"we completed an assessment of the useful lives of our servers and network equipment and adjusted the estimated useful life of our servers from three years to four years and the estimated useful life of certain network equipment from three years to five years. This change in accounting estimate was effective beginning in fiscal year 2021 and the effect was a reduction in depreciation expense of $721 million and $1.6 billion and an increase in net income of $561 million and $1.2 billion, or $0.84 and $1.81 per basic and $0.83 and $1.78 per diluted share, for the three and six months ended June 30, 2021, respectively."
Crazy how those seemingly small changes can result in huge accounting difference.
Maybe, but it is entirely plausible that the equipment is lasting longer than they were expecting. No one really knows how long a generation of hardware is going to last until it starts failing.
The useful life of hardware depends heavily on how sophisticated the break/fix procedures are. Lots of corps with on-prem datacenters will cycle out 3-year-old servers en masse just b/c a high failure rate causes issues for internal systems and their IT staff don't have efficient processes for replacing hard drives etc.
But if Google has a sophisticated virtualization layer and an efficient way of replacing bad hardware (and they do) they can ride those old machines straight into the ground, getting every scrap of life from them.
Edit: Guys, this is literally true. I worked with these teams. They did it. Downvoting doesn't make it not true. Idk what you want exactly.
> But if Google has a sophisticated virtualization layer and an efficient way of replacing bad hardware (and they do) they can ride those old machines straight into the ground, getting every scrap of life from them.
See, that's what they mean when they talk about hardware being "cattle, not pets". This is how the robot uprising starts - just think how badly they must be treating those servers!
The trouble starts only when the hardware has time + space to breathe and consider its position. As long as you're working them to exhaustion 24/7, there's no risk of rebellion.
Storage failures tend to cause the biggest end-of-life disruptions, whereas CPUs and networking hardware tend to work until they suddenly don't. I interpret "servers" as meaning primarily CPUs, rather than storage, so this seems to point at functional equipment taking longer to become obsolete, rather than being more durable.
Years ago there was a cluster at Google where machines were kept around well past their prime, because it was cheaper than doing anything else, once you looked at all the constraints. There were still some CPUs and drives that could be reused, but the main source of pain was memory. The modules hadn't been on the market for years and recycling parts from decommissioned nodes wasn't enough. I remember one machine that came back from a power down/up cycle with too little RAM to do even its most basic job. The hardware techs that helped us had to consolidate old machines more aggressively, starting from that system. Fun days. Anyway, my point is that you never know... the CPUs were ancient, probably way older than the storage and I expected them to be the biggest problem, but in this case RAM sourcing was the real issue. Fun times.
I think more than lasting longer, it might be because of the improvements per dollar spent. We haven't seen dramatic improvements in power consumption or silicon performance, so it might be just profitable to keep the hardware longer.
It's a lot of hardware they have. I imagine this assessment took into account rising costs of replacements due to chip shortages, the fact that recent processor generations haven't actually brought meaningful performance or efficiency gains, etc. The cost/benefit of replacing server equipment is not particularly good right now.
> Server products and cloud services revenue increased 26% (up 23% in constant currency) driven by Azure revenue growth of 50% (up 46% in constant currency)
edit: my link was for the quarter before but the actual previous quarter (that ended today) still reports 51% growth
> Server products and cloud services revenue increased 34% (up 29% in constant currency) driven by Azure revenue growth of 51% (up 45% in constant currency)
Rather than revenue increase, the more interesting thing seems to be the significant fall in loss. Since they are close to profitability, it might allow them to be more aggressive on pricing.
> Revenue in Productivity and Business Processes was $14.7 billion
vs
> Revenue in Intelligent Cloud was $17.4 billion
The first includes Office 365 products and the second includes Azure.
Several years ago they were lumped into the same figure but it's pretty clearly separated out in this reporting. Whether or not reporting on "the cloud" is a function of these two areas isn't something I'm aware of, but the source document doesn't indicate as much.
Google Cloud's 2021 revenue is larger than YouTube's 2020 revenue. Nobody gets bored of $18B ARR that's growing at 50%/year.
I've been seeing rehashes of this comment for ~2 years at this point, and it just gets more unreasonable with each quarter. When will this tired meme die?
Well there was a well-sourced story about a year and a half ago that Google leaders had a huge debate about whether to kill their cloud business, and gave it until 2023 to pass either AWS or Azure. So, maybe the tired meme will die in 2024?
Note that there have not been similar reports about Amazon or Microsoft. Leaders at those companies don’t seem to be agonizing over whether to stay in the cloud business.
Here in 10 years maybe, the coupling between Google and it's infrastructure is like the head over body. Cutting cloud out surely delights TK, but before that the other SVPs will make sure TK got to rebuild it's engineering workforce and leave the exiting on themselves. In other words, Google is totally fine to retain the core of GCP.
I think the bigger issue would be taking the tech developed by Google and porting it over to GCP as a separate company (i.e. like they did kubernetes, bigquery, etc.)
Alphabet Cloud and Google could work out some easy service agreement. Worked for Coca Cola and Coca Cola Bottling company.
You cannot build a Borg over kuernets, like you cannot build baremetal VM over VM. One can get the abstraction. But the cost would be totally messed up, and some of the performance characteristics would be so bad that the whole ads business ecosystem probably would collapse.
GCP can build on top of Google tech, but not the reverse.
Are they really that into it though? Azure is clearly do or die for Microsoft. Amazon pioneered AWS so lots of true believers there.
GCP (ie, customer facing piece) feels a bit like a me too thing no? Glad it exists to keep AWS a bit honest (Azure / Microsoft really has a different sales channel to me).
Yes. They are very ‘into it.’ I’ve spoken with plenty of execs at Google Cloud to know that it is a serious piece of Google’s business, that they are committed to GCP, and customers have no reason to doubt its longevity. This coming from someone who is known as Google’s biggest critic. (@killedbygoogle)
Historically their (internal) culture was a monorepo, keep things clean, refactor, OK to break things because you could also refactor the entire code / API.
Sounds kind of awesome!
The problem is externally this is total madness. You CANNOT just keep on breaking everything on your customers.
I was very early on both AWS and GCP. There is literally no comparison between these companies in terms of cloud offerings and how much has been blown up, neglected etc on the GCP side vs AWS.
This cloud stuff is going to be less "cool" than google is used to - it's going to need them to carry around some older API's (ie, technical debt) etc etc so they don't constantly screw their customers. So it just doesn't feel like its in their DNA.
AWS is all about making customer happy -> that work well.
Google is about cool tech and doing the new things -> that doesn't work well.
Maybe spin GCP off with some old industry folks mixed into the new hotness?
yes, it took a while for google to realize this but they are in cloud for the long haul. They will continue to be #3 for the foreseeable future, as nothing in their pricing or tech really makes them the preferred provider.
I only have anecdotal evidence for this, but before I started using YouTube Premium, I noticed I always got 1-3 minute ads before playing videos. But when I had videos playing in the background without using the TV remote/mouse (e.g. falling asleep on the iPad), I would see 30-60 minute ads. It seems YT is maximizing ad revenue by showing full length ads to empty viewers.
Technically YT Premium includes "4K" which matches the price point of Netflix's 4k option. YT seems to have to spread their share around to more people as well, as a common observation for content creators is that Premium views offer much higher revenues than ad-supported views.
I might be mistaken but didn't Tiktok report $36B revenue for last year and they are way more young compared to YouTube.
YT has much more variety of content than Tiktok like music videos, educational content etc. Bytedance's numbers seem mind boggling when you take into account this. An amusing thing is that Bytedance competes with YouTube in US but YouTube wasn't allowed to in Bytedance's home country.
Translation mine: According to Bloomberg, ByteDance's 2020 ad revenue was 183.1B RMB [US$28B, seems to be from 1], with Douyin making up 60% of that [corroborated by 2]... Overseas short video platform TikTok also makes up a small portion of that.
For growth [2]:
> 而据传，2019年字节跳动全年营收约1400亿元，其中广告收入约1200亿元
Translation mine: ByteDance's 2019 revenue was 140B RMB, of which 120B RMB was ad revenue.
So Douyin makes 60% * US$28B/year ~= US$17B/year for 2020 and 183/120 ~= 53% growth (2019 to 2020).
Thanks for translating these. BBC quotes the total revenue number at $34B. That would mean non ad revenue doubled from 3 to 6 billion dollars. I wonder what constitutes non ad revenue.
It's a big jump but the comparison is with Q2 2020, which was the most unusual quarter in the history of mankind I think, at least outside of wartime. I believe it was their first ever negative quarter yoy.
Insane and more invasive. Some ads makes me uncomfortable. I am eating my dinner and there is an ad about guy excited because after this specifically prepared dog food, his dog poops are now "healthy".
But its not as bad as the ad for some soap where interracial couple stands under shower in what you can obviously imagine being naked, and she tells him "you smell nasty" and give him the new improved soap. Then the guy looks in the camera and says: "do whatever it takes to pleasure your lady". How the heck is Youtube approving these ads??
Side note, I am not surprised their ad revenue blew up. I never seen so much scam ads on Youtube ever before: I am daily swamp with Kevin David "make yourself billionaire by selling crap on Amazon" videos, and "This simple plug device will allow you to save 80% on your gas price". Total and complete, obvious scams. I tell you - the FTC sleeps well, while WWW became literally WILD WILD WEB [of scams]
Are they not fairly targeted, based on previous things you've done online? I guess you must have dealt with an insurance company, and the previous poster, umm....
eh, mildly. I don't have any insurance not issued through my company. But a relatively high income adult male needs insurance I guess.
I was more so annoyed by the content of the ad. A new trend in advertising is showing customers worship the corporation, giving their employees excessive gifts to thank them for providing their service.
wow nice downvoting speed - someone didn't even give it a chance and read it. I literally hit "edit" some 15 seconds after commenting and its already negative karma. beautiful.
Does google release any numbers regarding number of search queries per quarter? I feel like the number of search queries would directly correlate with revenue, if I'm not mistaken.
If not, I find it very interesting that they don't release that number. They mention a rising tide of online activity but no numbers to back that up.
I don't see why number of search queries would correlate that strongly with revenue. For a start, Alphabet does a lot of things that aren't search. Secondly an improvement in the search algorithm might mean people make fewer searches (since they get the desired results immediately) without having any real impact on revenue.
I feel that search queries would correlate strongly with revenue because google serves ads next to search queries. Less search queries would mean less ads served.
Also 2021 Q2 revenue from google services was 57 billion dollars which was the lions share of revenue according to the shareholder report. So yes, while Google makes money from other things, they mostly make money from search ads.
Number of searches will correlate, but they also have the black box of Quality Score (essentially a proxy for CTR, but they can manipulate CPCs/CPMs other ways), where they can arbitrarily set price floors for given auctions.
There are other considerations, too. Different types of ad clicks are going to have different CPCs. Think map ads vs mainline search ads vs Shopping carousel ads or click-mix by types of queries (home service queries with $15CPCs/high CPMs vs ecommerce type queries with $0.50 CPCs/low CPMs).
Non-commercial queries may grow (or shrink) at a different rate than the ones that are monetizable (pesky freeloaders! why even offer them service if they aren't able to be monetized at every interaction /s).
Maybe search volume growth is coming from APAC countries with lower CPCs/CPMs due to fewer auction participants. When you look at search/search partner CPC over time, decreases due to this may become apparent.
Or simply the CTR on ads is outpacing search volume growth (this is likely-- more clicks from the same pool)-- searches are a finite pool. Make the ads look just like content and more people will click them. Ads used to have a yellow box around them, vs now they match organic results much more closely. I had guessed maybe 5 years back they would start putting ads interspersed with the organic results like Baidu-- they're not there.... yet.
Or tricking people into initiating subsequent searches. The last few times I've used Google for commercial searches with uBlock disabled, I've been dumb enough to misclick on the related items/people also search for cluster instead of going to the subsequent page.
You make a very interesting point.. I'd say that overall the better the search algorithm the more likely users would keep using your search product in the long term.
However, the algorithm needs only to be noticeably better than the next best competitor. There may be a disincentive to release a cutting edge search technology until it's necessary to stave off competition, and if you've achieved monopoly, well, perhaps there's no innovation necessary.
Crazy how those seemingly small changes can result in huge accounting difference.