Hacker News Reader

Comment by isusmelj | Hacker News Reader

Can anyone explain why servers are now more useful for longer than they used to be? I feel the world is spinning faster, and new tech comes out faster. Especially with faster networking and more compute.

In the document they mention this:

In January 2023, we completed an assessment of the useful lives of our servers and network equipment and adjusted the estimated useful life of our servers from four years to six years and the estimated useful life of certain network equipment from five years to six years. This change in accounting estimate was effective beginning in fiscal year 2023, and the effect was a reduction in depreciation expense of $977 million and $2.9 billion and an increase in net income of $761 million and $2.3 billion, or $0.06 and $0.18 per basic and $0.06 and $0.18 per diluted share for the three and nine months ended September 30, 2023, respectively.

View on HN

This is armchair analytics so take me with a grain of salt, but:

- server/cloud computing is largely all about performance per watt

- the latest couple gens intel seem to be struggling to innovate and are just getting better benchmarks by cranking up the power consumption. even AMD's latest release is along the lines of "x% faster per core, XX% more power draw".

So I wonder if this is percolating out into the server/cloud space where people are just watching the chipmakers trade off efficiency for benchmarks and are unimpressed.

> I feel the world is spinning faster, and new tech comes out faster.

Isn't it because it's precisely the opposite? Moore's law stopped a while ago, at least that's what a lot of people think.

So much of the new tech now seems to be around GPU's and LLM's, but that doesn't affect the Google Search or YouTube or Docs servers. So they just keep humming along as usual, with a totally different set of new-fancy-tech servers spun up for the new fancy stuff.

New tech in the server space comes out slower than it used to. Moore's Law has almost dried up.

Fifteen years ago, you could assume that you would have 50% larger hard drives and significantly more FLOPS/watt with each year's new systems. Today, that trend has almost entirely stalled.

In short: servers used to have short life-cycles not because they no longer worked, but because they had so much higher operating expenses (electricity, datacenter floor-space) than newer models. That trend no longer holds; the more relevant question today is "how long will the servers work".

My guess is that server farms are currently used at full capacity. So that while a few years ago the thought was to upgrade existing data centers to more efficient servers, now there is more data centers being built while slightly less efficient servers are left running where they are. This is the accounting change to reflect that: depreciate slower to reflect longer in-practice life.

Also "data center electrical power / compute" (which reflects both cooling and server power supply cost) has held relatively steady for a while which makes it easier to do that.

They aren't; this is an accounting change.

There are many tasks for which speed is not an absolute requirement.

I bet there are lots of places that parallelisation makes more computers within the ballpark of acceptable performance perfectly cromulent to keep around.

Heck, my 2015 mac is a perfectly serviceable media player when I watch sport.

None of that has anything to do with the accounting change that they made.

With repercussions. If they’re estimating a lifespan extension and later go back to the original estimate, they’ll have huge one-off depreciation charges, no?

No, it's an accounting change because they are.

My hot take is that servers could potentially last longer than 6 years so it is all really just accounting. Or at least is was never as short as 3 years.

When we were in a low/zero interest rate environment - there was a growth at all costs mentality among investors (and therefore management). Since investors didn't give any bonus points if a company was able to operate profitably - might as well depreciate to the max allowable by accounting law and not show a profit. In fact - not showing a profit was actually seen a positive features/badge of honor for a stock.

Now that interest rates have gone up, investors care about ability to generate current cashflow/net income and are prioritizing rational PE ratios - depreciating expenses over a longer time horizon (closer to reality) makes sense.

Reply on news.ycombinator.com