The answer is super complex and many factors play a role in the technical bits - if we ignore the commercial bits.
DSL lines are capacitively coupled to ground over their entire run length. This limits the frequency at which the line will operate reliably (a typical DSL modem will perform a whole series of tests to figure out which bands are 'safe' to use).
Other factors are dodgy connections in the line, other DSL lines in the same trunk cable causing interference, crappy termination causing impedance mis-match between the transmission lines and the driver hardware and limited usable bandwidth for a carrier wave to superimpose your data on top of. The smaller the available frequency band(s) the less data you can send.
Think of 'POTS' (the Plain Old Telephone Service) wiring as just about the worst possible means of transporting data that you could find. All things considered that DSL works at all is a small miracle and it took a good bit of engineering and some changes to the infrastructure (mostly the removal of the boost coils in the lines) to allow it to work.
LTE ('air' as you call it) has a huge advantage. It starts off as a GHz range carrier signal with modulation bands up to 200MHz wide (but more typically between 20 and 60MHz). That's an awful lot of bandwidth. The lack of connectivity is made up for by having a fairly dense network of base stations.
Of course you need to share those base stations with everybody else but as long as it is just you you can see phenomenal speeds, far in excess of what your DSL line is capable of.
If your DSL line could use carrier frequencies in the 100's of MHz without losing definition then DSL could be just as fast. Unfortunately the physical medium is crap and there isn't much to be done about that short of replacing it with coax (and that's called cable, and it's faster than DSL if you are on your own but - again - you will share that run of coax with a lot of your neighbours).
Indeed, brings to mind DSL over wet string[1] that was posted here a while back[2].
[1] https://www.revk.uk/2017/12/its-official-adsl-works-over-wet...
Real answer: money
DSL speed has an inverse relationship with wire length. The marginal benefit of VDSL over ASDL is almost entirely gone at 1km distance. Telcos that care about offering competitive speeds will install more VDSL hubs in neighborhoods to keep average length short. Or better yet they will install fiber. Telcos that rely on monopoly status to stay in business or think wireless is a more productive use of capital will leave DSL to rot.
LTE service usually comes with very low usage quotas and punitive overage fees or throttling. 150 Mbps burst speed is very different from 150 Mbps service you can freely use. Several thousand users have to share that 150 Mbps!
This is my understanding at least, I am open to correction/ nuance.
And it is also a matter of investment. The 4G, Smartphone evolution we are talking about 10s of billions of monthly recurring revenue worldwide, and a % of those are going into making faster 4G, more capacity, and 5G. I think we sort of forget 10 years ago when the iPhone first came out, No one was really using 3G ( Apart from Japan ). We literally went from End of 2G era to start of 5G in mere 10 years! Because consumers are willing to pay more for faster network access, better phones, better modem chips. You would have been told you are insane if you tell anyone in 10 years ago ~300M people on earth will spend average $750+ on a phone, EVERY YEAR.
A lot of these 4G / 5G tech, innovation are actually moving to other areas. WiFi upcoming 802.11ax is like mini LTE, the G.Fast for DSL and DOCSIS 3.1 for Cable also take many learnings from 4G.
As a matter of fact, someone is trying to use wireless signals in the tiny space between twisted cable wiring to get you Terabit DSL.
Restrictive data transfer quotas and the proliferation of free Wifi access make existing 4G+ implementations look fast.
Those incentives being a kickback/royalty payment to Wi-Fi operators who allow roaming and cheaper bills and higher quotas for customers. Right now I have a choice between $10/GB fast LTE and free but very slow store Wi-Fi. Free means the venue operator spends as little money as possible and users over-consume. A revenue stream for accepting off-loaded traffic would give them a business justification to subscribe to a decent connection and install additional access points.
To this day it seems like only the equipment manufacturers are profiting off Wi-Fi. Venue owners only indirectly profit by attracting more foot traffic. Cellular operators plug coverage holes with Wi-Fi calling, and address severely congested areas with efforts like attwifi, but they do not want to disrupt their main product.
I would love to either have a vpn that connects over 4g and wifi and seamlessly redirects traffic depending on conditions, or have widespread support for something like multipath-tcp that supposedly fixes this issue.
2. WiFi / Unlicensed Spectrum has limited power compared to licensed Spectrum. Which means full coverage is few order of magnitude harder.
3. I still think weak signal and hand off is an software and implementation issues. We should just disconnect WiFi if its signal is only 2 bars and not try to do edge location transfer.
The upcoming 802.11ax fixes that somewhat, and it should be able to handle 32+ concurrent users, and with up 8x8 antenna in phase 1 ( Not confirmed yet ) and hopefully 16x16 in phase 2. I am not sure if Phase 1 and Phase 2 are compatible in such a way so Phase 2 router would benefits Phase 1 users as well. After-all these manufacture are there to make you spend more money to upgrade. Another problem is all these are good on paper, but once 802.11ax are mixed with 802.11ac, its performance and capacity drops dramatically. So I actually wish in the far future we force more users on ax, drop ac from 5Ghz band and users uses N on the 2.4Ghz band.
Or better yet, LAA, ( License Assisted Access ). While technically superior I am not entirely comfortable of putting everything in carrier's hand. And Full LTE in Unlicensed Spectrum; MulteFire, is an entirely Qualcomm tech.
The provider A1 in Austria (around 45% marketshare for broadband) uses ADSL for most of its broadband lines, which is significantly slower than cable. A few years ago they merged with the largest wireless provider and now use hybrid-ADSL-LTE modems that use ADSL by default, but which automatically switch to LTE when higher bandwidths are required.
In addition there are several wireless providers (and MVNOs) that sell data flatrates which are cheaper and faster than ADSL. E.g., I recently replaced my ADSL connection with a LTE flatrate for 20 EUR a month that allows 20mbit down and 6mbit up (and almost all of the time it also reaches that speed).
I often hear the argument that LTE would be too congested if everyone uses it. But I can't see any evidence of that in places where LTE is heavily used.
Jealous! Here in Australia I'm paying A$70/month (~44 EUR) for 12/1 4G home broadband service. Still better than the 3/0.5 service I was getting with ADSL for the same price...
It can't. Coax can easily support Gbps connections (I used to have a 800 Mbps connection where the last few meters were over regular coax).
Now if you're talking about "why can't my copper twisted pair support that bandwidth", then it is a function of quality (the twisted pair is likely old, heavily bended, subject to interference, and running over old technologies.
Bandwidth is essentially a function of 2 things: spectrum availability (measured in Hz) and spectral efficiency (measured in bits/second/Hz).
For example, if a carrier "owns" 20 MHz of spectrum (let's say, from 700 MHz to 720 MHz), split in two symmetrical up/down channels, with a spectral efficiency of 4bps/Hz (a common value for LTE), and the cell towers have 3 sectors, each cell tower can provide 2043/2 120 Mbps up and down (although only 40 Mbps max per device, since a device can only be in one sector).
In wired communications, spectrum is less of a limit: every single wire has the full spectrum available. And if you are using fiber, then there is no interference, both from outside the wire or from different signals.
So air will never beat wired, because every single strand of wire has, in practice, the entire capacity of air.
PS: There are more things in play, such as attenuation, actual frequencies used (1 to 2 MHz has 1 MHz of spectrum; 1 to 2 GHz has 1000x more spectrum and therefore more capacity), etc.
[0] The speed of light, but not the speed of light.
I think you accidentally made a formatting typo: "each cell tower can provide 2043/2 120 Mbps", the 4 is probably surrounded by asterisks.
It should be 20 * 4 * 3 / 2 = 120