8 ms·

Boom, Bubble, Bust: The Fiber Optic Mania [pdf]

68 points by tim_sw 3y ago

ezekielmudd 3y ago
FYI, this article is from 2016 and discusses the dot com bubble that popped around 2001. Jeff Hecht is an excellent author.
anonylizard 3y ago
Instant analogy to GPUs... A similar story is found in the railway bubble bursts in the US. Railways are also incredibly revolutionary technologies that reshaped societies, doesn't mean they are immune from investment cycles. That being said, given ChatGPT was only released 8 months ago, we are probably very early in the GPU boom, far from bubble yet.
- hospitalJail 3y ago
  Computing seems pretty important. I have a usecase that can save our company millions of dollars, but unless my company is willing to spend hundreds of thousands on a computer, its not happening. (It may not work, and prototypes still have me conflicted if it will work). We also can't rent servers for this, too confidential. If we can get prices down 10x-100x, I can even do this for personal/home usecases. I know that sounds extreme/absurd, but given our office is nearly empty with WFH, we can use larger computers and no one would mind.
KirillPanov 3y ago
Yeah the phone companies laid those thick 144-strand bundles before wave division multiplexing. They figured the only way to get more bandwidth was more fibers, right? WDM came along and caught them with their pants down; now all of a sudden you could just attach different devices to the ends of the fiber and get 16x the bandwidth. No backhoes needed. Then CWDM let you split and combine channels in locations without a reliable power supply -- no electronics, just prisms! Unfortunately their response -- those who weren't killed off -- was to get the antitrust regulators to look the other way (i'm sure the "national security" card was played) while they bought each other up and carved countries into a checkerboard of regional monopolies. But hey, if you squint at it from far enough away you can see two colors, so that's not a monopoly, right?
- jacquesm 3y ago
  In the meantime though we've started using a substantial fraction of those thick fiber bundles and dark fiber is getting harder and harder to find. Barring another revolution in technology I suspect that at some point we'll be laying even more of those big bundles, or even bigger ones.
  - KirillPanov 3y ago
    > dark fiber is getting harder and harder to find Only because all of it along any given route now tends to be owned by a single entity. And that entity can simply say "not for sale". If the market were healthy dark fiber prices would skyrocket instead of availability plummeting.
- liotier 3y ago
  WDM guy here. Sure, 80 channels are space for lots of 400 Gb/s circuits... But we still fill cables and keep laying more. Though yes, the big 1288 fibers cables aren't common. Also, lambdas don't terminate at every node - so sparse network topology implies that filling all channels on a single link is a merely theoretical situation.
- rasz 3y ago
  >before wave division multiplexing. ... WDM came along Would that be Ciena in 1996?
- senttoschool 3y ago
  >Yeah the phone companies laid those thick 144-strand bundles before wave division multiplexing. They figured the only way to get more bandwidth was more fibers, right? I don't know anything about fiber tech but can't they just apply the newer techniques to the already laid out 144 strand bundles?
samuelkazeem 3y ago
>The investors who threw money at optics at the peak of the bubble probably would bankrupt us all if we ever hit a true “technological singularity.” If we really hit technological singularity, that will be the hail Mary of all bubbles.
- somat 3y ago
  conversely, I think we have been in the singularity since the late 1700's. For many thousands of years human manufacturing capability was effectively flat, a few blips and bumps but the advancement was generally steady and linear. And then in the late 1700's things changed. The graph went hockey stick and has stayed exponential since. However from inside this bubble there is a sort of tech dilatation where everything appears normal and flat and it is only the future that has exponential growth.
  - libealistand 3y ago
    Flat? I don't think so. I'm at a coffee shop and looking around me, this is very different from just 20 years ago. People with laptops and smaetphones around me. Just 20 years ago, that same place saw a few CS major nerds with their laptops, everybody else sat with a book or some paper to write on or reading a paper newspaper. Still with a latte, but tech has changed everybody's life dramatically in 20 years. 20 years from now I'm again expecting at least as big a change. Definitely exponential.
    - rcxdude 3y ago
      You're agreeing with the comment you replied to. they were saying it was flat until ~1700 or so.
      - libealistand 3y ago
        They wrote that it "appears flat". I tried to contradict that. > However from inside this bubble there is a sort of tech dilatation where everything appears normal and flat and it is only the future that has exponential growth.
libealistand 3y ago
I don't get the drama. To me this looks like a normal process in an industry that discovered a groundbreaking technology. A new technology pops up, everybody rushes to innovate, lots of actors do nonsense, some do the right thing, but for many of them you don't know beforehand which is which. So you throw money at all of them. The rush creates some really good ideas and innovations. And a lot of BS and wrong decisions. Within a few years, the successes emerge, everybody else is flushed out of the market. Today fiber is the network tech. Except for the last mile, everything is fiber. We're at 400G per port, soon 800G, and the market keeps growing and 1.6T is the obvious next step. Normal growth process. So, I don't really see the takeaway here. What to learn from it? Not to invest in hypes? That's not really how we grow.
- mangeld 3y ago
  Depends on the country, where I live it's pretty common to have fiber to the home. Everything runs on that, internet, cable tv and phone calls.
  - sandworm101 3y ago
    But in other areas most residential bandwidth increasingly comes from wireless techs. Even in homes with wired service, nearly everyone uses wifi rather than wires, let alone fiber to end machines. The last mile is generally not fiber, nor is it a literal mile.
    - cmrdporcupine 3y ago
      Indeed many people, including myself are now in the weird backwards world where the WAN is far faster than the LAN. I have a 3gbps fiber link straight to my house. The only machines in the house that can use anything close to it are the ones that hook up to my USB-C docking station which has a 2.5gbps ethernet port. Everything else is WiFi speeds. EDIT: This was a sudden change. I live rural and for the last 10 years I only had 10-15mbps speeds top, and most recently only by point to point wireless from a tower 5km away.
secretsatan 3y ago
I was involved in a UK company that strung fibre optics along the high voltage transmission network to quickly build a national network, there was a ridiculous amount of money flowing around hooking up companies, then there wasn't. I got out before it collapsed but you could see it was unsustainable. Later, due to my experience with the software involved, a colleague and I were tasked with recovering data from another huge firm that collapsed after the boom, walking into a massive empty office with a post it note with a server name and some passwords. Nothing on the db to recover but we had the skills to bypass the security.
christkv 3y ago
Got a friend working for Lumen and it's an overstaffed mess with declining income. He says it's pretty bad internally. They are all wondering if someone like Facebook, Google or Amazon will just snap it up but doubt if it will happen.
martinald 3y ago
FWIW this is happening again in the UK right now with FTTH deployments. There are about 10-20 companies of fairly sizeable scale, with hundreds of millions of billons of GBP of funding laying fibre. I think this is a good thing for the country (it has forced Openreach & Virgin Media to massively accelerate their FTTP plans); but it is going to be cataclysmic for the investors. Firstly, the build cost is astronomical and completely inefficient, apparently £1000/home is pretty standard, which is 3-10x what openreach and VM are paying (these altnets have access to all the openreach ducting and poles), due to very poor efficiencies in their build process and massively overpaying for materials and labour. Secondly, they overbuilding like crazy. I had 5 different FTTH providers at my old London apartment available, with totally separate infrastructure. This is meaning takeup is a fraction of what they expected it to be, as for the most part they expected to have at most one competitor at a property. Third, they seem to have absolutely no idea how to market and acquire customers, since their main skill seems to be how to pitch infrastructure funds to get capital. Their rollout is very patchy nationwide or even a local area so marketing campaigns are hard to do. There is going to be a huge wave of consolidation (at best) or bankruptcy in the sector as interest rates have risen and the calculus has totally changed.
- iamacyborg 3y ago
  I suppose this is what happens when infrastructure stops being funded by the government. I’ve got 3 current FTTH providers at my flat in London, I wasn’t even aware other companies were about and also offering the service. I guess this’ll be good for consumer costs in the short term before it inevitably implodes.
  - HPsquared 3y ago
    Nothing new: https://en.wikipedia.org/wiki/Railway_Mania https://en.wikipedia.org/wiki/Railway_Mania
    - iamacyborg 3y ago
      Isn’t it fun seeing history endlessly repeat itself.
aaron695 3y ago
[dead]
YesThatTom2 3y ago
The follow-up article should be about Google. Google saw the bandwidth glut (global fiber deployments were over-provisions and under-utilized) and took advantage of this. While earlier starts were in a "bandwidth is expensive" mentality, Google designed in a "bandwidth is cheap" mentality. It gave them a big advantage both in cost of operations and ability to design large scale systems differently than previous competitors had been able to. There was also a datacenter/colo glut. Google was buying up every datacenter square foot they could find. They could get it at pennies on the dollar just like bandwidth. This lasted until around 2009-ish when both gluts were exhausted. Google turned to running their own fiber and building their own datacenters just in time. TL;DR: Google's insight into the economics of global bandwidth and datacenter gluts became an advantage in the early years.
hospitalJail 3y ago
I'm not limited by my download/upload speed. Heck, 4G or 5G, whatever is on my phone, is absurdly fast. I don't see the demand for consumers. Heck, outside AI, the computer isn't even my bottleneck. Up until ChatGPT/Stable Diffusion was released, I thought computing was near end-game for consumer needs.
- malfist 3y ago
  640k should be enough for everyone right?
  - hospitalJail 3y ago
    Eh, we knew <1mb was limiting. The fact that I had multiple drives was a sign.
botswana99 3y ago
Fiber. Dot.coms, housing, crypto, AI, ….
nickdothutton 3y ago
The telco bust was much larger than dotcom bust. I have a couple of charts from the time here: https://blog.eutopian.io/nsa-prisms-commercial-cousin/ https://blog.eutopian.io/nsa-prisms-commercial-cousin/
beezle 3y ago
Bad, bad memories of MFNX debacle (metromedia fiber networks). Verizon and Kluge hosed everyone.

68 points by tim_sw 3y ago

FYI, this article is from 2016 and discusses the dot com bubble that popped around 2001. Jeff Hecht is an excellent author.

anonylizard 3y ago

Instant analogy to GPUs... A similar story is found in the railway bubble bursts in the US. Railways are also incredibly revolutionary technologies that reshaped societies, doesn't mean they are immune from investment cycles. That being said, given ChatGPT was only released 8 months ago, we are probably very early in the GPU boom, far from bubble yet.

hospitalJail 3y ago
Computing seems pretty important. I have a usecase that can save our company millions of dollars, but unless my company is willing to spend hundreds of thousands on a computer, its not happening. (It may not work, and prototypes still have me conflicted if it will work). We also can't rent servers for this, too confidential. If we can get prices down 10x-100x, I can even do this for personal/home usecases. I know that sounds extreme/absurd, but given our office is nearly empty with WFH, we can use larger computers and no one would mind.

KirillPanov 3y ago

Yeah the phone companies laid those thick 144-strand bundles before wave division multiplexing. They figured the only way to get more bandwidth was more fibers, right? WDM came along and caught them with their pants down; now all of a sudden you could just attach different devices to the ends of the fiber and get 16x the bandwidth. No backhoes needed. Then CWDM let you split and combine channels in locations without a reliable power supply -- no electronics, just prisms! Unfortunately their response -- those who weren't killed off -- was to get the antitrust regulators to look the other way (i'm sure the "national security" card was played) while they bought each other up and carved countries into a checkerboard of regional monopolies. But hey, if you squint at it from far enough away you can see two colors, so that's not a monopoly, right?

jacquesm 3y ago
In the meantime though we've started using a substantial fraction of those thick fiber bundles and dark fiber is getting harder and harder to find. Barring another revolution in technology I suspect that at some point we'll be laying even more of those big bundles, or even bigger ones.
- KirillPanov 3y ago
  > dark fiber is getting harder and harder to find Only because all of it along any given route now tends to be owned by a single entity. And that entity can simply say "not for sale". If the market were healthy dark fiber prices would skyrocket instead of availability plummeting.
liotier 3y ago
WDM guy here. Sure, 80 channels are space for lots of 400 Gb/s circuits... But we still fill cables and keep laying more. Though yes, the big 1288 fibers cables aren't common. Also, lambdas don't terminate at every node - so sparse network topology implies that filling all channels on a single link is a merely theoretical situation.
rasz 3y ago
>before wave division multiplexing. ... WDM came along Would that be Ciena in 1996?
senttoschool 3y ago
>Yeah the phone companies laid those thick 144-strand bundles before wave division multiplexing. They figured the only way to get more bandwidth was more fibers, right? I don't know anything about fiber tech but can't they just apply the newer techniques to the already laid out 144 strand bundles?

samuelkazeem 3y ago

>The investors who threw money at optics at the peak of the bubble probably would bankrupt us all if we ever hit a true “technological singularity.” If we really hit technological singularity, that will be the hail Mary of all bubbles.

somat 3y ago
conversely, I think we have been in the singularity since the late 1700's. For many thousands of years human manufacturing capability was effectively flat, a few blips and bumps but the advancement was generally steady and linear. And then in the late 1700's things changed. The graph went hockey stick and has stayed exponential since. However from inside this bubble there is a sort of tech dilatation where everything appears normal and flat and it is only the future that has exponential growth.
- libealistand 3y ago
  Flat? I don't think so. I'm at a coffee shop and looking around me, this is very different from just 20 years ago. People with laptops and smaetphones around me. Just 20 years ago, that same place saw a few CS major nerds with their laptops, everybody else sat with a book or some paper to write on or reading a paper newspaper. Still with a latte, but tech has changed everybody's life dramatically in 20 years. 20 years from now I'm again expecting at least as big a change. Definitely exponential.
  - rcxdude 3y ago
    You're agreeing with the comment you replied to. they were saying it was flat until ~1700 or so.
    - libealistand 3y ago
      They wrote that it "appears flat". I tried to contradict that. > However from inside this bubble there is a sort of tech dilatation where everything appears normal and flat and it is only the future that has exponential growth.

libealistand 3y ago

I don't get the drama. To me this looks like a normal process in an industry that discovered a groundbreaking technology. A new technology pops up, everybody rushes to innovate, lots of actors do nonsense, some do the right thing, but for many of them you don't know beforehand which is which. So you throw money at all of them. The rush creates some really good ideas and innovations. And a lot of BS and wrong decisions. Within a few years, the successes emerge, everybody else is flushed out of the market. Today fiber is the network tech. Except for the last mile, everything is fiber. We're at 400G per port, soon 800G, and the market keeps growing and 1.6T is the obvious next step. Normal growth process. So, I don't really see the takeaway here. What to learn from it? Not to invest in hypes? That's not really how we grow.

mangeld 3y ago
Depends on the country, where I live it's pretty common to have fiber to the home. Everything runs on that, internet, cable tv and phone calls.
- sandworm101 3y ago
  But in other areas most residential bandwidth increasingly comes from wireless techs. Even in homes with wired service, nearly everyone uses wifi rather than wires, let alone fiber to end machines. The last mile is generally not fiber, nor is it a literal mile.
  - cmrdporcupine 3y ago
    Indeed many people, including myself are now in the weird backwards world where the WAN is far faster than the LAN. I have a 3gbps fiber link straight to my house. The only machines in the house that can use anything close to it are the ones that hook up to my USB-C docking station which has a 2.5gbps ethernet port. Everything else is WiFi speeds. EDIT: This was a sudden change. I live rural and for the last 10 years I only had 10-15mbps speeds top, and most recently only by point to point wireless from a tower 5km away.

secretsatan 3y ago

I was involved in a UK company that strung fibre optics along the high voltage transmission network to quickly build a national network, there was a ridiculous amount of money flowing around hooking up companies, then there wasn't. I got out before it collapsed but you could see it was unsustainable. Later, due to my experience with the software involved, a colleague and I were tasked with recovering data from another huge firm that collapsed after the boom, walking into a massive empty office with a post it note with a server name and some passwords. Nothing on the db to recover but we had the skills to bypass the security.

christkv 3y ago

Got a friend working for Lumen and it's an overstaffed mess with declining income. He says it's pretty bad internally. They are all wondering if someone like Facebook, Google or Amazon will just snap it up but doubt if it will happen.

martinald 3y ago

FWIW this is happening again in the UK right now with FTTH deployments. There are about 10-20 companies of fairly sizeable scale, with hundreds of millions of billons of GBP of funding laying fibre. I think this is a good thing for the country (it has forced Openreach & Virgin Media to massively accelerate their FTTP plans); but it is going to be cataclysmic for the investors. Firstly, the build cost is astronomical and completely inefficient, apparently £1000/home is pretty standard, which is 3-10x what openreach and VM are paying (these altnets have access to all the openreach ducting and poles), due to very poor efficiencies in their build process and massively overpaying for materials and labour. Secondly, they overbuilding like crazy. I had 5 different FTTH providers at my old London apartment available, with totally separate infrastructure. This is meaning takeup is a fraction of what they expected it to be, as for the most part they expected to have at most one competitor at a property. Third, they seem to have absolutely no idea how to market and acquire customers, since their main skill seems to be how to pitch infrastructure funds to get capital. Their rollout is very patchy nationwide or even a local area so marketing campaigns are hard to do. There is going to be a huge wave of consolidation (at best) or bankruptcy in the sector as interest rates have risen and the calculus has totally changed.

iamacyborg 3y ago
I suppose this is what happens when infrastructure stops being funded by the government. I’ve got 3 current FTTH providers at my flat in London, I wasn’t even aware other companies were about and also offering the service. I guess this’ll be good for consumer costs in the short term before it inevitably implodes.
- HPsquared 3y ago
  Nothing new: https://en.wikipedia.org/wiki/Railway_Mania https://en.wikipedia.org/wiki/Railway_Mania
  - iamacyborg 3y ago
    Isn’t it fun seeing history endlessly repeat itself.

aaron695 3y ago

[dead]

YesThatTom2 3y ago

The follow-up article should be about Google. Google saw the bandwidth glut (global fiber deployments were over-provisions and under-utilized) and took advantage of this. While earlier starts were in a "bandwidth is expensive" mentality, Google designed in a "bandwidth is cheap" mentality. It gave them a big advantage both in cost of operations and ability to design large scale systems differently than previous competitors had been able to. There was also a datacenter/colo glut. Google was buying up every datacenter square foot they could find. They could get it at pennies on the dollar just like bandwidth. This lasted until around 2009-ish when both gluts were exhausted. Google turned to running their own fiber and building their own datacenters just in time. TL;DR: Google's insight into the economics of global bandwidth and datacenter gluts became an advantage in the early years.

hospitalJail 3y ago

I'm not limited by my download/upload speed. Heck, 4G or 5G, whatever is on my phone, is absurdly fast. I don't see the demand for consumers. Heck, outside AI, the computer isn't even my bottleneck. Up until ChatGPT/Stable Diffusion was released, I thought computing was near end-game for consumer needs.

malfist 3y ago
640k should be enough for everyone right?
- hospitalJail 3y ago
  Eh, we knew <1mb was limiting. The fact that I had multiple drives was a sign.

botswana99 3y ago

Fiber. Dot.coms, housing, crypto, AI, ….

nickdothutton 3y ago

The telco bust was much larger than dotcom bust. I have a couple of charts from the time here: https://blog.eutopian.io/nsa-prisms-commercial-cousin/ https://blog.eutopian.io/nsa-prisms-commercial-cousin/

beezle 3y ago

Bad, bad memories of MFNX debacle (metromedia fiber networks). Verizon and Kluge hosed everyone.