Hacker News Reader

Comment by vsskanth | Hacker News Reader

They assume YouTube's content costs will scale similarly to Netflix, which doesn't appear correct to me, mainly because YouTube content is largely user generated and I would assume orders of magnitude more new content is uploaded per day than Netflix. I don't know if you could even use Netflix CDN tricks since their content keeps changing constantly.

Considering pretty much every YouTube video I see these days is uploaded in 1080p or 4k, these costs would start to matter eventually.

I wonder how this affects their operating margin, considering this article calculates it to be between 5-18%.

View on HN

The cost of delivering the video decreases over time and unlike Netflix which has to pay for content YouTube doesnt which gives it tremendously better margins.

Just look at the debt load of Netflix.

The cost of acquiring content is so much greater than serving it especially now that every traditional media company has launched its own streaming service.

Also you can never compete with the viewer numbers between a product that is paid versus free so the most that YouTube has established is clearly much greater and visible in that it has no direct competition in the user generated video space for 15 years until perhaps maybe tiktok but tremendously different platforms.

Meanwhile Netflix has competition from Amazon Prime, Disney, Hulu, HBO, and others.

>The cost of delivering the video decreases over time and unlike Netflix which has to pay for content YouTube doesnt which gives it tremendously better margins.

Youtube has to share it's revenue with content creators, unlike Netflix.

>The cost of acquiring content is so much greater than serving it especially now that every traditional media company has launched its own streaming service.

Does this hold true for constantly changing HD video content, unlike traditional media ?

Youtube has to share it's revenue with content creators, unlike Netflix.

I don't know if this is true.

I can't say that I know the inner workings of any of the Netflix contracts, but in entertainment it is common to not only pay to acquire the rights to display the content, but then to pay again for each time it is viewed.

Sometimes it's paid monthly or yearly. In other cases (broadcast television, for example), it's part of the negotiating process for content renewal. ("Oh, the ratings for our show on your station are up 40%, so we want more money to renew.")

Netflix content contracts are typically fixed-fee ones (source: https://ir.netflix.net/ir-overview/top-investor-questions/de... section "How do you account for streaming content licensing?").

The providers of course generally want more money the more popular a title is (though Netflix does not usually disclose viewership numbers so they may have to rely on Nielsen etc.).

> Youtube has to share it's revenue with content creators, unlike Netflix.

Depends on the contract. That is why you see more and more content produced by Netflix, where they control the complete chain. However, producing Netflix content has an initial cost, but producing YouTube videos does not (for Google) and the videos are monetized only if the number of viewers are increased. Which also increases ad earning to Google.

Google might be in a much better position here regarding streaming quality (or traffic reduction control). You get YouTube videos for free and even if you request HD content, you don't have to get it every time; no one can complain because it is free. However, for Netflix, you pay for HD quality and you expect that almost always.

Google's production cost is uploading bandwidth and storage. If you look at a single popular video it's easy to think, "Oh, that's nothing." But consider the massive amount of video uploaded that only gets 1 - 10 views over its lifetime.

Some back of the envelope math suggests that a 30 minute video needs about 10 views a month to pay for its storage. So clearly there’s some amount of the popular stuff subsidizing the unpopular stuff but the break even point is much further down the long tail than people realize.

> Youtube has to share it's revenue with content creators, unlike Netflix.

From a financial modeling perspective, that’s more akin to a tax paid by YouTube than anything else, which makes it easy to plan for. That’s not the case with Netflix, which has all sorts of competition bidding up costs and making content costs much more difficult to predict per unit of revenue.

Correct, YouTube has much more favorable (lower) content cost model than Netflix

Youtube keeps adding and adding content. Even though storage is nearly free in capex, it still has opex costs.

Netflix on the other hand has small selection, which gets purged frequently.

Isn't YouTube now actively throwing out old content?

any proof? I can still see my content from 20 years(probably) ago .

Likely no more than 15 years ago, unless you just broke embargo on the Tardis project.

https://en.wikipedia.org/wiki/Youtube

Not that I am aware of.

It is possible that they are throwing away different quality transcodes of old and unpopular content and regenerating on the fly if needed. But my understanding is that videos are never automatically removed.

> videos are never automatically removed.

Unless the video trips some automated DMCA thing, or the owner of the video account runs afoul of some inscrutable algorithm, or journalists start asking about disturbing videos aimed at children, or...

But yes, probably not to save money on storage.

Especially since those don't actually remove the files, since there's almost always still a chance to get the account back.

I suspect YouTube would predict the popularity of videos and only cache where it matters (low anticipated popularity videos would be pushed to CDN if they performed better than expected). YouTube-dl in some cases gets my full pipe speed, in others I see it as low as 100kb/s which is evidence to my theory.

That being said the data storage problem they have is insurmountable. I don’t think they ever delete low view-count content, so I’d be curious to know about their storage tradecraft.

I remember in the early days this was very noticeable. Each time you watched a less popular video it would buffer in the beginning.

I experienced the opposite, because YouTube would allow their local caching servers in my ISP's datacentre to get congested. Popular videos wouldn't stream at all, but unpopular niche content was fine because it went straight out to the internet.

>I suspect YouTube would predict the popularity of videos and only cache where it matters (low anticipated popularity videos would be pushed to CDN if they performed better than expected).

That's actually a really interesting point, because YouTube can (partly) decide which video will become popular. They control what shows up as a suggested video, what appears at the top of search, what shows up in the subscription feed (unless the subscriber has clicked the bell icon).

There’s probably a bunch of ML they could use. I was just thinking CDN anything from popular artists, or with the string minecraft in the title ;)

I'd guess that, like most platforms, youtube content follows a power law concerning the activities, meaning that 99% of the traffic comes from 1% of the content (or something along those lines). In the Netflix CDN, each node has a fixed library that's restocked at fixed intervals, at least that's what I remember from reading their requirement document for ISPs interested in having a CDN node in their network. The Google CDN is likely more dynamic with nodes having region dependent libraries but nothing that you can't manage.

> I don't know if you could even use Netflix CDN tricks since their content keeps changing constantly.

Netflix CDN tricks like preloading content during non-peak hours probably doesn't work for YouTube; afaik, they don't have the same sorts of content available before publication date.

However, Google has a CDN appliance program too, and read-through caching works pretty well.

Reply on news.ycombinator.com