I disagree, lol.
There are still companies to this day with well over 200 employees, who use Windows XP.
The sheer inconceivable level of studpidity and wrecklessness when it comes to networking and server equipment is the single most common reason why respectable and known businesses are routinely compromised. Leaks, ransom and moles are a direct result of inadequate, cheap or incorrectly configured networking equipment and maintenance.
This is a very USian view on things. I am not trying to defend China* here by the way, but it is not that for a non-Chinese, non-US person either choice is all that attractive**. The argument about quality is something I do not share, you can produce high-quality products everywhere, and the same holds for low-quality products as well, just being manufactured in XXX does not guarantee high quality by itself.
*) China and Taiwan are clearly not the same entity and work by different rules, I do not think manufacturing in both countries is all that similar.
**) I am in no way impying the US are anywhere close to China in most dimensions here, but from a foreign perspective manufacturing clearly should be happening in the respective foreign country.
But this is pretty much a digression and a completely different line of argument than that that started this thread, no?
With lots of different twigs, like people optimizing for raw performance, for low energy consumption, or for performance/power.
As router OS? Hard to believe that this is a sizable fraction of companies with >= 200 employees...
For workstation use there is occasionally decent justification, like expensive but still useful legacy software that somehow requires XP, or where an OS upgrade would require additional investment in additional software/hardware updates. But again, that seems not to support your assertion that consumer routers/switches* have been used extensively in the past?
Glad you have this figured out as well as what kind of routers we all should be using. 
Ever heard of Zero-Day exploits? Just asking....
Of course one can produce high quality parts at a reasonable cost anywhere, but the obvious reason why chip/hardware manufacturers and any other major industry assemble their products in China is because they intend to achieve the opposite.
There are many Chinese and Taiwanese companies who manufacture quality hardware in an attempt to dispel that stigma. Asrock for example or Donghe.
However, the general rule of thumb still holds true. The vast majority of routers, chips and computer hardware in general is manufactured in China or Taiwan to drastically lower cost and increase profit margins. That is why the quality and longevity of those parts is usually subpar and consumers are forced to replace them or buy accessories compensate for lack of performance.
That is why the quality and longevity of those parts is usually subpar and consumers are forced to replace them or buy accessories compensate for lack of performance. Supporting this model of corporate greed, and human rights violations by purchasing a phone or router for the benefit of saving 50$ has more negavtive aspects than just a bad product.
That is not correct, this move is motivated by minimizing costs, some reduction in quality might be tolerated, but is not the primary goal. This is getting a bit off-topic though.
Heard of TMSC? They happen to produce some of the most modern and performant semi-conductors on the planet with customers including, Apple, AMD, and even Intel. As I said Taiwan is not China and not differentiating these somewhat devaluates your argument. Also I doubt that Taiwanese manufacturers aiming for high quality do so motivated by fear of a reputation for low-quality.
Yes, welcome to late-stage capitalism... not something we are going to change by all OpenWrt users switching to boutique routers manufactured in their respective home-countries.
There are plenty of low-quality products produced elsewhere. Also this has little bearing on this thread, as quite a number of your despised cheapish all-in-one routers work quite well for their owners for a long time. Few people I know personally change their internet access speed all that often, hence their performance requirements are a pretty static target, that can often be squelched by economically attractive plastic routers, that than will be operated for a long time (if the performance is sufficient for the task now, it will be so in a year, unless the task or the requirements what to perform change). Only accessories I have heard of are mostly WiFI mesh solutions that are add-ons even if people would start out with a juniper MX-960 as their home router, to pick something "professional" and ridiculously over-powered.
It is really really hard to buy competitive computing equipment that is not (at least partially) based on "corporate greed"*. Turning your moral black-white argument into a different-levels-of-gray argument, much less clear to interpret.
But didn't the whole thing start by an argument about performance/price/and energy consumption, why are we wading deep in a morality argument now?
*) Because almost all large enough companies operate like that.
Go look at the planes Boeing is assembling in South Carolina, then come back saying production in the US equals high quality standards...
Think it’s time for this thread to be locked as it’s going WAY off topic
Oh, drastically reducing cost is not the primary reason why various industries have built massive manufacturing plants with anti-suicide nets in China?
Then what is that reason?
How so? It's true that Taiwan is the world's largest manufacturer of semi-conducters, but it is also true that Taiwan faces an equally alarming amount of human rights violations and systemic labor exploitation of economically or socially disenfranchised minorities such as various indigenous populations and minors.
Taiwan has the same purposely ineffective legislation in those areas as China, in an obvious attempt to both fuel the economy by luring western industry with the promise of cheap labor under sub-human conditions and stealing their intellectual property in the process. This type of coordinated and willful economic exploitation of a country's own citizens has been occurring for centuries in many, previously developing countries, it's not just China and Taiwan, but they are the most prominent examples for the tech sector.
Read this article if you would like to educate yourself on the matter. Intel conducted a study, ranking tech corporations and their labor practices. https://www.newsgram.com/general/2018/06/19/intel-becomes-savior-of-exploited-workers
That was your claim, your sentence seemed to imply companies where intending to "achieve the opposite" of "high quality", which as we both agree was not the primary goal. (Not that the opposite of "reasonable cost" would make more sense as a sentence, unless you argue production cost in China is unreasonably low).
Well, cost as we apparently agree. I was a bit puzzled when you seemed to claim something else above.
Citation needed. But Even if true hat still does not make Taiwan and China equivalent, the world is not black and white so we need to actually look at the different levels of gray, otherwise we risk generalizing too much.
I am still not convinced that this is actually fully true for Taiwan.
You realize the list that is based on is per company for a selected list of companies, not aggregated (and not aggregatable) per country? Not sure you can robustly and reliably extract a country-ranking from that data.
But all of this is pretty far away from the performance/power/cost argument with which you started the thread.
Coming back to your original question, here is my (probably not entirely informed, very badly shortened, thus probably not 100% correct) argumentation again x86/amd64 routers, that adds on-top of the power usage arguement:
I'm not sure, that the "possible trust factor" is higher on x86,amd64.
My (perhaps not entirely correct) impression of things:
This, to me, translates to: "This platform is easier to audit, because it's hard/impossible? to hide [running] code/implants on it". (For better or worse) (Of course, this makes all those devices susceptible to evil-maid attacks.)
Some ARM-platforms (some Rockchips I think) come with sources to the ATF, and other "trusted" parts, and can thus be verified to be "kosher" (AFAIU). I'm not aware of any other ARM-verndor/product where this is the case, unfortunately. There may be other reasons for this, but -to me- the major 2 probably are "We are ashamed of the code-quality", and/or "We have stuff in there, that we don't want you to see."
On top of CPU-level vulnerabilities/problems you have chipset/mainboard issues like "Can my USB3-controller be used as turing-complete general purpose CPU, if you talk to it in the right way?" What about the DMA-controller, South/East/Northbridge, or any number of auxiliary CPU contained in e.g. the Wifi-Card.
Another facet is "chipset level bugs". Intel especially these last few years have shown, that while their stuff runs fast, they -like all humans- make mistakes, that considerably weaken system security, or performance.
So, in summary of my -less coherent than i'd like- rant: To me, there is no system/platform where I trust 100% for all built-in components, that this system is bug and back-door free, and will remain so, if I update my software regularly (from trusted sources)
I perceive a "repurposed PC" to have a much higher number of "dark corners" and other places to hide an implant/code. Probably hundreds of people of this forum could hide a persistent trojan in e.g. the DVD-drive firmware of that old PC.
On top of bugs/plausibly_deniable_backdoors there is the Hardware/Software in-place by law in many components to support "Lawful interception". I would expect until-proven-otherwise, that most (Intel, broadcom, huawei, etc.) network cards, either through (chipset-)FW, or implemented directly in HW, to have in-band debug/data-exfil-capabilities, if you send the right payload in a packet, and then some byte-code for some execution environment.
So, I may be wrong and/or misled by fake news, but I consider therefore an "old repurposed pc/pc_based_router" to the be the most untrustworthy option among the less untrustworthy alternatives.
I'm happy be to corrected on any point, or clarify thoughts, if written unclearly.
Your point about microcode issues and possible other vulnerabilities with older x86/x64 systems is valid, had to shoot you a 
for that.
Using outdated hardware is always a risk, especially on those platforms, I thought about this too the other day. You can load the typical non-free esque firmware packages in OpenWrt without issues, thankfully. On top I flashed the latest available BIOS from (2017!) on the MB. Apart from that you can't really do much but have an extremely tight firewall and not connecting any physical storage media that might be suspicious.
wow, i for one use to stand in the skeptical's group, encouraging people to challenge marketing bullshit and co, but you are on a different level. saying x86 is more untrustworthy than all-in-one platforms is a bit harsh when continuously we can hear about extreme stupid and vulnerable firmware issues, zero security with default passwords and backdoors in all-in-one device space (just one example CVE-2021-41653)
lawful intercept is a fact, all governments are doing, and i don't see why you think it is only related to x86 platform. not to mention, question is whether agencies can/want really intercept and process all collected information; or rather focus on specific threats(=whatever agencies specify as threat); therefore are they really want to analyses what i and millions of similar nobody's is doing on internet. how do you answer this question is up to you.
security is not just issue of x86, it is a generic problem. any end-of-support / end-of-life stuff is vulnerable, that's a fact, and platform independent. and this where all-in-one devices are even worth with their very short support and very quick sunset policies (planned obsolescence).
back to original topic: the OP was about why not everybody using x86 setup, simple answer is: because people are not one single mass. we have different needs, different resources (time, money, space etc) to invest into a network solution. OP said old x86 setup is better than any enterprise grade router - well, if you have had the chance to work with enterprise grade router you'd be not saying this. must be a reason why enterprises are using enterprise grade stuff and not old left over x86 parts.
therefore i am very glad openwrt community is very diverse and developers are covering large scale of different hw, and owrt is greatly scalable to provide very different functions. i would be very sad if it would be only just x86, or only capable of doing few things somebody forcing in name of all people.
but whether i am using x86 router? yes, i do. because for my needs it proved the best option as a router (+docker host). do i have all-in-one device? yes, i do, because as dumb AP and switch (a.k.a multiple functions in one box) they are pretty good.
anyhow, this topic turned into a political rant, so please feel free to argue with my post, but i am not going to add more questionable value comment from my side in this area, but would rather stay on technical ground and suggest the same to everyone.
thanks for reading. peace.
As a rule of thumb, for those who are paying around 10 to 14 cents per kWh of electricity, like many of us in the U.S., leaving a 1 watt appliance on all the time costs about 1 $US per year.
In other words, since I'm in this circumstance, I look at the wattage (actual average usage, not max peak) of appliances that I leave on all the time, and figure that's about how many $US per year it will cost me to leave that appliance (or light or whatever) always on.
This rule of thumb comes directly from your observation that "A watt-year is 8.76 kWh", and the simple calculation that (1 / 8.76) == 0.114, which happens to be near the price, in $US per kWh, that many of us here in the U.S. pay for electricity.
Given that most people live in large cities, I venture a guess that 75% of US population pays almost twice as much >20c/kW (electricity+delivery).
Edit: well, it looks like more than 50%, some pay over 40c/kW !
Personally, I pay enough that it is worth to me to shutdown part of my network at night and I am considering moving Home assistant to openWRT router so I can shutdown proxmox for a night as well.
Well, not anymore my friend. 
With the skyrocketing of gas prices, electricity has gradually been shooting up as well. In southern states there has been a steady increase of 11.5% with an upwards tendency.
It's true, electricity used to cost us nothing, but those times are over it seems.
Ehhh... even thats kinda high in power consumption.
Yeah the Raspberry Pi is an interesting choice but they are either sold out everywhere or cost double what they're worth. Buying parts on ebay has always been the best route imo.
On an x64 system you could also run pfsense or opnsense. Anybody tried them before?
Yup. The average price per kilowatt hour for August where I am was roughly 0.45 USD. That average conceals a large diurnal variation, so the price is high for the hours of the day where there is peak usage, and low in the dead of night. The power usage is monitored by 'smart meters', so a kilowatt used between 10 a.m. and 11.a.m could easily cost 50% more.
That translates into just over 4 USD per watt year, so I'm prepared to pay a premium for capable low-power draw devices.
Just switched my ISP, from 1Gbit/s fiber to 1Gbit/s Ethernet one.
One of the reasons: old ISP provided glorified fiber-to-ethernet box was consuming 6W, no matter how loaded.
At 50+ eurocents per KWh, it's like 26 EUR per year spent heating your room, pushing light rays to copper.
Over the last 2 months:
You need to look for RPIs in the right places: https://rpilocator.com/
x86 as some people might have pointed out already, is not energy efficient.
I've also used a (old Pentium3) x86 computer back in late 2000s, until i got tired of the energy consumption and also moved out of my parents, and looked for something good enough, and found the TP-Link Archer C7, then on revision 2, around 2010-2012.
Nowadays though, you can get awesome ARM systems to do these tasks or other things, for a much smaller energy footprint.
On point. I have a decent 3 antenna WiFi ac card on my desktop. It's rare to find something similar, and on a good price.
@ally
here in Portugal, standalone fiber internet only:
standard on the last years has been more 100/30 or 100/100 and 200/100 or 200/200Mbps, but this changes often
As already pointed out, this belief is incorrect - you can very well get modern x86_64 devices which chug under 10 watts out of the wall (down to roughly half of that) in practice.
In those hardware generations, 120-130 watts idle power consumption where indeed rather common, but the times have changed. Sandy-bridge brought that down to ~75 watts, ivy-bridge to ~30 watts, since haswell you can go down to 11-15 watts. On the Atom side, ~20 watts were possible with the venerable N270, since baytrail-d (j1900) you can achieve 6 watts (and maybe even less).
ARM is not a guarantee for low power, I have seen ARMv8 based wireless routers using around 6 watts, but others that idle around the 25 watt mark - it all depends on the hardware, the SOC, the kind of wired- and wireless capabilities.
Yes, an all-in-one 'traditional' plastic router has every opportunity to win against a system of x86_64 router, managed switch and AP(s) in terms of power consumption, but the delta doesn't need to be significant - and once you cross the border of 4+1 ethernet ports and a single AP, it might disappear altogether. Towards the upper end of the WAN speed scale, ARM based routers have a hard time to compete with x86_64 in terms of routing throughput and their ability to perform traffic shaping (sqm), x86_64 just has a lot more performance in the back (at least in regards to systems you can actually purchase as mere mortal). Tips for getting cheap used x86-based firewall with full Gbit NAT (a PC Engines APU) if you are in the US has a number of low-power examples that can compete quite well on all fronts, once you detach yourself from the dated PC engines designs in the title (e.g. sophos, gateprotect, etc.; we are talking about 6-15 watts idle depending on the actual model and four 1000BASE-T ports) - even compared against RPi4 or NanoPi r4s/ r5s setups (which also need switches and APs in addition, rendering the power consumption deltas quite negligible).