explanation is the patent minefield.
And I'd like to address this as well:
> Which has resulted in a number of "ubiquitous" OpenGL extensions that aren't in the standard solely because they cost money to license, but can be accessed through de-facto standard extension names
Yet another very fine example how the patent system that was supposed to further innovation actually stifles it.
In the west, Qualcomm won. There's basically no one else even trying to make medium or high end chips. The patent encumbrance seems total.
Even the mega giant Apple is reportedly just giving up on trying. And at huge real world impact, as well as market drive. Apple designs tend to use a third or more of their phone motherboard's real estate supporting Qualcomm's forest of different parts required for cellular. It's been stubbornly hard to integrate well, forever.
The other IT field I'd cite is GPUs, where it's not at all a secret that companies would love to make more open hardware & stop using so many firmware blobs, but those blobs are the obfuscation layer that gets them legal protection, by making it unclear how the hardware works.
In these regards, it's just a miracle wifi and bluetooth and consumer GPUs exist. That anyone can build any chips, given our shitty onerous forsaken legal IP system, seems like a miracle. Qualcomm feels like the dark world hell nightmare we can never escape from but miraculously nightmare-shit-world is in only one key part of information technology, not all of IT. Somehow. Thank the stars.
I get technology is always advancing, but I do wonder what the landscape will look like in 10 years when a lot of these patents start to expire. As a consumer, it feels like we have been hitting diminishing returns for a while now.
Sure, they will have released 11G, and all of the ads will tell me how vital it is the government can track me with millimeter precision, but I think all of my needs were already met with a five+ year old device.
I have a lot of hope the open core movement is going to help, but right now it's still bonkers hard and expensive & special to attach a good modern bus to a chip. Having ram or USB3+ or PCIe or soon CXL - much less BT or wifi! - is a totally different kind of feat that open silicon seems far off from. The LiteX folks come to mind as deserving as big nod here but I think mainly fpga not basic targeting. A connected computer still feels so far off.
It's pretty interesting seeing some plateau forming commercially. The new Intel n100 - n30p chips look awesome. But they compete with mini-pc systems built around 20-65W desktop chips from 3-10 years ago that have a huge cheap second hand market. And they are a bit more power efficient but not vastly, producing similar power. But, good news, Intel & oems can have new systems using these chips on the market for competitive prices. Being able to compete with your own secondary market is a great sign to me.
It’s kind of an untold story, but I grew up near San Diego. My dad knew a bunch of early Qualcomm employees (in the first 50). Many of them had most of their equity clawed back through unethical legal agreements
Patents, inscrutable tools for working with low-level firmware, and a completely separate kernel line for each SoC based on an old kernel not even remotely upstreamed, ...
The chip is the most important part of a computer, but most chip companies do not capture much of the value created with the computer. An OEM pays about as much for an intel i3 chip as they pay of a windows license, and intel actually has to produce the chip.
This is why Apple has been so successful with Apple Silicon: Apple is able to capture way more of the value created with Apple devices, which is why they're able to invest more in chips then Intel/AMD and yes, even Qualcomm.
Oh, is Qualcomm the reason why adding cellular to an iPad is so expensive? I never understood how such a relatively small feature would cost that much. The alternative is that Apple just makes more money when setting prices like that. Would be curious how much of the price difference is extra license fees for Qualcomm.
I'm not sure why you're inclined to pin the blame on qualcomm, when the pricing is set by a company that charges $200 for 256GB of nvme storage[1]. For that price you can get a high end pcie 4.0 2TB nvme drive.
No, that's apple marketing, eg. apple upselling 32 or 64 GB phones or ipads to 256 or 512 etc gives them huge margins on the SDDs they resell by a $100s. If that was its own business it would be larger than 3M.
I question who this article is for - having worked at Qualcomm, they beat you over the head on how their business is built on patents and how they were going to go belly up some 20 years ago had they not pivoted to this model. I can't imagine a single analyst or anyone interested in the company not aware of their business model in the same way it doesn't take long to figure out that Apple is in the business of selling iPhones, mac books, etc.
I think this is valuable for those outside of Qualcomm to understand why they have the role that they do in the mobile space. Having worked there as well I have two observations:
- So much of that wall of patents was undergrad level CS concepts with ", for mobile device" tacked on the end. Paraphrasing, don't sue me (kidding..), but that was the gist of it.
- Their actual development process was pretty bad. I submitted a bug fix within the first few weeks I was there. Due to their CI process being entirely manual and very broken I was informed nearly a year later that my fix did not work. This tied in pretty well with my department being told they had a "budget"[1] of lines of code they could change and that budget being pretty limited.
Overall I left with an impression that they were mostly there to license very old technology and make money from patents. Maybe their hardware team was better, I couldn't say.
[1] Most of the effort of our team was spent debugging crash dumps determined to be in "our" part of the code. Nearly all of these were caused by other teams calling into an API we maintained with a null function pointer that "our" code would call. When I asked if we could refactor the API to return an error code if the function pointer was null I learned about the "code budget" and how it would cost too much to refactor.
About a decade ago I had to help with due diligence on a potential acquirer in the San Diego area. Qualcomm was the largest tech employer there and from the developers I talked to they seemed to have a terrible reputation at least among software folks.
I got one patent while I was a Qualcomm for some fairly inconsequential software thing. I was awarded $3k USD. Had the patent been filed in other countries and awarded, I would have been awarded I think 1.5k USD per country. This is/was the standard deal at Qualcomm when I was there (2013-2021). I would not be surprised if core researchers are awarded more for stronger and more applicable patents.
The headline doesn't cover the main thrust of the article, which is:
> "...like other semiconductor majors, the San Diego-based company also protects its inventions with intellectual property. But Qualcomm has a unique model at the heart of the way it uses those patents"
When was that? They had naming rights to a Major League Baseball & football stadium from ‘97 to 2017. Maybe good for recruitment but seemingly pointless on the consumer side to maintain the contract:
>Also in 1997, the facility was renamed Qualcomm Stadium after Qualcomm Corporation paid $18 million for the naming rights.[8] The naming rights belonged to Qualcomm until 2017, after which the rights were purchased by San Diego County Credit Union.
Surprisingly cheap! Not that I think it is a good use of money, but less ostentatious than I expected.
Readit News
Motorola makes 18% [https://valustox.com/MSI]
Broadcom, Nvidia, and even Texas Instruments are around 40%! [https://valustox.com/AVGO, https://valustox.com/NVDA, https://valustox.com/TXN].
Intel and AMD bring up the rear with 1.8% and 0.5%. [https://valustox.com/INTC, https://valustox.com/AMD]
For reference, Apple is at 25% and Tesla is at 11%. https://valustox.com/AAPL, https://valustox.com/TSLA]
https://en.m.wikipedia.org/wiki/GlobalFoundries
Through TSMC for example has margin >40%. Which isn't that surprising I mean there is not much competition. Intel doesn't (yet) provide their capabilities to anyone else, and the other competition is behind enough to not be an option for many things, especially high end and/or high efficiency.
What is quite fascinating is that while TSMC is a close to monopoly(/duopoly if you count Intel) in many sectors and takes advantage of it they seem to have successfully resisted the common problem of such companies stopping/slowing innovation and similar problem. I'm not sure if it's related to the additional pressure of them being on of the main lifelines for Taiwan or if they have just pretty good management or if it's because they still competition oligopoly main customers do push them to do so.
But all those factories have to be paid for no matter income.
So if income goes down your profit margin goes to terrible.
Intel has been burning their cash to try and catch up in both Foundry and CPU business.
Probably better if those % were Non-GAAP numbers. But that is another debate.
They’ve been betting the farm on delivering the highest performance chips on the market for the last 25 years.
For many years, they were unopposed technically. Now, though, their technological advantage is starting to erode.
Intel is in deep shit as a business. How much runway they have is anyone’s guess. But they’re in trouble.
And the decreasing return, in practical terms, for their customer base when upgrading since most basic computation tasks have long been solved and it's only exotic high-end workloads (plus constant bloat scaling factor that is largely a consequence of the added performance, rather than it's target) that benefit from added performance.
In short, was Intel making percent, or even single digit margins back in the 80s or even early '90s?
For Intel, it looks like they kept one dollar out of every four until the recent crash. I don't know the reason for the decline, but I know I wouldn't want to compete against Apple.
The other bit that blows me away is that sometimes it feels like Qualcomm's perpetual inability to integrate chips is a vast submarine point no one using Qualcomm sees coming, but which forever works out in Qualcomm's favor. I don't know how many chips a MediaTek cellular modem takes, but iPhone 15 pro has 6+ chips dedicated to cellular, probably more. Modem, two transceivers, an envelope tracker, and probably a bunch of in this case non-qualcomm power supplies but most probably use Qualcomm pmics, maybe plural. https://www.ifixit.com/Guide/iPhone+15+Pro+Max+Chip+ID/16532... . There's also separate but non-Qualcomm solutions for wifi & bt too, where-as again I rather expect a company like MediaTek has much more working together.
I'm really interested to see chip identifications of the upcoming desktop/server class Qualcomm Oryon, supposedly by the Nuvia team (one of the strongest possible cases for anti-trust I have ever seen). https://www.semiaccurate.com/2023/09/26/whats-going-on-with-... is not partial journalism, but this cavalcade of ridiculousness is entirely the sort of poorly integrated Qualcomm solution I'd expect, that both builds terrible terrible products but which also lets Qualcomm fail upwards by requiring 5x more board components than is reasonable. Terrible for consumers, OEMs are in tears, but oh sorry, your design is late hot over priced and seemed reasonable but you're going to have to add a bunch more pmic's to get to market because we are Qualcomm and we spend more on lawyers than design.
This is an interesting turn of phrase, but I don't understand what it means - are you able to explain? (Google is not helping)
The 20% margin this year was an anomaly. It’s usually closer to 29%+
The last thing this world needs is more Qualcomm. I'd rather see the entire patent system go down in flames than this utter crap continuing.
Qualcomm has been a vampire on technology for years now, it's time for this company to end once and for all. We have to get technology under the people's control again - it is ridiculous and stifling progress that governments legally prescribe standards that are not open to everyone to implement. And especially it's ridiculous how few notable manufacturers of mobile phone SoCs remain - it's either Qualcomm or Mediatek.
Standards-essential patents are required to be licensed this way. It's called "RAND licensing": reasonable and non-discriminatory. Everyone who participates in, say, ISO or the ITU, is required to fill out a form documenting if they own part of the standard and what terms they agree to license it on. If they refuse RAND licensing, their inventions get ripped out of the standard[0].
What you want is royalty-free RAND licensing. There are a few standards bodies that insist on royalty-free licensing, notably the W3C[1] and Khronos Group[2], but it's uncommon. Most standards bodies are in fields of endeavor that don't actually require free redistribution - in fact, they don't want it. They pay the people doing research for the standards on a "for exposure" basis[3]: they get their patented inventions in the standard, and as payment for their work, they get to charge a $2/unit royalty on every user, which they can then pour into research that yields more patents.
I'm going to be honest, royalty-bearing RAND (as in, pay up or don't implement) is probably the right fit for a lot of industries. We hate it because it steps on our (FOSS) toes, and if we can develop the software as a community, why can't we invent the underlying technology too? This is the idea behind royalty-free video codecs like AV1 - it turns out a community of amateurs can do video codec research after all[4]. But at the same time, cellular radio technology is quite possibly the worst thing to leave to amateur[5] software development. There's a lot of very expensive lab tests you need to do to make sure the technology works, to ensure hundreds of devices can communicate with the towers without jamming each other, and to ensure each one of those devices isn't pumping the user with shittons of radiation. There's a lot of boring research and regulation needed to make radio work at all, it's not something that can be hacked together in a weekend and collaborated on over time.
The real problem with RAND is that a lot of the licensing bodies that require it don't actually enforce that requirement all that well. For example, 4G and 5G cellular technology has a RAND requirement. If you want to put a 5G modem in a cellphone, there's a defined fee structure you pay that's the same deal everyone else gets. BUT, if you want to put a 5G modem in a laptop, that's not covered by the RAND promise, so Qualcomm reserves the right to charge whatever price it wants to each individual licensee, or even just not license it. This is why laptops with cellular modems in them are rare[6] and hideously expensive.
[0] For the record, most standards bodies do NOT have a means to do this to people who refuse royalty-free licensing, which is why ISO has failed to create a royalty-free video standard. Every time they try someone says "I own this, here's your royalty bearing RAND license", and the standard is dead in the water. In fact, not only did they fail to make the standard, they kicked Leonardo Chiariglione, co-founder of MPEG, out of MPEG for making too much of a fuss about it.
[1] Specifically as a reaction to the University of Minnesota trying to charge licensing fees for Gopher
[2] Which has resulted in a number of "ubiquitous" OpenGL extensions that aren't in the standard solely because they cost money to license, but can be accessed through de-facto standard extension names
[3] https://blog.chiariglione.org/a-crisis-the-causes-and-a-solu...
[4] To be more accurate, AOM AV1 was only partially done by the community. Xiph.org contributed their work on Daala, but Google also included what would have become VP10. So there's still a lot of Google monopoly money in there. Furthermore, there's at least two companies claiming patent ownership on AV1 anyway, but I've yet to hear of any enforcement actions.
[5] Amateur as in ham radio, not amateur as in unskilled layman
[6] Apple doesn't sell a laptop with a cellular modem in it, and why the only Surface tablet Microsoft sells is the Qualcomm ARM version, not the x86 version. Qualcomm wants cellular capability to be a selling point of their ARM processors, so they (refuse to) license appropriately.
However, I do disagree on some points:
> They pay the people doing research for the standards on a "for exposure" basis[3]: they get their patented inventions in the standard, and as payment for their work, they get to charge a $2/unit royalty on every user, which they can then pour into research that yields more patents.
The problem at the core IMHO is that Western nations removed themselves from R&D financing following the end of the Cold War, and instead large corporations and foreign nations moved in that had other interests in mind than the common good.
> Every time they try someone says "I own this, here's your royalty bearing RAND license", and the standard is dead in the water.
There used to be at least three competing patent pools for h264 and HEVC (until Velos Media disbanded). That even the largest commercial standardization enterprises like MPEG could not make sure that there was at least exactly one consolidated license shop just shows to me how nuts the entire patent situation has become.
> But at the same time, cellular radio technology is quite possibly the worst thing to leave to amateur[5] software development. There's a lot of very expensive lab tests you need to do to make sure the technology works, to ensure hundreds of devices can communicate with the towers without jamming each other, and to ensure each one of those devices isn't pumping the user with shittons of radiation. There's a lot of boring research and regulation needed to make radio work at all, it's not something that can be hacked together in a weekend and collaborated on over time.
Actually it can. Fabrice Bellard (yes, the same person as ffmpeg and qemu...) wrote an entire 4G/5G tower implementation that he currently sells via his own company, there's a FOSS project implementing the core services for a 5G network [2] and I think Osmocom has also made some progress in getting some sort of client side support.
Yes, the interop and jamming concerns are still valid, but those with enough brains to work in this field in the first place generally tend to know WTF they're doing - and commercial efforts definitely have the resources to properly work without endangering anyone.
The telling thing for me is that there is virtually only Qualcomm and Mediatek around who sell high performance modems, in addition to Samsung who mostly keep their stuff reserved for their own phones. Where's Apple? Where's Broadcom? Where's anyone to compete with Qualcomm's extortion tactics and Mediatek's IMHO embarassing code quality (just look at one of their Android BSP packages and the kernel patches they do)? For me, this state of the market shows that there is something seriously broken, and given that Apple has more cash on hand than entire countries' GDP and some of the most capable semiconductor engineers in the world managing to kick the arses of behemoths like Intel, the only explanation is the patent minefield.
And I'd like to address this as well:
> Which has resulted in a number of "ubiquitous" OpenGL extensions that aren't in the standard solely because they cost money to license, but can be accessed through de-facto standard extension names
Yet another very fine example how the patent system that was supposed to further innovation actually stifles it.
The system is completely broken.
[1] https://bellard.org/lte/
[2] https://open5gs.org/
Deleted Comment
Even the mega giant Apple is reportedly just giving up on trying. And at huge real world impact, as well as market drive. Apple designs tend to use a third or more of their phone motherboard's real estate supporting Qualcomm's forest of different parts required for cellular. It's been stubbornly hard to integrate well, forever.
The other IT field I'd cite is GPUs, where it's not at all a secret that companies would love to make more open hardware & stop using so many firmware blobs, but those blobs are the obfuscation layer that gets them legal protection, by making it unclear how the hardware works.
In these regards, it's just a miracle wifi and bluetooth and consumer GPUs exist. That anyone can build any chips, given our shitty onerous forsaken legal IP system, seems like a miracle. Qualcomm feels like the dark world hell nightmare we can never escape from but miraculously nightmare-shit-world is in only one key part of information technology, not all of IT. Somehow. Thank the stars.
Sure, they will have released 11G, and all of the ads will tell me how vital it is the government can track me with millimeter precision, but I think all of my needs were already met with a five+ year old device.
It's pretty interesting seeing some plateau forming commercially. The new Intel n100 - n30p chips look awesome. But they compete with mini-pc systems built around 20-65W desktop chips from 3-10 years ago that have a huge cheap second hand market. And they are a bit more power efficient but not vastly, producing similar power. But, good news, Intel & oems can have new systems using these chips on the market for competitive prices. Being able to compete with your own secondary market is a great sign to me.
This is why Apple has been so successful with Apple Silicon: Apple is able to capture way more of the value created with Apple devices, which is why they're able to invest more in chips then Intel/AMD and yes, even Qualcomm.
[1] price difference between an mac mini 256GB vs 512GB: https://www.apple.com/shop/buy-mac/mac-mini/apple-m2-chip-wi...
- So much of that wall of patents was undergrad level CS concepts with ", for mobile device" tacked on the end. Paraphrasing, don't sue me (kidding..), but that was the gist of it.
- Their actual development process was pretty bad. I submitted a bug fix within the first few weeks I was there. Due to their CI process being entirely manual and very broken I was informed nearly a year later that my fix did not work. This tied in pretty well with my department being told they had a "budget"[1] of lines of code they could change and that budget being pretty limited.
Overall I left with an impression that they were mostly there to license very old technology and make money from patents. Maybe their hardware team was better, I couldn't say.
[1] Most of the effort of our team was spent debugging crash dumps determined to be in "our" part of the code. Nearly all of these were caused by other teams calling into an API we maintained with a null function pointer that "our" code would call. When I asked if we could refactor the API to return an error code if the function pointer was null I learned about the "code budget" and how it would cost too much to refactor.
Deleted Comment
Seems like the message behind 'this is core to our business' is easily communicated with financial incentives backing that up.
> "...like other semiconductor majors, the San Diego-based company also protects its inventions with intellectual property. But Qualcomm has a unique model at the heart of the way it uses those patents"
The article dives into that model.
https://en.m.wikipedia.org/wiki/San_Diego_Stadium
But I get it that in sports, sometimes the big cheques are written because you can.
Surprisingly cheap! Not that I think it is a good use of money, but less ostentatious than I expected.
Deleted Comment