No, I think that your environment is quite common.

Allot of organisations see ongoing IT costs as being fairly limited, usually just to internet/network payments and IT staff. Budgets for repairing hardware are often forgotten and when hardware breaks is when it's replacement is budgeted for.

The bigger organisations can accept the idea of an equipment life cycle as this allows their accountants to do "deprecation" and devaluing of assets. They love the cloud because they totally understand that a server costing $200,000 is deprecated by 25% each year. They hate the idea that an amount would hit a budget randomly so a cloud service that can be at a fixed price per annum without fluctuation works fine for them.

I doubt if all universities are in the same predicament as you. Allot are grant funded and Grants like "big one off costs up front". In another organisation, it was easier to buy a new 4WD then get funding to do a grease and oil change. No one wants to fund maintenance because maintenance is not attractive to grant funds. The attitude towards maintenance is that it should be self sufficient, which often does not match the reality.

Universities are often a bottled society. They are insulated from the happenings of the modern world but not insulated from funding cuts. Hence they will react differently because a different person and mindset is in university accounting then in the real world.

All in all the question you are asking is should you buy a Cloud service or a physical server. That will depend if you can afford the initial cost of the original server and if you believe that the life of the server will give a good Return on your investment. Different accounting cultures will have different values.

Research organisations (ie, basically grant-funded organisations) do definitely have problems with accounting features which (heavily) favour CapEx over OpEx.

Yet, ironically, in a big picture view OpEx is sometimes more appropriate for a grant-funded piece of fixed-term research than CapEx: instead of buying Their Very Own Expensive Machine (tm) at the start of the research, which ends up being underutilised and having the research finish with some amount of that cost not depreciated, the grant money can fund N months of OpEx for "cloud" computing (within-organisation centralised computing or otherwise), at less cost to the grantees. The fact that currently seems unattractive is (IMHO) a failure of accounting. (Particularly since even now continuing to run a 5-7 year old, generic, machine is often an inefficient use of money compared with buying/running a more modern one, from a total-cost-of-use point of view.)

As a counter point, the same accounting that has problems paying OpEx for computing power appears not to have a problem paying OpEx for the electricity to continue to use the computer purchased. I imagine the answer to this is "but it's another budget category". However that just reinforces my point that it's the means of accounting which is causing this problem, not an inefficient use of money.

Finally, there's clearly a business opportunity here. For instance, in the telecommunications industry, there's a desire to own (long haul) fibre rather than pay monthly telecommunications charges. But it's very expensive to install your own. So there's a system of buying multi-year (often 10-20 year) Rights Of Use to some portion of an installed fibre bundle. For a single, up front, CapEx, payment. It's great for the organisation installing the fibre (helps solve their cash flow issues) and gives the purchaser a fixed-for-N-years cost. But it also turns what would be OpEx (eg, fibre circuits) into CapEx.

By analogy, some of the large cloud companies, for instance, could sell, say, 3/5/10 year contracts which for some CapEx (paid up front) entitled the organisation to, say, N core-hours and M storage. Where N core-hours was, for instance, C cores * hours-in-time-period. Presumably with some discount for "bulk purchase" to make it seem more attractive than buying your own.

That this hasn't happened yet is, IMHO, another failure of accounting. The whole "cloud" computing accounting/payment model seems to have been designed by accident, at present: few organisations want $N,000 credit card bills for core services. It's just that some of them are willing to tolerate it for the flexibility offered (eg, on-demand extra power for bursts, multiple locations, etc).

Ewen

Having spent the better part of a decade in a research university, you're nit alone.

I know Amazon allows up-front purchases of instances (they call these "reserved instances"), which might make funding agencies more amenable to cloud resources. I agree with the previous caller who said this is a failure of accounting.