9 ms·

I always assumed YouTube’s technical foundations were significantly subsidized by the mothership, but I have no data to base that on other than looking at the $

by jl6 4y ago

I always assumed YouTube’s technical foundations were significantly subsidized by the mothership, but I have no data to base that on other than looking at the $29bn/year revenue figure and wondering whether that amount of money would be enough to provision the amount of compute, storage and bandwidth that YouTube needs, plus all other expenses (including revenue sharing with uploaders). Yes, $29bn is a lot of money, but there is a lot of YouTube.

Aromasin 4y ago
From what I'm aware, a lot of these larger companies basically treat other parts of the business the same they would customers; very important customers, but customers all the same. If YouTube wants more cloud budget, it comes out of their budget and goes into revenues of the Cloud business unit. Not quite that simplistic of course, but in essence it's meant to drive competitiveness. No special privilege's just because you're part of the same team. If an external company is paying more, for more compute, the business goes to them.
- bpodgursky 4y ago
  Yeah I think this is generally true, although YouTube is a weird case because there's literally no equivalent customer with even a fraction of the compute volume.
- curiousllama 4y ago
  That stuff still comes with special privileges. E.g., yes they'll "pay" for compute, but they'll do so at cost. And they're also allowed to deploy a monitoring widgets to other orgs' data centers to find when they're at low utilization, and then argue "at cost" is actually near 0, because the compute resources were just sitting around anyway. And then have the search team agree because YouTube promises to build the new data API search has been asking for forever. If you ultimately report earnings on the same P&L, USD is just monopoly money you trade around to get things done.
kevincox 4y ago
When I worked at Google I definitely noticed that YouTube was great at (ab)using untracked or very cheap resources. They definitely get huge benifits of otherwise idle data centers in sleeping regions. As the regular service traffic died off YouTube would put those idle cores to work transcoding video. It is definitely the case that YouTube gets lots of really cheap compute by being part of Google. But that isn't a bad thing, they are making use of an otherwise wasted resource, effectively lowering the cost for products that need daytime capacity to serve humans.
- jelling 4y ago
  The commentators point was that YouTube may not be as strong of a stand-alone company as the article believes if the COGS is inaccurate because YT gets tons of free resources via Google.
- BooneJS 4y ago
  “In our case, we developed a custom chip to transcode video, as well as software to coordinate these chips. And we put it all together to form our transcoding special brain – the Video (trans)Coding Unit (VCU). We’ve seen up to 20-33x improvements in compute efficiency compared to our previous optimized system, which was running software on traditional servers.” https://blog.youtube/inside-youtube/new-era-video-infrastructure/ https://blog.youtube/inside-youtube/new-era-video-infrastruc...
  - kevincox 4y ago
    If I had to guess these are used for urgent video transcodes. New uploads to popular channels and on-demand transcoding. I suspect that CPUs are still often used for background batch encoding. But maybe 20x is enough that the power cost of extra CPU usage isn't worth it and they will do it all in hardware.
- fisf 4y ago
  You are basically making the case for the anti-trust case here. i.e. it is very hard for another company to compete without "free compute".
jsnell 4y ago
Note that the content isn't free either; more than half of the revenue goes to the creators.
klodolph 4y ago
Think of it less as a subsidy. Instead, think of this— Team A is running a database and needs to buy disks with a certain amount of IO capacity. Team B is running YouTube and needs to store tons of data. You add up the IO + storage requirements from team A and the IO + storage requirements of team B and you buy disks capable of providing that combined amount of IO + storage. The resulting cost is much lower than the cost of buying disks for team A and disks for team B separately.
nerdjon 4y ago
This was my first thought as well. Is YouTube viable as its own company at the scale it operates? I have to imagine the infrastructure costs for YouTube would be extremely high and would likely cause YouTube to downgrade quality or try even harder to put a subscription behind even 1080. Now obviously contacts can be signed and I highly doubt they would be paying normal prices in any cloud environment given this hypothetical situation. But it wouldn't be the same as it would be now.
- chrisfinazzo 4y ago
  Yep. Does YT need Google's backing in order to survive because they don't have the resources strictly in-house to build and pay for infrastructure? Disclaimer: Not an Economist, not even close. If in the unlikely case they spun it off and another competitor emerged, would YT's scale make it hard to compete with this new entrant? Read: Red tape, bureaucracy, etc, etc. Essentially back to where they were as an independent company when they didn't have Google to "protect" them... The inverse case is what usually comes to mind first though: Would the company wither and die if not for an acquisition? Think about Sprint+TMobile. Until the deal, Sprint was in bad shape, second to Verizon with outdated tech - in the sense that emerging technologies were based more off of an evolution of GSM than CDMA. If TMo doesn't acquire them, Sprint probably goes "poof" and their customers flee to the remaining carriers, further entrenching their positions. This seems like it is usually the card European regulators tend to play, worrying more about market dynamics and how that will affect costs rather than just what people pay for service.
recuter 4y ago
And yet up until a couple years ago at least Netflix served a higher percentage of web traffic. Don't know about now, maybe that changed. There is only so many hours in the day and so many eyeballs watching, that is your upper bound. Bandwidth like that is not available to mere mortals, you would have to colocate with a lot of ISPs all over the place. Would take years to setup, not really even a question of money. In the case of Netflix their content library fits inside one such colocated server, which can (or pretty soon will) do 1tb/s. Hundreds of thousands of streams per box are within reach, especially at YouTube potato resolutions. In the case of Youtube while the long tail is very very long, I bet the most popular content of the week would fit into a similar sort of server. So the serving portion at least is an endeavor on the order of magnitude of say 1 billion dollars a year. No idea how to estimate letting anybody upload unlimited amounts of video of unlimited length and transcoding it into a dozen sizes. I think they did that early on to solidify their moat and it seems like a huge waste.

by jl6 4y ago

Aromasin 4y ago

From what I'm aware, a lot of these larger companies basically treat other parts of the business the same they would customers; very important customers, but customers all the same. If YouTube wants more cloud budget, it comes out of their budget and goes into revenues of the Cloud business unit. Not quite that simplistic of course, but in essence it's meant to drive competitiveness. No special privilege's just because you're part of the same team. If an external company is paying more, for more compute, the business goes to them.

bpodgursky 4y ago
Yeah I think this is generally true, although YouTube is a weird case because there's literally no equivalent customer with even a fraction of the compute volume.
curiousllama 4y ago
That stuff still comes with special privileges. E.g., yes they'll "pay" for compute, but they'll do so at cost. And they're also allowed to deploy a monitoring widgets to other orgs' data centers to find when they're at low utilization, and then argue "at cost" is actually near 0, because the compute resources were just sitting around anyway. And then have the search team agree because YouTube promises to build the new data API search has been asking for forever. If you ultimately report earnings on the same P&L, USD is just monopoly money you trade around to get things done.

kevincox 4y ago

When I worked at Google I definitely noticed that YouTube was great at (ab)using untracked or very cheap resources. They definitely get huge benifits of otherwise idle data centers in sleeping regions. As the regular service traffic died off YouTube would put those idle cores to work transcoding video. It is definitely the case that YouTube gets lots of really cheap compute by being part of Google. But that isn't a bad thing, they are making use of an otherwise wasted resource, effectively lowering the cost for products that need daytime capacity to serve humans.

jelling 4y ago
The commentators point was that YouTube may not be as strong of a stand-alone company as the article believes if the COGS is inaccurate because YT gets tons of free resources via Google.
BooneJS 4y ago
“In our case, we developed a custom chip to transcode video, as well as software to coordinate these chips. And we put it all together to form our transcoding special brain – the Video (trans)Coding Unit (VCU). We’ve seen up to 20-33x improvements in compute efficiency compared to our previous optimized system, which was running software on traditional servers.” https://blog.youtube/inside-youtube/new-era-video-infrastructure/ https://blog.youtube/inside-youtube/new-era-video-infrastruc...
- kevincox 4y ago
  If I had to guess these are used for urgent video transcodes. New uploads to popular channels and on-demand transcoding. I suspect that CPUs are still often used for background batch encoding. But maybe 20x is enough that the power cost of extra CPU usage isn't worth it and they will do it all in hardware.
fisf 4y ago
You are basically making the case for the anti-trust case here. i.e. it is very hard for another company to compete without "free compute".

jsnell 4y ago

Note that the content isn't free either; more than half of the revenue goes to the creators.

klodolph 4y ago

Think of it less as a subsidy. Instead, think of this— Team A is running a database and needs to buy disks with a certain amount of IO capacity. Team B is running YouTube and needs to store tons of data. You add up the IO + storage requirements from team A and the IO + storage requirements of team B and you buy disks capable of providing that combined amount of IO + storage. The resulting cost is much lower than the cost of buying disks for team A and disks for team B separately.

nerdjon 4y ago

This was my first thought as well. Is YouTube viable as its own company at the scale it operates? I have to imagine the infrastructure costs for YouTube would be extremely high and would likely cause YouTube to downgrade quality or try even harder to put a subscription behind even 1080. Now obviously contacts can be signed and I highly doubt they would be paying normal prices in any cloud environment given this hypothetical situation. But it wouldn't be the same as it would be now.

chrisfinazzo 4y ago
Yep. Does YT need Google's backing in order to survive because they don't have the resources strictly in-house to build and pay for infrastructure? Disclaimer: Not an Economist, not even close. If in the unlikely case they spun it off and another competitor emerged, would YT's scale make it hard to compete with this new entrant? Read: Red tape, bureaucracy, etc, etc. Essentially back to where they were as an independent company when they didn't have Google to "protect" them... The inverse case is what usually comes to mind first though: Would the company wither and die if not for an acquisition? Think about Sprint+TMobile. Until the deal, Sprint was in bad shape, second to Verizon with outdated tech - in the sense that emerging technologies were based more off of an evolution of GSM than CDMA. If TMo doesn't acquire them, Sprint probably goes "poof" and their customers flee to the remaining carriers, further entrenching their positions. This seems like it is usually the card European regulators tend to play, worrying more about market dynamics and how that will affect costs rather than just what people pay for service.

recuter 4y ago

And yet up until a couple years ago at least Netflix served a higher percentage of web traffic. Don't know about now, maybe that changed. There is only so many hours in the day and so many eyeballs watching, that is your upper bound. Bandwidth like that is not available to mere mortals, you would have to colocate with a lot of ISPs all over the place. Would take years to setup, not really even a question of money. In the case of Netflix their content library fits inside one such colocated server, which can (or pretty soon will) do 1tb/s. Hundreds of thousands of streams per box are within reach, especially at YouTube potato resolutions. In the case of Youtube while the long tail is very very long, I bet the most popular content of the week would fit into a similar sort of server. So the serving portion at least is an endeavor on the order of magnitude of say 1 billion dollars a year. No idea how to estimate letting anybody upload unlimited amounts of video of unlimited length and transcoding it into a dozen sizes. I think they did that early on to solidify their moat and it seems like a huge waste.