4 ms·

> In January 2021, we completed an assessment of the useful lives of our servers and network equipment and adjusted the estimated useful life of our servers fr

by easton 5y ago

> In January 2021, 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 Does anyone know what the observation was here that made this possible? Have Google's internal workloads not grown in size as quickly as they thought, or is this a "Moore's law is slowing down" thing?

kyrra 5y ago
Googler, opinions are my own. I know nothing about this. I wonder if this was Google's response to the chip shortage? Maybe the cost/benefit (and just sheer inability to get certain parts) made them hold on to hardware for longer.
- DennisAleynikov 5y ago
  That would be an interesting adjustment
- bduerst 5y ago
  Could also be the fact that Google has built a ton of new data centers in the last two years. Newer hardware, infra, utilization, etc. could lead to better quality/lifespans.
sidibe 5y ago
I don't think the workloads would have much to do with this, probably saw that they were having less failures than anticipated or memory/chip shortages have them keeping equipment around longer than they planned.
jdavis703 5y ago
My assumption is the hardware was more reliable than expected. If it was about slowing workload growth one might expect them to switch to a usage-based (instead of time-based) depreciation schedule.
DennisAleynikov 5y ago
Seemingly the latter. Less benefit between generations and more stress resistance in hardware compared to earlier data center practice
cj 5y ago
I wouldn't be surprised if the decision was driven by Google's army of accountants. Extending the life of an asset would affect the depreciation schedule, which would affect when depreciation is recognized / recorded on their books in a given tax year.
slo_news 5y ago
While I'm unsure of the the reason the computers are lasting longer, I would bet this observation was originally triggered from the public company accounting firm that audits Alphabet's financials. The accounting firm likely asked Alphabet to prove out why they used a 3 year estimate and found out 4 years was more accurate (FYI 3 years is a standard useful life for computers for most companies - sometimes even shorter).
- hogFeast 5y ago
  It doesn't work this way. If Alphabet wanted to prove they lasted three years, they would have found a way to prove it. Auditors don't go around challenging on things like this. In particular, they don't go around challenging companies on depreciation because that is a key lever that management can pull if they need to hit numbers. It is very difficult to discern exactly why they will have done this but the functional answer is: they have done this because they needed earnings to increase. In this case, it looks like the difference between a 15% and a 20% beat (I believe the change was made in January, I think they had quite a big beat in that quarter after some flattish momentum, middle of the year very strong anyway).
  - slownews45 5y ago
    Umm, intel flatlined basically on chip development. Chip shortages also hit the sector. And the long tail has gotten to point where folks keep systems in workloads even if "better" solutions exist - because existing systems are working fine. 4 years is not unreasonable (now). Once chip shortages ease and development picks up again at intel useful life may decrease again.
  - slo_news 5y ago
    But sometimes it does. I worked in audit for several years and the auditors will challenge any estimate made by management that the auditors deem material to the financial statement readers. And the useful life of how long computers last is an estimate by management. While at most companies useful lives might not be an estimate worth looking at, Alphabet has enough computers for this to be a material estimate (likely due to GCP) that is subject to management bias. The public auditors are there to ensure (to a reasonable extent) that management can't manage earnings by changing estimates like this.
bpodgursky 5y ago
This is probably less exciting than you imagine. It's hard to estimate HDD and CPU failure rates 5 years into the future, especially on new-generation hardware. 3 years was an informed guess.
greenail 5y ago
AFAIK it comes down to some degree to a tradeoff between the projected density and power efficiency gains of newer gear vs the cost savings of keeping older gear around and in the racks. Some datacenters simply can't support higher density due to power constraints and if the newer gear isn't offsetting the capital costs by pulling in more revenue then you might as well keep the old gear around longer.
- jeffbee 5y ago
  I don't know why people always mention density. Google isn't constrained on floor space. If they need another 100 acres of Iowa for their servers to stand on, they'll just buy that for a comparative $0 cost. Density is an issue for people who are still renting from Equinix.
  - greenail 5y ago
    the substation has a limit, so do all of the power distribution systems.
    - jeffbee 5y ago
      Sure but that's the opposite trend from density, is it not? A dual 64-core EPYC Rome machine - and we know they are using these for web search, from published research papers - has 400W of CPU TDP alone, which is a more than double the per-host TDP of a Westmere EP (as an example of a retired platform). So if anything there is more empty space in the racks in 2021 than there would have been in 2011, based on the design power load of existing buildings.
      - greenail 5y ago
        The cycle for servers is 3 years, the cycle for the building is 30 years. What I'm trying to say is that if you can deliver more widgets per watt per sq ft by replacing the server than you go for that. If you can't why not keep the already depreciated server operating for another year. The incremental costs are mostly the power (and ac) plus the opportunity cost of putting a more efficient server in a year later. If it fails you just put in the newer server. If it doesn't you may save a bit.
mattlondon 5y ago
Probably because intel chips feel like they have stagnated somewhat performance-wise while they milked the market IMHO. Why upgrade to the latest intel CPU when you get a paltry 100Mhz clock increase for the bother? May as well wait another year if 3 year old CPUs are pretty much the same performance as current models (give it take a few %)
jeffbee 5y ago
Uh, both 3 and 4 are radical understatements of the economic lives of servers. They are TCO+ for 10 years in reality. For a company like Google or any other company designing their own equipment the usual reason to retire equipment (end of warranty) does not exist. There are still Sandy Bridge CPUs available in GCE. Sandy Bridge came out in Q2 2011.
ChuckMcM 5y ago
No, computers typically have a service life of 5 years, but depreciating them faster gives you more loss to to write down against your income. What you are seeing though is the trick they used to make 'net income' pop. Let's say they had a million servers and a server costs $5K each. When depreciated at 3 years that is $1.6 billion dollars a year of depreciation. But you stretch that out to 4 years and now its only $1.25 billion dollars in appreciation. Since depreciation is subtracted from revenue you just "bumped" up your revenue by $350M and you didn't have to do anything at all (except change how you treated your assets). That they had to resort to this level of shenanigan to get their revenue numbers up is interesting to me.
- gradys 5y ago
  > Since depreciation is subtracted from revenue you just "bumped" up your revenue by ... No, you bumped your net income. Revenue is unaffected by the depreciation change. Revenue is up 61%. Also, nobody would be fooled by an accounting trick like this. Analysts routinely compute EBITA, earnings before interest, taxes, and depreciation, exactly for this reason - filtering out the more purely financial/virtual expenses that are less informative for understanding the core business. They likely did it because you're required to report things like depreciation in a way that reflects reality. There could be tax implications for instance, since you can count depreciation expenses against your taxable earnings (though often companies maintain separate depreciation accounting for financial reporting vs taxes due to the different rules for each).
  - ChuckMcM 5y ago
    Fair, it bumps net income not revenue. When I was younger, I took an accounting class from Santa Clara University (it was part of their Executive MBA program although I did not get an MBA) because I knew that I'd like to start my own company some day and needed to know how accounting worked. > Also, nobody would be fooled by an accounting trick like this. The course specifically covered "accounting tricks" that would create the most favorable impression in the eyes of the public and in the eyes of the analyst community. Playing around with depreciation rates was one of them. The assertion in class (which I've never had a chance to test) was that the IRS rarely, if ever, cared about your depreciation tables. That was damn close to 20 years ago so it most certainly could be different now, and I am not a tax accountant. If it has changed since then I am sure someone who has taken the class will correct me. I don't think it "fools" analysts, but I suspect they might trade on an understanding that retail investors might have a different take on what was reported than what the analysts read.
green-eclipse 5y ago
This was announced back in July: https://www.convergedigest.com/2021/07/google-extends-depreciation-lifecycle.html https://www.convergedigest.com/2021/07/google-extends-deprec...

by easton 5y ago

kyrra 5y ago

Googler, opinions are my own. I know nothing about this. I wonder if this was Google's response to the chip shortage? Maybe the cost/benefit (and just sheer inability to get certain parts) made them hold on to hardware for longer.

DennisAleynikov 5y ago
That would be an interesting adjustment
bduerst 5y ago
Could also be the fact that Google has built a ton of new data centers in the last two years. Newer hardware, infra, utilization, etc. could lead to better quality/lifespans.

sidibe 5y ago

I don't think the workloads would have much to do with this, probably saw that they were having less failures than anticipated or memory/chip shortages have them keeping equipment around longer than they planned.

jdavis703 5y ago

My assumption is the hardware was more reliable than expected. If it was about slowing workload growth one might expect them to switch to a usage-based (instead of time-based) depreciation schedule.

DennisAleynikov 5y ago

Seemingly the latter. Less benefit between generations and more stress resistance in hardware compared to earlier data center practice

cj 5y ago

I wouldn't be surprised if the decision was driven by Google's army of accountants. Extending the life of an asset would affect the depreciation schedule, which would affect when depreciation is recognized / recorded on their books in a given tax year.

slo_news 5y ago

While I'm unsure of the the reason the computers are lasting longer, I would bet this observation was originally triggered from the public company accounting firm that audits Alphabet's financials. The accounting firm likely asked Alphabet to prove out why they used a 3 year estimate and found out 4 years was more accurate (FYI 3 years is a standard useful life for computers for most companies - sometimes even shorter).

hogFeast 5y ago
It doesn't work this way. If Alphabet wanted to prove they lasted three years, they would have found a way to prove it. Auditors don't go around challenging on things like this. In particular, they don't go around challenging companies on depreciation because that is a key lever that management can pull if they need to hit numbers. It is very difficult to discern exactly why they will have done this but the functional answer is: they have done this because they needed earnings to increase. In this case, it looks like the difference between a 15% and a 20% beat (I believe the change was made in January, I think they had quite a big beat in that quarter after some flattish momentum, middle of the year very strong anyway).
- slownews45 5y ago
  Umm, intel flatlined basically on chip development. Chip shortages also hit the sector. And the long tail has gotten to point where folks keep systems in workloads even if "better" solutions exist - because existing systems are working fine. 4 years is not unreasonable (now). Once chip shortages ease and development picks up again at intel useful life may decrease again.
- slo_news 5y ago
  But sometimes it does. I worked in audit for several years and the auditors will challenge any estimate made by management that the auditors deem material to the financial statement readers. And the useful life of how long computers last is an estimate by management. While at most companies useful lives might not be an estimate worth looking at, Alphabet has enough computers for this to be a material estimate (likely due to GCP) that is subject to management bias. The public auditors are there to ensure (to a reasonable extent) that management can't manage earnings by changing estimates like this.

bpodgursky 5y ago

This is probably less exciting than you imagine. It's hard to estimate HDD and CPU failure rates 5 years into the future, especially on new-generation hardware. 3 years was an informed guess.

greenail 5y ago

AFAIK it comes down to some degree to a tradeoff between the projected density and power efficiency gains of newer gear vs the cost savings of keeping older gear around and in the racks. Some datacenters simply can't support higher density due to power constraints and if the newer gear isn't offsetting the capital costs by pulling in more revenue then you might as well keep the old gear around longer.

jeffbee 5y ago
I don't know why people always mention density. Google isn't constrained on floor space. If they need another 100 acres of Iowa for their servers to stand on, they'll just buy that for a comparative $0 cost. Density is an issue for people who are still renting from Equinix.
- greenail 5y ago
  the substation has a limit, so do all of the power distribution systems.
  - jeffbee 5y ago
    Sure but that's the opposite trend from density, is it not? A dual 64-core EPYC Rome machine - and we know they are using these for web search, from published research papers - has 400W of CPU TDP alone, which is a more than double the per-host TDP of a Westmere EP (as an example of a retired platform). So if anything there is more empty space in the racks in 2021 than there would have been in 2011, based on the design power load of existing buildings.
    - greenail 5y ago
      The cycle for servers is 3 years, the cycle for the building is 30 years. What I'm trying to say is that if you can deliver more widgets per watt per sq ft by replacing the server than you go for that. If you can't why not keep the already depreciated server operating for another year. The incremental costs are mostly the power (and ac) plus the opportunity cost of putting a more efficient server in a year later. If it fails you just put in the newer server. If it doesn't you may save a bit.

mattlondon 5y ago

Probably because intel chips feel like they have stagnated somewhat performance-wise while they milked the market IMHO. Why upgrade to the latest intel CPU when you get a paltry 100Mhz clock increase for the bother? May as well wait another year if 3 year old CPUs are pretty much the same performance as current models (give it take a few %)

jeffbee 5y ago

Uh, both 3 and 4 are radical understatements of the economic lives of servers. They are TCO+ for 10 years in reality. For a company like Google or any other company designing their own equipment the usual reason to retire equipment (end of warranty) does not exist. There are still Sandy Bridge CPUs available in GCE. Sandy Bridge came out in Q2 2011.

ChuckMcM 5y ago

No, computers typically have a service life of 5 years, but depreciating them faster gives you more loss to to write down against your income. What you are seeing though is the trick they used to make 'net income' pop. Let's say they had a million servers and a server costs $5K each. When depreciated at 3 years that is $1.6 billion dollars a year of depreciation. But you stretch that out to 4 years and now its only $1.25 billion dollars in appreciation. Since depreciation is subtracted from revenue you just "bumped" up your revenue by $350M and you didn't have to do anything at all (except change how you treated your assets). That they had to resort to this level of shenanigan to get their revenue numbers up is interesting to me.

gradys 5y ago
> Since depreciation is subtracted from revenue you just "bumped" up your revenue by ... No, you bumped your net income. Revenue is unaffected by the depreciation change. Revenue is up 61%. Also, nobody would be fooled by an accounting trick like this. Analysts routinely compute EBITA, earnings before interest, taxes, and depreciation, exactly for this reason - filtering out the more purely financial/virtual expenses that are less informative for understanding the core business. They likely did it because you're required to report things like depreciation in a way that reflects reality. There could be tax implications for instance, since you can count depreciation expenses against your taxable earnings (though often companies maintain separate depreciation accounting for financial reporting vs taxes due to the different rules for each).
- ChuckMcM 5y ago
  Fair, it bumps net income not revenue. When I was younger, I took an accounting class from Santa Clara University (it was part of their Executive MBA program although I did not get an MBA) because I knew that I'd like to start my own company some day and needed to know how accounting worked. > Also, nobody would be fooled by an accounting trick like this. The course specifically covered "accounting tricks" that would create the most favorable impression in the eyes of the public and in the eyes of the analyst community. Playing around with depreciation rates was one of them. The assertion in class (which I've never had a chance to test) was that the IRS rarely, if ever, cared about your depreciation tables. That was damn close to 20 years ago so it most certainly could be different now, and I am not a tax accountant. If it has changed since then I am sure someone who has taken the class will correct me. I don't think it "fools" analysts, but I suspect they might trade on an understanding that retail investors might have a different take on what was reported than what the analysts read.

green-eclipse 5y ago

This was announced back in July: https://www.convergedigest.com/2021/07/google-extends-depreciation-lifecycle.html https://www.convergedigest.com/2021/07/google-extends-deprec...