Meta wants to sweat its servers for longer – at a cost of $60b

Meta's share price has taken a twenty percent dive after investors marked down the social media company's datacenter infrastructure spending, as well as weakness in the ad market. The home of Facebook, Instagram, WhatsApp, and various metaverse experiments today reported [PDF] Q3 revenue of $27.7 billion – a four percent dip …

[Gordon 10]

Can someone parse this statement for me?

“ Having more datacenter headroom will allow us to extend the life of servers over time because we won't have to replace them due to power constraints," Wehner said. "Part of that is to get more efficiency over time out of the CPU-based server fleet.”

Sounds like BS.

[Anonymous Coward]

Re: Can someone parse this statement for me?

Agree it sounds like BS. I parse it as newer generation servers are more power (and thus thermally) efficient, so put less strain on the datacentre infrastructure. Less strain per server means they can fit more servers in without having to upgrade the power/cooling again.

[ChoHag]

Re: Can someone parse this statement for me?

"We have spare capacity in our racks. We can burn through that rather than replacing kit as it dies until we are running on very thin ice indeed."

"Power constraints" is a red herring except inasmuch as power costs are increasing.

[T. F. M. Reader]

Re: Can someone parse this statement for me?

It certainly has the look and feel of corporate BS, but it may actually be quite sensible when you do parse it. My understanding is, "let's overprovision data centres today, presumably in specs and numbers alike, in the hope that we won't need more CAPEX investment for replacements for longer".

I assume Meta cannot easily do what just about everyone else does: swap CAPEX for OPEX by renting servers in the cloud: 1) major cloud providers are competitors in one sense or another; 2) Meta's scale is too huge; 3) Meta's requirements are too unique even beyond scale. So CAPEX it is, and optimizing that over time is generally a good idea.

Their adopted approach may or may not be right - one can only hope that they have done a deep analysis before jumping head first.

[Charlie Clark]

Re: Can someone parse this statement for me?

Ψ The Company Formerly Known as Facebook is big enough to set up a subsidiary to buy hardware which the various tentacles then lease at about cost. CAPEX is thus maximally efficient and OPEX is just gravy.

[Anonymous Coward]

Re: Can someone parse this statement for me?

My reading of the statement is that for a given amount of CPU compute, that through improved power/space/cooling efficiency they can grow at a defined rate for 5 years instead of 3

Looking at the Open Compute Rack V2 specs, it looks like Facebook (and others using this standard) have increased network bandwidth, power efficiency, cooling and density to the point where they can fully support 96 compute "sleds" and supporting infrastructure in a 13.2kW/48U/21" rack when they used to support 64. It looks like networking capacity and power efficiency improvements led to the increase.

[Anonymous Coward]

Facebook

How quaint. All the boomers love it, I hear.

[Anonymous Coward]

"Investors clearly aren't sure Zuck's plan has any connection to reality – virtual or otherwise."

This made me laugh ...

Indeed, General Zuckpioca* has now completely moved to virtual reality, it seems.

* In memory of General Tapioca, the dictator, in "Tintin and the Picaros" from Hergé

[Charlie Clark]

The future is the Matrix, apparently

The Zuck says Third is more social interactions move to messaging. We are developing a flywheel between discovery and messaging that is going to make these apps stronger.

I take that to mean that nowtrage insertions are coming soon to messaging to keep people "engaged" aka hooked watching the mixture of banal shit from friends and family with the stuff that scares us and makes us really angry.

[Andy Non]

So they want to spam the feed even more?

"our discovery engine work allows us to recommend all types of content beyond Reels as well, including photos, text, links, communities, short and long-form videos and more. Second is that we can mix this content alongside posts from your family and friends, which can't be generated by AI alone."

The main reason I've pretty much given up on visiting facebook is all the extra crap they put in the feed: Irrelevant reels, irrelevant suggestions for other groups, irrelevant adverts that bypass my adblocker. The recommended videos are a total joke... utterly unrelated to my interests. The content to noise ratio is already bad but they are striving to make it even worse. The more sh1t they shove into my face trying to hook me, just drives me further away. Nowadays I maybe log in once a week to see if there are any interesting posts in the two remaining groups I'm a member of; which usually means in and out again in five minutes. Facebook really have lost their way.

In a little twist, last week Facebook suggested I send birthday wishes to one of my friends (who I knew in real life), which I would have done had they not been dead for a year. I guess it is just down to everyone to "unfriend" the deceased.

[MJB7]

Re: Dead Friends

Actually, it is supposed to be down to the nearest and dearest to "memorialize" the dead account. It's a tough problem; they probably _could_ scan everyone's feeds with AI and detect when someone dies - but they would inevitably get it wrong from time to time, and the publicity would be awful.

Of my 141 friends, four are dead. I expect that proportion to grow over time.

https://what-if.xkcd.com/69/

[Boris the Cockroach]

I can do the

AI programming for next to nothing

1. Load user history

2. Check videos watched

3. Offer more videos from the same source

4. If no clicks then offer videos of same subject

5. Goto 3

[localzuk]

Metaverse and VR

$9.4 billion on VR and the metaverse!? What on Earth are they spending that money on?

That's jus under $26m a day on it. Nearly $20,000 a minute.

What is that money actually being spent on!?