I'm not sure if that's something label as 100% of cause of these number, but it is certainly part of the picture.
Here's a release that has more long term data:
http://jonpeddie.com/press-releases...other-sizeable-step-forward-in-q409-says-jon/
Notice Q3 '08 is at 854K units, by the time the recession was in full swing, that number dropped to around 600 units for several quarters. By Q4 '09 it was back to ~717K units.
So here's a smattering of sales figures over that last 4-5 years:
Q4 09: 717K
Q2 10: 795K
Q3 10: 850K
Q4 10: 903K
Q1 11: 860K
Q2 11: 907K
Q3 11: 1.02M
Q4 11: 998K
Q1 12: 918K
Q2 12: 883K
Q3 12: 932K
Q4 12: 934K
So, I think there is some validity to your idea of companies catching up after the recession, but it happened in Q3/Q4 2011, not today. After sliding from that post-recession catch-up. Sales have climbed back up to near record highs. Remember, the 934K mark represents about a 10% improvement to what was seen pre-recession. The Dow and S&P haven't climbed 10% from pre-recession peaks. While certainly not the best measure for comparison and kinda quick and dirty. The argument that workstation sales are outperforming the general market certainly has a leg to stand on.
There are certainly peaks and valleys involved, and the result of various causalities, such as Intel & vendor product cycles and financial cycles (quarters/half/annual), but lets examine this from a growth POV, which is what we've been discussing specifically.
Lets take a look at 2012. Except for Q2, it was fairly steady. Not bad, but not really any growth to speak of (appears that immediate need issues have stabilized, so no sudden upshot in sales to fulfill overdue need).
Now consider that given the Recession's impact as well as a longer term outlook, we need to go back and get a larger chunk of data, say from 2006 up until the Recession registered in the sales figures, which would reflect pre-Recession conditions.
From JPR (I'll stick with this site since it's public and you seem to like their data
):
Workstation market experiences record quarter in Q4'07, reports Jon Peddie Research
Q3CY06 618.9 (K units)
Q4CY06 686.7
Q1CY07 694.9
Q2CY07 719.9
Q3CY07 762.6
Q4CY07 847.9
Now if you notice, and the article's author has done the math, you'll notice a nice and steady growth rate between Q2 2006 to Q4 2007.
The markets been on a roll, posting yearly returns in 05 and 06 of 22.1% and 21.2%, respectively.
For the year [2007], over 3 million workstations shipped, posting a gain of 20.5% over 2006.
Then the Recession hit, and scarred the crap out of everyone. Thus causing most companies to sit on their cash to wait it out and see what was going to happen (stalled hiring, if not layoffs due to reduced if any growth in the company in question, and system lifecycles were extended). So replacement system purchases and any growth that may have occurred (hiring new employees), was staggered out between 2009 & 2012 as a result of the machine age and usage in question (i.e. system that should have been replaced in 2008 at the latest, wasn't replaced until sometime in 2009, and say a machine slated for a 2009/10/11 replacement date was pushed to 2012).
So both the recession and delayed replacement systems purchase dates had a significant effect on the overall sales figures.
Where I see a major discrepancy however when looking at the data, is in the growth rates Y-o-Y before and after the Recession hit, and any recovery began. The previous 21 - 22% YoY growth rate that was occurring before the Recession has vanished (3.6% uptick for 2009, based on the link you provided).
Now consider the technical advances (i.e. elimination of FSB, ..., architecture improvements that came with Nehalem and later systems), and the overall loss of jobs (layoffs greater than any subsequent hiring in general in the sorts of fields that require workstations from what I'm seeing), it's no wonder there's not much if any genuine growth in workstation sales.
Where I see future (near future, not 10+ years), is in the ever increasing core counts (CPU & GPU), and reduced costs of high speed networking (i.e. cheaper fiber solutions, say networking capabilities added to an optical Thunderbolt interconnect), will combine further where a single employee can do the work of more than one employee based on human productivity figures from 2005/6 (whether independent workstations or on some form of cluster that's cheaper to implement). Thus less people doing more work on fewer workstations in the near future.
Please understand, I can't but help also look at both employment figures and other technologies (directly related to workstation users) when thinking about the workstation market (or other markets for that matter).
Also, think of PC sale during the same time period. Here's some idea:
http://money.cnn.com/2013/04/10/technology/pc-sales/index.html
Granted that starts to bring in issues from PCs being cannibalized by tablets and laptops. But on the other hand, weren't PCs supposed to be eating into Workstations?
Don't underestimate the tablets and laptops. Desktops have been losing ground steadily due to lower cost laptops and a closer parity in performance for similar funds.
Desktops still have their market, which is in developing nations. But those figures wouldn't be reflected in JPR's data, as they're from non-US system vendors.
In my field (genomics) there has certainly been a good deal of expansion requiring purchases of workstations.
There will be a few exceptions here and there to be sure. But it's not enough to put the growth rate back to what it was in 2005/6 (bit over 20%).
Sorry, BW users? Can you clarify?
Bandwidth.
But that $100 probably represents on the order of .1% of the cost of the employee/year using the computer (and presumably the computer should last at least 3 years, right?). So, its not hard to justify from cost/benefit analysis. And if it is 1K systems, that's 1K employees. $10K for company with 1K employees isn't a huge number. Maybe if its all in single month, that could be hard to swallow from a bill paying stand point. But spread out over a year, it would hopefully be a drop in the bucket in total expenses.
A lot of smaller entities are cash strapped vs. say Fortune 500 or larger. And the smaller they are, usually the worse cash flow is, so every little bit counts, and without a proper TCO analysis, the wrong choice may be made (consumer models used instead of workstations).
For larger entities, it's more about cutting every little bit to increase the gross margin instead of their ability to purchase equipment.
If the computational work doesn't get more demanding at the same pace as the computers get faster, then yes, PCs will eat into workstations more and more. But that's a two headed monster, because that also means computational work being off loaded to clusters is now easier to put on a workstation. So, I think these forces balance out in general, which is what will allow for workstations to maintain moderate and steady growth that roughly matches the economy as a whole.
This is basically how I see it (also considering other factors mentioned, such as fewer employees working in these fields relative to the workload produced).
And although workstations can be used in clusters, rack mount systems will likely rule here due to thermal requirements, physical space, ... (same Xeon processors and so on, but the different case will cause it to be classified as a server).
I also wonder if a few years in the future PCs will just be the odd man out. Either you want a laptop or you need a workstation. In effect the PC will get squeezed from both sides until its for the large part just gone. And given PC's sale figures, I'm more convinced that will happen than the workstation market will shrink.
I don't see things going this way.
Especially when you consider the global market (i.e. China and India for example, that are still buying desktops at a far greater rate than the West), the business market continues to buy desktops (globally), and overlap.
What I mean in regard to overlap, is that Intel has continued to reduce power consumption, where it would be possible to use the same CPUID's for either a laptop or desktop, thus reducing the total CPUID's necessary to cover the various segments. So instead of high power variants for mainstream desktops and low power variants for laptops, one part would do. Thus giving them a greater economy of scale in this regard.
The high power variants, would be suitable to the high-end consumer (enthusiast) parts as well as the Xeons (servers & workstations). Just leave ECC enabled for the Xeons, and cut the traces on the enthusiast variants as they currently do.
